A major strategic divide is emerging over the future of agentic commerce. Shopify is opening its merchant ecosystem to Meta’s new AI shopping agent, Muse, while Amazon has taken the opposite approach and blocked the same agent from shopping on Amazon.com. The contrast reveals two very different visions of how AI should interact with online retail. Shopify appears willing to let third-party agents become a new commerce channel, while Amazon is defending tighter control over its marketplace, customer experience and commercial infrastructure.

On one side, Shopify is working with Meta so Muse can help users discover products and complete purchases across Shopify-powered stores using Shop Pay. On the other, Amazon says Meta did not obtain permission for Muse to access its marketplace and has blocked the agent from shopping on behalf of customers. The dispute raises a wider question for the entire e-commerce industry: should AI agents be welcomed as a new layer of distribution and checkout, or treated as external intermediaries that could weaken retailer control over customers, data and monetization?

Shopify embraces Meta’s Muse as an AI commerce channel

Shopify has chosen the more open approach. The company is partnering with Meta to integrate Shop Pay agentic checkout into Muse, allowing the AI assistant to facilitate purchases from merchants operating on Shopify. Shopify had already started preparing for this shift earlier in September by adding Meta as an AI channel inside Shopify Admin. Merchants can manage whether product catalogs are shared with Meta, control direct checkout access and monitor performance from within Shopify’s agentic-commerce settings.

This is significant because Shopify is not merely allowing Muse to browse merchant websites. It is connecting Meta to structured commerce infrastructure, giving the AI cleaner access to product data, availability and checkout functionality. That reduces the need for the agent to interpret conventional websites designed for humans. Instead, Shopify can act as the commerce layer beneath the AI interface, providing the product and transaction infrastructure while Meta controls the consumer-facing experience.

Shop Pay becomes the transaction layer

The partnership also highlights the strategic value of Shop Pay. Muse users will be able to complete purchases through Shopify’s accelerated checkout infrastructure across participating stores. That means the customer may begin and remain inside Meta’s AI environment while the transaction itself still flows through Shopify’s commerce stack. For Shopify, the loss of the traditional storefront as the primary interface does not necessarily mean losing the economic value of the transaction.

This is a critical distinction in the emerging agentic-commerce model. Shopify does not primarily depend on consumers visiting Shopify.com, because its business is built around providing infrastructure to merchants wherever commerce happens. If discovery increasingly moves into AI assistants, Shopify can still capture value through product feeds, checkout, payments and merchant operations. Its strategy therefore appears less focused on protecting the storefront and more focused on becoming indispensable infrastructure underneath AI-driven commerce.

Amazon takes the opposite approach

Amazon has chosen a far more restrictive position. The company blocked Muse from shopping on Amazon.com after Meta reportedly declined Amazon’s request to exclude its marketplace from the agent’s shopping capabilities. Users attempting to shop on Amazon through Muse began seeing a message stating that continued access by an unauthorized AI agent violated Amazon’s Conditions of Use. Amazon says the issue centers on authorization, transparency and how third-party agents interact with its platform.

According to Amazon, Meta did not provide advance notice that Muse would shop on Amazon.com, the agent does not adequately identify itself while browsing, and the company has concerns around customer credentials and account access. Amazon argues that third-party applications making purchases on behalf of users should operate transparently and respect whether a service provider wants to participate. The company’s position is that consumer permission alone is not enough; the merchant platform should also be able to decide whether an external agent can access and transact through its systems.

Meta disputes the credential-risk concern

Meta has said Muse does not directly see users’ passwords or payment credentials. According to the company, credentials are stored securely and can be used by the AI without being exposed to the model itself. Muse operates inside a virtual machine and can use a browser to interact with services when a public API is not available. Meta also says the agent seeks confirmation before taking sensitive actions such as sending messages or completing purchases.

This difference in interpretation is central to the dispute. Meta views the consumer as authorizing the agent to act on their behalf, while Amazon argues that the platform being accessed also needs to consent. That tension could become one of the most important legal and commercial questions in agentic commerce. It effectively asks whether an AI agent should be treated like a user-controlled browser, an automated bot, a commercial intermediary, or something entirely new.

Amazon wants AI agents to identify themselves

Amazon’s position is also closely linked to transparency and identity. The company argues that an external AI agent entering customer accounts, browsing product pages and initiating transactions should clearly identify itself to the retailer. This reflects a broader issue emerging across e-commerce and payments: merchants increasingly want to know which agent is acting, who authorized it, what permissions it has, whether it is operating within those limits and who is responsible if something goes wrong.

These are no longer theoretical questions. Payment networks including Mastercard and Visa are already developing identification and authorization layers specifically for AI agents. Amazon is effectively applying the same logic at the marketplace level. If agentic commerce is to scale, retailers may demand a standardized way to recognize trusted agents, verify user intent and establish accountability before allowing automated systems to interact deeply with their commerce infrastructure.

Shopify sees AI agents as another distribution channel

Shopify’s strategy appears fundamentally different. Rather than requiring consumers to begin every transaction within a Shopify-controlled environment, the company is treating AI assistants as another sales and distribution channel. This resembles the way Shopify already connects merchants with Google, social networks, marketplaces, mobile commerce and point-of-sale systems. Agentic platforms now become another route through which merchants can reach potential customers.

That approach fits Shopify’s architecture particularly well. Its value lies in powering merchants across multiple channels rather than owning a single consumer destination. If an AI platform becomes the place where shoppers discover and compare products, Shopify can still remain the system handling structured catalog data, merchant operations, payments and checkout. In that model, losing control of the interface does not necessarily mean losing control of the transaction infrastructure.

Amazon has more to lose from external agents

Amazon’s incentives are very different because its marketplace is not simply a transaction engine. The company controls product discovery, search, advertising, reviews, recommendations, checkout, fulfillment and customer service inside one highly integrated environment. External AI agents could bypass several of those layers by searching, comparing and selecting products on behalf of users without presenting Amazon’s traditional marketplace interface.

This matters because Amazon has built a huge advertising business around product discovery. Sponsored placements, search positioning and marketplace visibility are all tied to shoppers navigating Amazon directly. If an AI agent evaluates products based mainly on price, reviews, availability and delivery terms, some of the commercial value of those interfaces could be reduced. Amazon therefore has more strategic incentive than Shopify to preserve control over how shoppers discover and select products.

The fight is about who controls product discovery

This is why the Shopify-Amazon split matters beyond Muse itself. Agentic commerce is changing where product discovery happens. Today, shoppers typically search, browse, compare products, open listings, add items to a cart and complete checkout. An AI agent could compress that process into a single instruction such as: “Find me the best stroller under €500, with strong reviews and delivery before Friday.”

The agent could then search across multiple merchants, evaluate the available options and complete the transaction according to the user’s preferences. If that behavior becomes common, the center of gravity in e-commerce may move away from individual marketplace interfaces and toward AI platforms. The strategic question then becomes less about who owns the product listing and more about who owns the interface between shopper and merchant.

Shopify is betting on infrastructure, Amazon on ecosystem control

The two companies are effectively making different bets on the next phase of e-commerce. Shopify is betting that infrastructure remains valuable even if another company controls the consumer interface. Its role can still include catalog data, merchant operations, payments, checkout and order management. Amazon appears more determined to preserve control over both the infrastructure and the customer-facing environment, including search, advertising, recommendations and transaction flow.

Neither position is necessarily permanent. Amazon could eventually negotiate access agreements with selected AI platforms, while Shopify could impose stricter controls if agentic commerce creates problems for merchants. But the current split exposes two competing models. One is open and infrastructure-led, while the other is more closed and ecosystem-driven. Those models could shape how retailers interact with AI agents over the next several years.

Amazon is already challenging other AI shopping agents

Muse is not the first AI shopping agent Amazon has opposed. The company has also attempted to restrict access by agents from Perplexity, Google and OpenAI. Its dispute with Perplexity has already moved into the courts, where Amazon initially secured a preliminary injunction before a US appeals court later rejected part of its argument under federal anti-hacking law. Contractual claims based on terms of service remain a separate issue.

That legal background helps explain the language Amazon is now using with Muse. The company is framing the dispute around authorization and contractual access rather than purely around cybersecurity. This suggests Amazon is trying to establish a broader principle: AI agents should not assume unrestricted access to its commercial environment simply because an individual user authorizes them to act.

Amazon itself uses agentic shopping technology

Amazon is not opposed to AI shopping agents in principle. The company has developed its own agentic-commerce capabilities, including Buy for Me, which can help customers purchase products from external merchant sites, and Alexa for Shopping, which can research products and provide recommendations. Amazon argues that the difference lies in transparency and merchant consent.

According to Amazon, Buy for Me identifies itself when interacting with external merchants and allows businesses to opt out. That distinction reinforces the company’s central argument: the problem is not autonomous shopping itself, but how external agents enter and use another company’s commerce environment. In other words, Amazon supports agentic commerce when the rules of access are negotiated and visible to all participants.

Muse is gaining traction quickly

The dispute matters partly because Muse is already attracting significant attention. Meta launched the personal AI agent on September 8, positioning it as a system capable of handling multi-step tasks including shopping, appointment booking, email management and other everyday activities. Shortly after launch, Muse climbed rapidly through Apple’s US App Store rankings, indicating strong early consumer interest.

Growing adoption increases the urgency for retailers and commerce platforms to decide how they want to interact with these agents. For Shopify merchants, Meta represents a potential new source of product discovery and checkout. For Amazon, Muse currently represents an unauthorized intermediary operating inside a commercial environment Amazon has spent decades building and monetizing.

Agentic commerce could disrupt e-commerce advertising

Advertising may become one of the areas most affected by this shift. Traditional marketplaces monetize shopper attention by selling visibility within search results and product pages. Retailers pay to move products higher in rankings or increase exposure across marketplace interfaces. AI agents may operate differently, especially if they prioritize objective criteria such as price, delivery speed, return policies and product specifications.

If consumers increasingly delegate purchasing decisions to agents, traditional sponsored placement could become less influential. An AI may not respond to the same visual hierarchy and advertising formats as a human shopper. That creates a potential challenge not only for Amazon but also for Google and other platforms whose commercial models depend heavily on monetizing search and product discovery.

Merchants face their own strategic decision

Retailers will also need to decide how open they want to be to AI agents. Participation could bring significant benefits by creating new channels for product discovery, customer acquisition, conversion and cross-border sales. At the same time, merchants may lose some control over brand presentation, customer relationships, upselling and the overall shopping journey.

A consumer who shops entirely through Muse may never see the merchant’s homepage, marketing funnel or carefully designed product pages. Competition could therefore shift toward machine-readable factors such as price, stock availability, delivery speed, product attributes and returns. That would change how retailers think about optimization and could make structured product data increasingly important.

Structured commerce data becomes more valuable

Shopify’s role demonstrates this shift clearly. When Shopify shares catalog data directly with Meta, Muse does not need to scrape and interpret every merchant website independently. Instead, the agent can access standardized commerce information such as product descriptions, variants, prices, stock levels and checkout capabilities.

That gives Shopify an important strategic position in the agentic ecosystem. The company already sits between merchants and numerous external sales channels, and AI assistants can become another distribution layer connected to that infrastructure. As more shopping moves through agents, the platforms controlling high-quality structured commerce data may become increasingly valuable.

The future could split into open and closed commerce ecosystems

The Shopify-Amazon divide may point toward a broader fragmentation of agentic commerce. Some platforms may operate as open commerce infrastructure, allowing multiple AI agents to access catalogs and transactions through standardized integrations. Others may remain more closed, allowing only approved agents or proprietary AI experiences to participate.

The result could resemble earlier battles over mobile operating systems, app stores and platform APIs. Retailers and marketplaces may have to choose between reach and control, while AI companies may need to negotiate access to the largest commerce ecosystems rather than assuming they can browse and transact freely.

A defining battle over the next e-commerce interface

The dispute between Shopify, Amazon and Meta ultimately comes down to control. Meta wants Muse to become an interface through which consumers can complete everyday tasks, including shopping. Shopify sees an opportunity to make its merchant ecosystem accessible through that interface. Amazon sees an external agent entering a shopping environment it has spent decades building, optimizing and monetizing.

All three positions are commercially rational, but they point toward very different futures for e-commerce. Shopify’s model says the consumer can shop anywhere while Shopify provides the infrastructure underneath. Amazon’s model says the shopping environment itself remains strategically important and should not be accessed by third-party agents without permission. As AI shopping agents become more capable, the next major battle in e-commerce may not be over assortment or delivery speed, but over which AI is allowed to stand between the shopper and the store.

The post Shopify Opens Its Stores to Meta’s Muse While Amazon Blocks the AI Shopping Agent appeared first on Cross-Border Magazine.

SHIP.com has launched two artificial intelligence tools designed to bring agentic commerce into one of the most operationally complex areas of e-commerce: shipping and fulfillment.

The company announced SHIP AI, an AI-powered shipping manager for online sellers, and SHIP MCP, a Model Context Protocol integration that exposes more than 35 shipping capabilities to Claude and other AI applications.

Together, the tools automate tasks such as order analysis, rate comparison, address correction, label preparation, and tracking. They also allow developers to connect AI agents directly to shipping infrastructure. The announcement is significant because much of the recent discussion around agentic commerce has focused on product discovery and payments. SHIP.com is extending the same concept deeper into the physical fulfillment layer.

That means AI agents are moving closer to handling not only what consumers buy and how they pay, but also how those purchases are shipped.

What Is SHIP AI?

SHIP AI is designed to function more like a digital shipping manager than a traditional chatbot. Instead of simply responding to questions, the system can analyze incoming orders and prepare fulfillment work before the seller begins processing shipments.

According to SHIP.com, the system can determine package details, compare shipping options, correct addresses, flag inconsistencies, and prepare multiple orders for shipment simultaneously. Users can also ask the system to update dashboards, run reports and analyze variables such as shipping costs and delivery times.

Joe DiSorbo, founder and CEO of SHIP.com, described the distinction by saying that a general chatbot provides answers, while SHIP AI is designed to perform operational work. The goal is to move sellers away from manually processing every order and toward reviewing work AI has already prepared.

From Manual Processing to AI-Assisted Fulfillment

Shipping remains one of the most repetitive and time-consuming areas of online retail operations. A merchant processing multiple orders may need to verify addresses, calculate package dimensions, compare rates, select carriers, purchase labels and update order information.

SHIP.com says SHIP AI can automate much of that preparation. The company gives the example of a 20-order batch that could take around 25 minutes to process manually. Depending on the level of automation enabled, SHIP.com says the same batch could potentially be prepared in less than a minute with SHIP AI.

That claim is based on the company's own testing and should therefore be treated as a vendor-reported performance example rather than an independent benchmark.

Nevertheless, the broader operational shift matters. The merchant's role shifts from performing each shipping action manually to supervising and approving AI-generated decisions.

Sellers Retain Control Over Financial Decisions

SHIP.com is not presenting SHIP AI as completely autonomous by default. Financial transactions remain under the merchant's control. Purchasing shipping labels, adding insurance or reloading an account balance normally requires seller approval, and the relevant costs are displayed before the transaction is completed.

This human-in-the-loop approach reflects a central challenge facing agentic commerce. Businesses want AI systems to remove repetitive operational work, but financial authority introduces considerably greater risk. SHIP.com's approach lets the AI analyze and prepare decisions while merchants decide how much authority they are willing to delegate.

According to the company, sellers could gradually expand the degree of automation as they become more comfortable with the system.

Early Adoption Suggests Sellers Are Willing to Test AI Shipping

SHIP AI has been available to SHIP.com users since August 11, 2026. The company says that during the first week of beta availability, more than 30% of orders were purchased through SHIP AI.

This figure does not necessarily mean that 30% of all SHIP.com customers have adopted the technology, nor does it demonstrate long-term adoption. However, it provides an early indication that merchants are willing to experiment with AI-driven shipping workflows when those tools are integrated directly into the software they already use.

SHIP is offering the complete AI experience free to users for at least 30 days during the launch period.

SHIP MCP Brings Shipping Capabilities Directly to AI Agents

The second product may ultimately have even larger implications for agentic commerce. SHIP MCP connects SHIP.com's shipping infrastructure to AI applications through Model Context Protocol. MCP is an emerging standard that allows AI systems to connect with external tools, software and data sources.

SHIP.com says its implementation exposes more than 35 shipping capabilities that developers can integrate into AI applications. These capabilities include rating shipments, creating labels, tracking packages, processing order information and managing shipping workflows.

Rather than forcing users to leave an AI environment and manually open separate shipping software, these functions can potentially operate directly inside the AI application.

Claude Users Can Access SHIP.com Shipping Functions

SHIP MCP is available through Claude Desktop CLI and can also be found by searching for SHIP.com within Claude's connector directory. Developers can additionally access the technology directly through SHIP.com when building their own AI applications.

SHIP.com says some early customers have already embedded its capabilities into their own AI applications, enabling orders to be rated, labeled and tracked through those systems. DiSorbo described these integrations as early steps toward fully autonomous agentic commerce.

The distinction matters. SHIP AI primarily focuses on making existing merchant shipping operations more efficient, while SHIP MCP could make shipping itself a machine-accessible service that autonomous AI agents can call.

Agentic Commerce Moves Beyond Checkout

This development reflects a broader shift across e-commerce. Many of the highest-profile agentic-commerce initiatives have concentrated on the early stages of the customer journey. AI agents can increasingly discover products, compare alternatives, recommend purchases and initiate payments. Companies including Stripe, Alipay and major AI providers are already building infrastructure around these interactions.

But purchasing a physical product creates another requirement: the item still needs to leave a warehouse and reach the customer. A genuinely autonomous commerce system eventually needs access to inventory, fulfillment systems, shipping rates, carrier networks, delivery estimates, labels, tracking information, and returns processes.

SHIP.com is attempting to make part of that logistics infrastructure directly accessible to agents. The emerging agentic-commerce stack could therefore begin to resemble Discovery → Decision → Payment → Fulfillment → Delivery, rather than stopping once the transaction is authorized.

The Logistics Layer Could Become Machine-Readable

Agentic commerce also changes the way logistics technology needs to function. Most shipping platforms were originally designed for humans. A user logs into a dashboard, reviews orders, chooses a carrier, generates a label, and manages exceptions through a graphical interface.

AI agents require something different. They need structured, programmatic access to shipping capabilities so they can evaluate different options and perform actions automatically. SHIP MCP represents one approach to solving that problem. Rather than requiring an agent to navigate a human-designed web interface, the system exposes specific shipping actions through standardized tools.

That could eventually allow an AI agent to determine that a particular order needs expedited delivery, one carrier offers a better rate, an address needs correcting, or an alternative service would meet the customer's requested delivery date. The agent could then act on that information within predefined limits.

What This Could Mean for E-Commerce Sellers

For merchants, the immediate benefit is operational efficiency. Shipping teams spend a lot of time on repetitive tasks that follow predictable rules. AI systems are particularly well suited to analyzing those workflows and preparing recommended actions.

But the longer-term implications extend further. As agentic commerce develops, merchants may increasingly operate in environments where software agents communicate directly with logistics platforms.

Instead of a human seller manually selecting a shipping method, an AI system could optimize thousands of shipments based on cost, speed, service level, and delivery probability. Smaller merchants could access capabilities that previously required dedicated logistics teams, while larger businesses could use agents to manage significant order volumes and let human employees focus on exceptions and strategic decisions.

AI-Driven Shipping Still Introduces Risk

Autonomous fulfillment also introduces important limitations. SHIP.com's own terms explicitly warn that AI-generated outputs may be inaccurate, incomplete or outdated and may not always reflect current carrier rules or rates.

Users are therefore responsible for reviewing and approving AI-generated outputs, including shipping labels, addresses, rate selections and customs or regulatory classifications. That is particularly important in cross-border commerce. Incorrect customs information, classifications or addresses can result in delays, additional costs or regulatory problems.

SHIP.com's terms also address automated clients such as AI assistants and autonomous agents. Actions carried out by an authorized automated client can be treated as actions taken by the user. As businesses give agents greater authority, questions of accountability will consequently become increasingly important.

Shipping Platforms Could Become Infrastructure for AI Commerce

The most important aspect of SHIP.com's announcement may not be automated label creation itself. It is the idea that AI systems can call logistics capabilities directly. That turns a shipping platform from a traditional software interface into infrastructure that autonomous commerce applications can use.

The same transformation is already occurring elsewhere in e-commerce. Payment companies are building agent-compatible payment mechanisms. Marketplaces and retailers are exposing product information to AI systems. AI platforms are creating tools that allow consumers to discover and purchase products without leaving a conversational interface. Shipping represents another piece of that puzzle.

The Next Phase of Agentic Commerce Reaches the Warehouse

Agentic commerce has often been presented primarily as a change to the shopping experience. Consumers tell an AI assistant what they want, and the assistant searches for the best option. But commerce does not end when the shopper presses, or no longer presses, the checkout button. Orders still need to be processed, packed, shipped, tracked, and delivered.

SHIP.com's new tools show how AI agents are beginning to move into that operational layer. SHIP AI focuses on automating the work merchants currently do inside shipping software. SHIP MCP goes one step further by allowing external AI applications to interact directly with shipping capabilities.

If this model expands, logistics platforms could increasingly become invisible infrastructure behind AI-driven commerce. The customer may interact with an AI assistant, the merchant may approve only exceptions, and software agents could manage much of what happens between the purchase and the doorstep.

For e-commerce, that would mark an important evolution in agentic commerce: from automating the transaction to automating its fulfillment as well.

The post SHIP.com Brings Agentic Commerce Into E-Commerce Fulfillment With New AI Shipping Tools appeared first on Cross-Border Magazine.

Stripe believes one of e-commerce's most familiar elements could eventually become obsolete. Will Gaybrick, Stripe’s President of Product and Business, has argued that the traditional checkout page could disappear as artificial intelligence, digital wallets, and agentic commerce increasingly collapse product discovery and payment into a single transaction.

Speaking on the A16z Show, Gaybrick said the checkout page could become unnecessary not only for AI agents, but eventually for human shoppers as well. His argument is straightforward: when payment credentials, identity and authorization are already securely available through tools such as Stripe Link, consumers should increasingly be able to purchase directly from the product or discovery interface rather than navigating several checkout screens.

The prediction offers a glimpse of how Stripe expects online commerce to change as AI agents become increasingly capable of finding products, evaluating options, and completing purchases on behalf of consumers.

From checkout page to instant transaction

The traditional e-commerce journey is familiar:

A customer discovers a product, visits a merchant website, adds it to a cart, proceeds to checkout, enters delivery information, chooses a payment method and confirms the order. Agentic commerce challenges almost every step of that model.

Stripe defines agentic commerce as online shopping in which AI agents can find, compare, and potentially purchase products on users' behalf. The shopper provides an objective — for example, finding a particular product under a certain price with delivery before a specified date — and the agent handles much of the searching and decision-making.

In this environment, the shopper may never need to navigate a conventional checkout page. Instead, product discovery, decision-making, authorization and payment could take place inside an AI assistant, search environment or other digital interface. Stripe explicitly notes that customers using an AI agent may not need to see a checkout page at all.

Why Stripe thinks checkout could disappear

Checkout disappearing doesn't mean the underlying processes vanish. Merchants would still need to calculate taxes, confirm inventory, determine shipping costs, process payments, prevent fraud, manage orders, and handle fulfillment. The difference is that those processes increasingly happen behind the interface.

Instead of presenting each step to a human shopper, commerce infrastructure could communicate directly with an AI agent. Gaybrick has pointed to systems such as Stripe Link and Shopify's Shop Pay as early examples of how stored payment credentials and accelerated checkout can already reduce the number of steps required to purchase online.

Agentic commerce takes that concept considerably further. Rather than simply speeding up a checkout page, AI could make the page unnecessary.

Stripe is already building infrastructure for agentic commerce

Stripe's prediction is closely connected to products the company has been developing over the past year. At Stripe Sessions 2026, the company expanded its Agentic Commerce Suite and announced additional partnerships with companies including Meta and Google. Stripe also introduced new ways for AI agents to pay using Link's agent wallet.

Stripe describes its Agentic Commerce Suite as infrastructure that connects businesses, agents, and buyers. Merchants can make their products available to AI agents while maintaining control over pricing, descriptions, fulfillment and the customer relationship. AI applications can then embed checkout directly inside their own interfaces.

This model represents a fundamental shift in the merchant website's role. Instead of every purchase requiring consumers to reach a retailer's checkout interface, transactions could increasingly be completed wherever the customer discovers the product.

AI agents need their own payment infrastructure

Giving an AI agent purchasing authority creates problems that conventional payment systems were never designed to solve.

An agent may need to:

• Identify the correct product;
• Confirm availability;
• Compare shipping options;
• Understand return policies;
• Operate within a spending limit;
• Prove that the consumer authorized the transaction;
• and complete payment without gaining unrestricted access to the customer's financial credentials.

Stripe has been developing infrastructure specifically for these scenarios.

Its Link wallet for agents can issue a one-time-use card or Shared Payment Token backed by payment methods already stored within Link. According to Stripe, the agent itself does not receive the user's underlying payment credentials. This allows agents to transact while preserving spending controls and payment security.

Stripe has also introduced Issuing for agents and other mechanisms designed to give autonomous software controlled purchasing power.

Machine Payments Protocol removes checkout entirely

The transformation becomes even clearer when commerce takes place between machines rather than between a human and a retailer. In March 2026, Stripe and Tempo introduced the Machine Payments Protocol (MPP), an open standard that lets AI agents transact directly with businesses and other services.

Traditional online payment flows assume that a human will visit a website, select a product, and complete a checkout form. MPP instead allows software agents to interact programmatically with paid services. Stripe says conventional financial infrastructure was built primarily for humans, meaning AI agents often struggle with tasks such as navigating pricing pages, selecting subscription tiers, creating accounts or entering payment details.

Machine-readable payment protocols are designed to remove those obstacles. If adoption grows, some forms of commerce may therefore never generate a visible checkout interface at all.

E-commerce UX could become commerce infrastructure

The implications for retailers could be substantial. For years, merchants have invested heavily in optimizing their checkout pages.

Businesses test:

• Button positioning
• Page layouts
• Payment methods
• Form length
• Guest checkout
• Upselling
• Cart recovery
• Conversion rates.

Agentic commerce could shift part of that optimization away from visible user interfaces and toward APIs, structured data, and machine-readable commercial rules. Stripe advises businesses preparing for agentic commerce to expose structured product information and build checkout flows that can function without a human directly interacting with them.

That means an AI agent needs reliable access to information including:

• Product availability
• Pricing
• Shipping times
• Returns policies
• Taxes
• Merchant identity
• Fulfillment guarantees.

In other words, the checkout experience increasingly becomes an infrastructure problem rather than purely a web-design problem.

Product discovery and checkout could merge

One of the biggest changes could be eliminating the boundary between discovering something and buying it. Today, discovery often happens outside the retailer's website.

Consumers may discover products through:

• Google
• Social media
• Marketplaces
• Influencers
• Comparison services
• Increasingly, AI assistants.

Historically, shoppers had to leave that environment and visit a merchant's checkout. Agentic commerce lets the transaction stay inside the discovery layer.

Stripe has already worked with OpenAI on the Agentic Commerce Protocol, designed to allow businesses to sell through AI interfaces while maintaining control over their transactions. Stripe said the protocol's first real-world implementation supported Instant Checkout inside ChatGPT.

The company is also working with AI and technology platforms including Microsoft Copilot, Anthropic, Perplexity, Vercel and others as it develops infrastructure for agent-driven commerce.

Merchants risk becoming invisible to AI agents

The change could also create a new form of competitive pressure. Traditional SEO focuses on ensuring that search engines can discover and understand a merchant. Agentic commerce introduces another audience: machines that are actively choosing which merchant should receive a transaction.

Stripe argues that businesses need structured product catalogs, clean APIs, transparent policies and real-time availability if they want AI agents to transact with them successfully. A retailer that is difficult for an AI agent to understand could theoretically become less competitive even if its consumer-facing website remains excellent.

The industry may therefore see the emergence of what could effectively become agent optimization alongside traditional SEO and conversion-rate optimization.

Trust remains the biggest challenge

Despite rapid technological development, agentic commerce still faces substantial barriers. Consumers must trust AI systems enough to allow them to spend money. Businesses must also distinguish legitimate AI agents from malicious automated systems.

Traditional fraud tools rely heavily on signals associated with human behavior, devices and payment credentials. Agent-initiated transactions can change those patterns. Stripe has acknowledged that agentic commerce challenges many assumptions built into existing permission and fraud systems.

Its Shared Payment Tokens, Link agent wallet and Stripe Radar infrastructure are designed partly to address this problem by attaching authorization and risk information to agent transactions. Human approval is also likely to remain important for many purchases, particularly expensive or sensitive transactions.

Agentic commerce therefore does not necessarily imply giving autonomous software unlimited purchasing authority. Instead, the emerging model relies on controlled delegation.

The checkout page may disappear, but checkout will not

Stripe's prediction should therefore be understood carefully. Commerce will still require authorization, payment processing, taxes, fraud prevention, shipping and fulfillment. What could disappear is the visible page where those processes traditionally come together.

For consumers, that could mean fewer forms, fewer redirects, and fewer abandoned carts. For merchants, however, it could require substantial behind-the-scenes changes. Retailers may increasingly need commerce systems that can communicate directly with AI agents while preserving pricing control, customer data, brand identity, payment security and fulfillment rules.

Stripe is betting that this transition will happen quickly. The company said during Stripe Sessions 2026 that it expects AI agents eventually to account for a significant share of online transactions and is positioning its payments infrastructure accordingly.

E-commerce is moving beyond the website

The broader significance of Gaybrick's comments is not simply that checkout pages may become shorter or more convenient.

They suggest that the website itself could gradually lose its position as the central interface of digital commerce.

In an agentic-commerce environment, consumers may increasingly interact with AI assistants rather than individual retailer websites. Those agents could search across merchants, evaluate products, negotiate constraints and complete transactions without reproducing the traditional browsing journey.

The merchant remains responsible for the product, pricing, payment and fulfillment. But the interface connecting the merchant to the consumer may belong to someone else. That makes Stripe's prediction about the death of the checkout page part of a much larger transformation. The next era of e-commerce may not be defined by better checkout pages. It may be defined by commerce that no longer needs them.

The post Stripe says traditional checkout pages could disappear in the Agentic Commerce Era appeared first on Cross-Border Magazine.

Alipay has launched what it calls China’s first full-stack agentic commerce platform, introducing new infrastructure designed to help merchants shift from conventional digital commerce to operations increasingly managed and accessed by artificial intelligence agents.

The platform was presented at Alipay’s AI Ecosystem Partner Conference and brings together tools that make businesses' products, services, workflows, and payment capabilities accessible to AI agents. According to Alipay, businesses with different levels of AI maturity can adopt different components of the system, rather than building their own agentic-commerce infrastructure from scratch.

The announcement represents another step in the rapid development of agentic commerce in China, where Alipay has already been developing AI-native payment technologies, trust protocols and consumer-facing AI agents.

What is Alipay’s agentic commerce platform?

Traditional e-commerce has largely been designed around humans interacting directly with websites, apps and marketplaces.

Agentic commerce introduces another layer: AI agents that can search, compare, select and potentially purchase products or services on behalf of consumers.

Unlike conventional chatbots, agentic AI systems can interact with external software and perform actions toward a defined objective with limited human supervision.

For merchants, this creates an important technical challenge. A product being visible on a website or searchable through Google does not automatically mean that an AI agent can understand the service, determine whether it is appropriate for the user, and complete the required transaction.

Alipay’s new platform is intended to bridge that gap. The company says merchants without their own AI capabilities can use a one-stop solution to convert product and service processes into tools that AI agents can understand and invoke. Alipay describes these as agent-ready Skills and MCP-based tools.

This could allow merchants to make existing commercial services available through emerging AI interfaces without rebuilding their entire technology stack.

From digital storefronts to AI-accessible businesses

One of the most important implications of agentic commerce is that the traditional online storefront may no longer be the only interface between a consumer and a merchant. Instead of opening several websites and comparing products manually, a consumer could potentially tell an AI assistant:

“Find me the best hotel near my meeting, book it within my budget and pay using my preferred payment method.”

The AI agent would then need to interact with multiple businesses, evaluate availability and pricing, verify relevant information and complete the transaction. For this model to work at scale, merchants need infrastructure that can communicate with those agents.

Alipay is positioning its new platform as part of that infrastructure. Its approach combines merchant services, AI tools, payment capabilities, identity systems, risk controls, and transaction infrastructure to support the entire agent-driven commercial process.

Ah Bao becomes a gateway to Alipay’s AI commerce ecosystem

A major part of Alipay's strategy is Ah Bao, the company’s consumer-facing AI service agent. Launched in June 2026, Ah Bao provides access to more than 10,000 AI-enabled everyday services, Alipay says. Consumers can use natural-language conversations to do everything from paying utility bills and locating electric-vehicle charging stations to booking services.

Alipay is now developing Ah Bao as a gateway connecting consumers with businesses participating in its agentic ecosystem.

This marks an important difference between generative AI assistants that simply recommend a service and systems that can actually complete the requested task.

The commercial opportunity increasingly lies in moving AI from recommendation to transaction.

For merchants, participating in these ecosystems may eventually become as important as optimizing products for search engines or marketplaces today.

Businesses may need to ensure not only that consumers can discover their products, but also that AI agents can understand, compare, and transact with them.

Payments are becoming AI-native

Payments are another essential piece of the agentic commerce model.

Earlier in 2026, Alipay introduced new AI payment infrastructure including AI Wallet and Token Pay, designed to enable transactions involving AI agents and AI-powered applications.

The company has also developed an Agentic Commerce Trust Protocol, intended to establish a standardized framework for interaction between AI systems and service platforms.

The challenge is significant.

Allowing an AI agent to make purchases raises questions about authorization, identity, transaction limits, fraud prevention, and whether the AI's actions accurately reflect the user's intention.

Alipay has therefore built security and trust mechanisms alongside its payment infrastructure, rather than treating AI payments as just another checkout method.

This development is not limited to Alipay's domestic ecosystem.

Alipay+ has also introduced an Agentic Mobile Protocol designed to let AI agents initiate payments through compatible digital wallets while users maintain control over spending intentions, budgets, and security.

The infrastructure supports scenarios where AI agents can transact without requiring consumers to manually open a payment application for every purchase.

Why agentic commerce matters for e-commerce merchants

The arrival of full-stack agentic commerce platforms could eventually change how online businesses approach customer acquisition.

Today, e-commerce companies invest heavily in areas such as:

SEO
marketplace optimization
paid advertising
social commerce
mobile applications
conversion optimization

Agentic commerce could add another requirement: AI discoverability and transactability.

An AI shopping agent needs structured information about products, pricing, availability, delivery, returns and payment options before it can confidently select a merchant.

That means product feeds and APIs could become increasingly important parts of customer acquisition infrastructure.

Commerce may consequently evolve from businesses optimizing primarily for human interfaces toward businesses optimizing simultaneously for humans and machines.

Competition for the AI commerce layer is intensifying

Alipay is not developing agentic commerce in isolation.

Technology companies, payment networks, marketplaces and AI providers are increasingly exploring ways for autonomous AI agents to interact with commercial services.

The strategic question is becoming which platforms will control the infrastructure connecting:

Consumers → AI agents → merchants → payments → fulfillment

The companies controlling those connections could gain significant influence over how products are discovered and transactions are routed.

Alipay already occupies an important position in digital payments and consumer services in China, giving it an existing ecosystem from which to develop an agentic commerce layer.

Its latest platform therefore represents more than the introduction of another AI assistant.

It aims to provide the underlying commercial infrastructure through which AI agents can interact with thousands of businesses.

Agentic commerce still faces major challenges

Despite rapid development, fully autonomous commerce remains an emerging model.

Security is one of the largest concerns.

Recent academic research examining agentic-commerce systems has highlighted potential vulnerabilities at the protocol level, particularly when agents interact with credentials, payment systems and external services.

Merchants and payment providers must also address questions surrounding authorization, liability, refunds and disputes.

If an AI agent buys the wrong product, exceeds a budget or incorrectly interprets a consumer's request, responsibility for that transaction could become difficult to determine.

Reliable identity verification and clear user controls will therefore be critical before highly autonomous purchasing becomes mainstream.

The next phase of e-commerce could be machine-to-business

Alipay’s launch shows how quickly agentic commerce is moving from experimentation to commercial infrastructure. The internet was originally designed for humans to visit websites.

Mobile commerce shifted that interaction toward apps. Marketplaces then centralized product discovery and purchasing within large commercial ecosystems. AI agents could represent the next change in that interface.

Consumers may increasingly describe what they want to achieve, while software agents determine which merchants, products, and services can fulfill that request.For e-commerce businesses, that could ultimately create a new competitive requirement. Visibility may no longer be enough.

Merchants may also need to become visible, understandable and transactable to AI agents. Alipay’s full-stack agentic commerce platform is an early indication of what that future could look like.

The post Alipay Launches China’s First Full-Stack Agentic Commerce Platform appeared first on Cross-Border Magazine.

Alibaba Group has agreed to sell its gaming business Lingxi Games to Asian private equity firm Trustar Capital, marking another significant step in the Chinese technology giant’s strategy to concentrate resources on e-commerce, artificial intelligence and cloud computing.

The transaction involves Alibaba transferring its entire stake in Lingxi Games to Trustar Capital. Reuters reported that the deal is worth more than $2 billion, citing a source familiar with the transaction, although an internal memo confirming the sale did not disclose the final valuation or a closing date.

The divestment reflects Alibaba’s increasingly focused strategy as it redirects capital from businesses it considers non-core to areas it sees as central to its future growth.

Alibaba exits Lingxi Games

Lingxi Games is a Guangzhou-based game developer best known for titles including Three Kingdoms: Strategy Edition, developed in collaboration with Japanese game company Koei Tecmo.

According to reports, Lingxi CEO Zhou Bingshu told employees that Alibaba would transfer full ownership of the company to Trustar Capital. Management and operations are expected to remain largely stable following the acquisition.

Trustar Capital, formerly known as CITIC Capital, emerged as the buyer after reports in early August suggested several potential bidders had been evaluating the gaming company.

The disposal comes after Alibaba spent several years building a presence in China's highly competitive gaming sector. However, gaming has become increasingly peripheral to the group's wider strategic direction.

Instead, Alibaba is consolidating around two areas that management has repeatedly identified as fundamental to the company’s future: commerce and artificial intelligence.

E-commerce and AI become Alibaba's strategic priorities

Alibaba's decision to sell Lingxi is not an isolated restructuring move.

The company formally identified e-commerce and AI technology as its most strategically important directions in 2024, when it reorganized its domestic and international commerce operations under a newly integrated e-commerce business group.

Since then, Alibaba has been reducing its exposure to businesses outside those priorities.

The company previously disposed of non-core retail assets including its holdings in Sun Art and Intime. Alibaba said monetizing these investments would allow it to concentrate resources on developing its core businesses.

Selling Lingxi Games fits directly into that strategy.

Rather than maintaining a broad portfolio spanning retail, entertainment, gaming and other sectors, Alibaba is increasingly positioning itself as a technology and commerce company built around a combination of e-commerce platforms, cloud infrastructure and AI.

Alibaba is spending heavily on AI

The shift is particularly significant because Alibaba is committing enormous amounts of capital to artificial intelligence infrastructure.

In February 2025, the company announced plans to invest at least RMB380 billion, approximately $53 billion at the time, in AI and cloud infrastructure over a three-year period. Alibaba said the investment would exceed its total spending on AI and cloud infrastructure during the previous decade.

That investment covers computing infrastructure, AI models, cloud services, and technologies capable of supporting increasingly advanced AI applications.

Alibaba CEO Eddie Wu has positioned AI as one of the company's largest technological opportunities. The group has since expanded its Qwen family of AI models, increased its cloud computing capacity, and developed AI services for both consumers and businesses.

By the end of Alibaba's 2026 fiscal year, AI-related products represented around 30% of Alibaba Cloud's external revenue, while external cloud revenue grew 40% during the final quarter of the year, according to the company.

The numbers demonstrate why Alibaba increasingly views AI and cloud infrastructure not simply as technology investments but as major future revenue engines.

AI is also becoming part of Alibaba's e-commerce strategy

For e-commerce businesses, however, Alibaba's strategy matters because it is not developing AI and commerce as separate divisions. Instead, Alibaba is increasingly integrating artificial intelligence throughout its commerce ecosystem.

AI can potentially influence almost every layer of an online marketplace, from product discovery and search to advertising, merchant services, customer support, logistics optimization and personalized recommendations.

Alibaba has already deployed AI solutions across Taobao and Tmall and has increasingly connected its Qwen ecosystem to its commerce platforms. In 2026, for example, the company integrated the Qwen app with Taobao's product catalog and introduced Qwen-powered shopping capabilities inside Taobao. This creates an important strategic advantage.

Alibaba operates major consumer and B2B commerce platforms while simultaneously controlling cloud infrastructure and developing its own foundation AI models. Those capabilities could allow Alibaba to build AI shopping tools without relying heavily on third-party technology providers.

The race toward agentic commerce

The evolution could become even more significant as e-commerce moves toward agentic AI. Traditional online shopping still requires consumers to search for products, compare options, evaluate sellers and manually complete purchases.

AI agents could increasingly perform parts of that process independently. A consumer might eventually ask an AI assistant to find a specific product within a budget, compare delivery times, evaluate seller reliability, and recommend or potentially execute the best purchase.

For marketplace operators, this changes the importance of both product data and the technology controlling the interface between consumers and merchants. Alibaba has already described agentic AI and Model-as-a-Service as important components of its long-term AI strategy.

Integrating those technologies with Taobao, Tmall, Alibaba.com, AliExpress and other commerce platforms could give the company a large real-world environment in which to deploy AI-driven shopping services.

International e-commerce remains important

Alibaba's commerce strategy also extends far beyond China. Its international operations include platforms such as AliExpress, Lazada, Trendyol and Alibaba.com, giving the company exposure to both consumer and business-to-business cross-border commerce.

Alibaba previously reported that international commerce revenue was being supported by growth in businesses including AliExpress and Trendyol. The group therefore has an opportunity to apply AI capabilities across markets with very different languages, consumer behaviors and merchant ecosystems.

AI tools could prove especially valuable in cross-border commerce, where sellers must deal with challenges including multilingual product information, localization, customer service, search relevance and international marketing.

As AI becomes more deeply embedded in Alibaba's international platforms, these services could lower some of the barriers merchants face when selling into unfamiliar markets.

A leaner Alibaba emerges

The Lingxi Games sale therefore represents more than simply the disposal of a gaming studio. It also signals how Alibaba is reshaping itself after years of operating as a highly diversified technology conglomerate.

The company has repeatedly stressed portfolio discipline and the disposal of non-core assets while increasing spending on businesses that can support long-term growth. That increasingly means concentrating resources around consumption, e-commerce, cloud infrastructure and artificial intelligence.

Gaming may remain a large global digital industry, but maintaining a major gaming studio does not necessarily strengthen Alibaba's position in its most strategically important markets.

Selling Lingxi lets the company simplify its portfolio while potentially freeing up billions of dollars in capital as investment requirements in AI infrastructure continue to grow.

What the Lingxi Games sale means for e-commerce

For the wider e-commerce industry, Alibaba's restructuring illustrates how quickly AI is becoming intertwined with competition between the world's largest commerce platforms.

Amazon, Alibaba, Google, Meta and other technology companies are all exploring how generative and agentic AI could change product discovery and purchasing. The competition is no longer simply about which marketplace has the largest selection of merchants or the fastest logistics network.

Increasingly, it could also depend on who controls the AI systems consumers use to discover, compare and buy products. Alibaba's combination of marketplaces, merchant relationships, consumer data, cloud infrastructure and proprietary AI models puts it in a particularly strong position to pursue that opportunity.

The sale of Lingxi Games suggests the company is willing to give up significant businesses if they no longer fit that objective. And as Alibaba continues to invest tens of billions of dollars into AI and cloud infrastructure, the direction of travel is becoming increasingly clear: the company wants e-commerce and artificial intelligence to reinforce each other as the foundations of its next stage of growth.

The post Alibaba Sells Lingxi Games as It Doubles Down on E-Commerce and AI appeared first on Cross-Border Magazine.

A new phase of the European Union’s Artificial Intelligence Act has begun, bringing important transparency obligations into effect for companies using chatbots, generative AI and synthetic content.

From 2 August 2026, Article 50 of the EU AI Act applies across the European Union. The rules are designed to ensure that consumers understand when they are interacting with an AI system and when apparently authentic content has been artificially generated or manipulated.

For e-commerce companies, the changes may affect customer-service bots, virtual shopping assistants, AI-generated advertising, product imagery, synthetic influencers and automated editorial content.

However, the regulation does not require retailers to label every product description or marketing image created with AI. The obligations depend on the type of system, the nature of the content and whether the company is acting as an AI provider or simply using a third-party tool.

E-commerce chatbots must identify themselves as AI

One of the clearest requirements concerns AI systems that interact directly with consumers.

Providers of these systems must ensure that users are informed that they are communicating with AI, unless this is already obvious to a reasonably informed person.

For online retailers, this can include:

• Customer-service chatbots
• Virtual shopping assistants
• Automated returns agents
• Conversational recommendation tools
• AI-powered order-tracking systems
• Voice-based shopping assistants

The disclosure should appear no later than the first interaction. A message such as “Hello, I’m an AI-powered shopping assistant” or a persistent “AI Assistant” label within the interface may satisfy the requirement.

Retailers should not assume that a robot icon, unusual bot name or automated writing style makes the artificial nature of the system sufficiently clear.

Although the legal design obligation falls primarily on the system provider, e-commerce companies should verify that the disclosure is displayed properly across their websites, applications and customer-service channels.

Vendor contracts should also clarify who is responsible for implementing, testing and maintaining the notice.

Not all AI-generated content needs a visible label

Article 50 also covers AI-generated and manipulated audio, images, video and text.

Providers of generative AI systems must ensure that covered outputs can be identified in a machine-readable format. This may involve metadata, watermarking, content credentials or another technical method.

This technical marking requirement does not mean that every retailer must place a visible “AI-generated” label on every piece of AI-assisted content.

The visible disclosure duties placed on businesses using the content are narrower. They primarily concern:

• Deepfakes
• Certain AI-generated texts dealing with matters of public interest
• Emotion-recognition and biometric-categorization systems

A standard product description, category page or promotional email does not automatically need a visible AI label simply because generative AI helped produce it.

What this means for product descriptions

Retailers increasingly use AI to create:

• Product titles and descriptions
• Marketplace listings
• Category pages
• Buying guides
• Marketing emails
• Translations
• Frequently asked questions

Most ordinary commercial copy will not require a visible AI-generated label.

The specific disclosure obligation for AI-generated text applies when the material is published to inform the public about a matter of public interest.

This may become relevant when an e-commerce company publishes AI-generated content involving:

• Product safety
• Public health
• Environmental claims
• Sustainability
• Financial services
• Consumer rights
• Legal or regulatory developments

For example, a simple description of a product’s dimensions is unlikely to trigger the rule. An AI-written article claiming that the product provides medical benefits or complies with environmental regulations may require closer scrutiny.

Human editorial review remains important

Even when AI-generated text addresses a matter of public interest, a visible label may not be required if the content has undergone meaningful human review and a person or company assumes editorial responsibility.

The review must be more than a basic spelling or grammar check.

A qualified editor should assess the substance of the content, verify its sources, correct inaccurate claims, and have the authority to approve or reject the final version.

Retailers should document:

• Who reviewed the material
• Which claims were checked
• What sources were used
• Who approved publication
• Who assumes editorial responsibility

This is particularly relevant for company blogs, sustainability reports, regulatory explainers and product-safety content.

When AI-generated images may need disclosure

AI-generated advertising and product imagery do not automatically require a visible label.

The main visible-disclosure obligation applies when the content qualifies as a deepfake.

Under the AI Act, a deepfake is an AI-generated or manipulated image, audio or video content that resembles a real person, object, place, organization or event and could falsely appear authentic.

In e-commerce, this could include:

• A synthetic celebrity endorsement
• An artificial customer testimonial
• A fake video of a real person using a product
• A fabricated representation of a warehouse or factory
• A manipulated demonstration that exaggerates product performance
• A realistic virtual influencer presented as a genuine person

When content qualifies as a deepfake, its artificial nature must be disclosed clearly when the viewer first encounters it.

Machine-readable metadata alone is not enough. Consumers must be able to understand that the material is artificial without using specialized software.

Virtual models and synthetic influencers

Not every virtual model will necessarily qualify as a deepfake.

A clearly fictional animated character or obvious digital mascot is less likely to mislead consumers than a photorealistic virtual influencer presented as a real customer.

Retailers should consider:

• How realistic the content appears
• Whether it resembles a real person
• Whether viewers are likely to believe it is authentic
• Whether it imitates a testimonial or endorsement
• Whether the artificial nature is clear from the context

Even when the AI Act does not require a label, EU advertising and consumer-protection rules may still apply.

Companies should not create the false impression that a genuine customer used, reviewed or recommended a product when the person is entirely synthetic.

AI editing of product photography

Routine image editing does not automatically trigger the AI Act’s transparency requirements.

Basic adjustments such as resizing, color correction, sharpening and noise removal are less likely to be treated as synthetic-content generation.

The risk increases when AI is used to make substantial changes, including:

• Altering the size or proportions of a product
• Simulating features the product does not have
• Changing materials or colors inaccurately
• Adding accessories that are not included
• Manipulating how clothing fits a model
• Creating fabricated before-and-after results

These practices may violate consumer-protection rules even when they fall outside the strict deepfake definition.

AI-generated reviews remain a major risk

The new transparency rules should not be interpreted as permission to create fake customer reviews.

Publishing fabricated reviews, endorsements or testimonials can already breach EU consumer-protection law.

An AI-generated testimonial may also qualify as a deepfake when it depicts a realistic person describing an experience that never occurred.

Retailers should clearly distinguish between:

• A genuine customer review
• A paid endorsement
• A fictional demonstration
• A virtual spokesperson
• A simulated use case

The safest policy is to prohibit synthetic customer reviews altogether.

Marketplaces face additional challenges

Online marketplaces may need to manage AI-generated content produced by thousands of third-party sellers.

They should review whether their systems:

• Preserve machine-readable AI markings
• Allow sellers to disclose deepfakes
• Prohibit synthetic reviews
• Prevent misleading AI-generated product demonstrations
• Offer reporting tools for deceptive content
• Retain disclosures when listings are reformatted or syndicated

Marketplaces that provide their own AI listing generators should also assess whether they are acting only as deployers or may assume additional responsibilities as providers.

A practical compliance checklist

E-commerce companies should now take several immediate steps.

Audit customer-facing AI

Identify every chatbot, shopping assistant, voice agent and automated support tool that interacts with customers.

Confirm that each system clearly identifies itself as AI.

Define legal responsibilities

Establish whether the company is acting as a provider, deployer, distributor or importer for each AI system.

Do not rely only on the terminology used by the vendor.

Review synthetic media

Identify where AI is used to generate or substantially alter images, audio or video.

Assess whether any material could qualify as a deepfake.

Protect technical markings

Check whether website optimization, image compression, marketplace uploads or content-management systems remove metadata or watermarks added by the AI provider.

Introduce editorial controls

Create a documented human-review process for AI-generated content involving safety, health, sustainability, regulation or other public-interest topics.

Update supplier contracts

Require AI vendors to explain:

• How chatbot disclosures are implemented
• How synthetic content is marked
• Whether the marking survives export and compression
• What compliance documentation is available
• How future regulatory changes will be communicated

Train relevant teams

Marketing, legal, customer service, e-commerce, content and technology teams should all understand the transparency requirements.

Compliance is not solely an IT responsibility.

Penalties and commercial risk

Breaches of the relevant EU AI Act obligations can result in fines of up to €15 million or 3% of the company’s worldwide annual turnover, whichever is higher.

Authorities must consider proportionality, including the size of the business and the seriousness of the infringement.

However, financial penalties are not the only concern.

Misleading AI-generated content may also lead to:

• Consumer complaints
• Advertising investigations
• Platform sanctions
• Product-information disputes
• Reputational damage
• Loss of customer trust

Transparency as part of the customer experience

The EU AI Act does not prevent retailers from using chatbots or generative AI.

Its central principle is that consumers should not be misled about whether they are interacting with a person or viewing content that appears authentic but is artificially generated.

A clearly labeled chatbot can still provide fast and effective service. AI-assisted content can still improve productivity when its claims are properly reviewed. Virtual models and synthetic media can still be used when their presentation is transparent and not deceptive.

For e-commerce businesses, the priority should be to identify where AI interacts directly with consumers, where synthetic content could be mistaken for reality, and where human editorial responsibility is required.

As AI becomes more deeply integrated into online retail, transparency will become both a legal obligation and a competitive trust signal.

The post EU AI Act enforcement begins: What e-commerce businesses must disclose about chatbots and AI-generated content appeared first on Cross-Border Magazine.

ESW has launched a new agentic commerce solution designed to help international brands make their products discoverable, purchasable and operationally supported inside artificial intelligence platforms.

The cross-border e-commerce provider announced that Microsoft Copilot will be the first integration for ESW Agentic Commerce, allowing participating brands to connect AI-powered product discovery directly with secure checkout and payment capabilities.

The launch represents another important step in the development of agentic commerce: a model in which AI assistants do more than recommend products. They can increasingly compare options, guide purchasing decisions and complete transactions on behalf of shoppers.

For international retailers, however, connecting an AI recommendation to a completed order is considerably more complicated than adding a buy button to a chatbot. Payments, localisation, fraud prevention, import duties, compliance, fulfilment and returns must still operate correctly behind the interface.

ESW is positioning its new Agentic Commerce infrastructure as the layer that manages that complexity.

What is ESW Agentic Commerce?

ESW Agentic Commerce is a new infrastructure solution intended to connect retailers’ existing e-commerce operations with AI-powered shopping environments.

The system allows brands to integrate and optimise their product catalogues for discovery by AI platforms. Once a shopper finds a suitable product, the transaction can be completed securely within the AI experience rather than requiring the customer to move to a separate retailer website.

According to ESW, the solution works alongside a retailer’s existing e-commerce infrastructure. Brands therefore do not need to replace their current commerce platform to participate in AI-assisted shopping channels.

The technology is supported by the ESW Agentic Hub, which acts as an infrastructure layer connecting AI agents with the company’s international commerce capabilities.

Through this layer, AI platforms can potentially access the product, checkout and operational services required to move a shopper from initial discovery to a completed cross-border order.

ESW says Microsoft Copilot will be the first integration, with support for additional agents and platforms expected in the coming months. The service is immediately available to brands and customers in the United States, while broader availability is planned.

Microsoft Copilot becomes the first integration

Microsoft has been developing Copilot into a transactional shopping environment rather than limiting it to search, recommendations and product comparisons.

Copilot Checkout allows shoppers to discover, evaluate and purchase products without leaving the conversational interface. Microsoft says the merchant remains the merchant of record and retains ownership of the transaction, customer data and commercial relationship.

The ESW integration extends this concept towards international commerce.

A customer could theoretically ask Copilot for a product recommendation, compare several options, select an item and complete the purchase inside the same conversation. ESW would then provide the underlying international commerce infrastructure required to process and support the transaction.

That infrastructure can include:

• Localised checkout experiences
• International payments
• Fraud management
• Duties and tax calculations
• Regulatory compliance
• Cross-border shipping
• Fulfilment
• Customer service
• International returns

The customer may experience the journey as a simple conversation, but the transaction still depends on a complex network of commercial and logistical systems.

This operational layer is particularly important when the shopper, merchant, payment provider and delivery destination are located in different countries.

From conversational search to conversational checkout

Generative AI has already changed how some consumers research products. Shoppers can ask an AI assistant to compare specifications, interpret reviews, recommend suitable products or narrow a large catalogue according to highly specific requirements.

Until recently, however, most AI-assisted shopping journeys ended with a link.

The AI could recommend a product, but the shopper would usually have to leave the conversation, visit the merchant’s website, find the correct product again and proceed through a conventional checkout.

Each additional step created an opportunity for the customer to abandon the purchase.

Agentic commerce attempts to remove that separation between discovery and transaction. Instead of acting primarily as a search interface, the AI assistant becomes an active participant throughout the purchasing journey.

Microsoft reported that shopping journeys involving Copilot produced 53% more purchases within 30 minutes of an interaction than journeys without Copilot. When clear shopping intent was present, journeys involving Copilot were reportedly 194% more likely to result in a purchase.

These figures were published by Microsoft and should be interpreted as early platform data rather than independent measurements of the wider e-commerce market. Nevertheless, they suggest that reducing the distance between product recommendation and checkout could have a measurable effect on conversion.

Why agentic commerce is especially relevant to cross-border retail

Domestic transactions are already complicated. International orders add multiple layers of friction and risk.

A product recommended by an AI assistant may be available in one country but restricted in another. The displayed price may need to include local taxes, import duties or customs charges. Payment preferences, consumer regulations, delivery options and return procedures may also differ by market.

An AI assistant cannot reliably complete an international order simply by identifying the right product. It must also connect with systems capable of answering operational questions in real time.

These may include:

• Is the product available in the shopper’s country?
• What is the final landed cost?
• Which currency and payment methods should be offered?
• Are import restrictions applicable?
• How long will delivery take?
• Who is responsible for collecting taxes and duties?
• Can the item be returned locally?
• How will fraud and payment risk be managed?

ESW already provides international commerce services covering payments, compliance, localisation, shipping and returns across more than 200 markets. The company says it works with more than 150 global brands and processes more than eight million cross-border orders annually.

Agentic Commerce applies those capabilities to a new customer interface.

Rather than building a separate international transaction system for every AI platform, ESW’s proposed model gives brands a central infrastructure layer that can be extended across multiple agents.

Product data becomes a commercial priority

The emergence of AI-powered shopping will also increase the importance of product data.

Traditional e-commerce search is often built around keywords, categories, filters and paid placements. AI discovery is more conversational and contextual.

A shopper may not search for a particular product name. Instead, the shopper might ask:

“Find me a waterproof jacket suitable for a week in Iceland that fits in hand luggage and costs less than €200.”

To respond accurately, the AI must understand far more than a product title and category. It may need structured information about materials, sizing, climate suitability, weight, delivery availability, price, inventory and return conditions.

Brands whose product information is incomplete, inconsistent or inaccessible to AI systems may become less visible even when their products would otherwise meet the customer’s requirements.

This creates a new optimisation discipline for merchants. In addition to search engine optimisation, marketplace optimisation and feed management, brands will increasingly need to consider how products are interpreted and selected by AI agents.

Important areas will include:

• Accurate structured product attributes
• Consistent descriptions across channels
• Real-time pricing and inventory
• Clear shipping eligibility
• Detailed sizing and compatibility information
• Transparent return conditions
• Verified brand and product information
• Machine-readable commercial policies

In agentic commerce, product data is not merely descriptive content. It becomes part of the decision-making infrastructure.

ESW says retailers can retain their existing platforms

One of the most significant aspects of the ESW proposition is that retailers are not expected to replace their existing e-commerce platforms.

ESW Agentic Commerce is designed to operate alongside the merchant’s current technology stack. The retailer can therefore extend its catalogue and transaction capabilities into AI environments while continuing to use its existing platform for core commerce operations.

This reduces one of the most obvious barriers to agentic commerce adoption.

Large international retailers often operate highly customised systems involving commerce platforms, enterprise resource planning software, product information management, order management, warehouse systems and multiple payment providers.

Replacing that infrastructure solely to participate in a new AI sales channel would be costly and risky.

An integration layer allows retailers to test agentic commerce without treating it as a complete platform migration.

Nevertheless, brands will still need to determine how orders generated through AI interfaces are attributed, managed and measured. Questions around customer identity, consent, marketing permissions, returns and post-purchase communication will become increasingly important as the channel develops.

The retailer must remain visible

Agentic commerce could reduce purchasing friction, but it may also weaken the direct relationship between shoppers and retailers.

When a customer shops through an AI assistant, the assistant may control much of the experience:

• The initial recommendation
• The comparison criteria
• The products displayed
• The explanation of advantages and disadvantages
• The checkout interface
• The post-purchase conversation

This creates a strategic risk for brands. The AI platform could become the primary interface, while the retailer becomes an invisible fulfilment provider.

Microsoft has attempted to address this concern by emphasising that merchants participating in Copilot Checkout remain the merchant of record and retain control of their transaction data and customer relationship.

However, formal ownership of the transaction does not necessarily guarantee control over the customer experience.

Retailers will need to ensure that AI commerce integrations preserve brand identity, communicate accurate product information and create opportunities for continued customer engagement after the purchase.

Otherwise, brands may become increasingly dependent on AI platforms in much the same way that many merchants became dependent on search engines, marketplaces and social networks.

Agentic commerce will create new compliance questions

The ability of an AI system to complete a purchase introduces legal and regulatory questions that conventional product recommendation tools do not face.

Retailers must be able to demonstrate that customers were shown accurate prices, material product information, delivery conditions and return rights before the transaction was completed.

For cross-border orders, the process may also involve:

• Consumer protection requirements
• Data protection rules
• Product safety obligations
• Customs declarations
• Sanctions screening
• Import restrictions
• Tax collection
• Payment authentication
• Accessibility requirements
• Records of customer consent

There is also the question of responsibility when an AI assistant recommends an unsuitable product, presents incorrect information or completes a purchase based on an ambiguous instruction.

The commercial agreement between the retailer, commerce provider and AI platform will need to define how responsibility is distributed.

These issues will become more significant as AI agents gain greater autonomy. Recommending a product is different from selecting quantities, accepting terms, choosing delivery services or authorising payment.

Agentic commerce infrastructure must therefore provide both convenience and traceability.

AI platforms are becoming a new distribution channel

ESW’s launch should be understood as part of a broader shift in e-commerce distribution.

Retailers have previously adapted their businesses for successive generations of digital channels:

• Desktop websites
• Mobile commerce
• Online marketplaces
• Social commerce
• Retail media
• Voice commerce
• Conversational commerce

AI assistants may now become another major entry point.

ESW cites McKinsey research estimating that agentic commerce could represent a global opportunity of between $3 trillion and $5 trillion by 2030. The company also references a Gartner prediction that traditional search engine volume could decline by 25% as consumers increasingly use AI chatbots and virtual agents.

These projections should not be treated as guarantees that AI agents will replace conventional e-commerce journeys. Consumer trust, technical reliability, product coverage and regulatory oversight will all influence adoption.

However, the direction of investment is increasingly clear.

Microsoft, Google, OpenAI, payment companies, commerce platforms and major retailers are all developing systems that allow AI assistants to participate more directly in shopping and transactions.

The strategic question for retailers is therefore shifting from whether AI will influence product discovery to how much of the commercial journey AI platforms will ultimately control.

What retailers should do now

Most brands do not need to rebuild their entire commerce strategy around autonomous AI agents immediately. However, waiting until the channel is mature could leave retailers with significant product-data and infrastructure work to complete.

Retailers preparing for agentic commerce should begin with several practical steps.

Audit product information

Product catalogues should contain detailed, consistent and structured information that AI systems can interpret accurately.

Missing attributes, inconsistent naming and vague descriptions could reduce both visibility and recommendation quality.

Improve real-time data access

AI shopping systems require current information about prices, promotions, inventory, delivery options and geographical availability.

Static or frequently outdated feeds will not be sufficient for transactional experiences.

Review international checkout capabilities

Retailers should assess whether their existing systems can calculate local prices, duties, taxes and delivery conditions in real time.

Agentic discovery will not generate sustainable revenue if customers encounter inaccurate landed costs or failed international orders.

Establish clear governance

Brands should define which actions an AI platform can perform, what information it can access and when customer confirmation is required.

Higher-risk actions should include appropriate human approval or explicit shopper consent.

Protect the customer relationship

Retailers should understand what customer information they receive, how post-purchase communication works and whether they can continue serving the customer outside the AI platform.

Measure the channel separately

Agent-driven traffic and transactions should be distinguishable from conventional search, marketplace and direct website sales.

Without clear attribution, retailers will struggle to evaluate conversion, acquisition costs, repeat purchases and platform dependency.

A significant step, but not the finished model

ESW Agentic Commerce provides a clearer view of how AI-assisted shopping could become operational at an international scale.

The announcement moves the conversation beyond experimental chatbots and product recommendations. It focuses on the less visible infrastructure required to turn an AI interaction into a compliant, fulfilled and serviceable transaction.

Microsoft Copilot gives ESW an established AI interface through which to launch the solution. ESW contributes the international checkout, payment and operational capabilities required behind that interface.

The initial availability is currently limited to the United States, and broader adoption will depend on how quickly additional markets, retailers and AI platforms are integrated. Questions also remain regarding customer ownership, compliance, attribution and the influence AI assistants will have over product selection.

Even so, the direction is important.

E-commerce is entering a stage in which the storefront may no longer be a website, app or marketplace page. It may be a conversation.

For international brands, success in that environment will depend not only on whether an AI assistant can find their products, but whether the entire cross-border transaction can operate correctly after the recommendation is made.

The post ESW Launches Agentic Commerce with Microsoft Copilot, Connecting AI Discovery to Checkout appeared first on Cross-Border Magazine.

London-based live commerce platform Tilt has raised $26 million in fresh funding, marking one of the most notable recent funding rounds in Europe’s e-commerce startup ecosystem. The investment comes as live shopping, social commerce and AI-powered seller tools move from experimental formats to serious strategic priorities for online retailers, marketplaces and recommerce platforms.

The round includes Vinted Ventures, the investment arm of second-hand marketplace Vinted, as a new investor, alongside existing backers TQ Ventures, Balderton Capital, Earlybird, Seedcamp and others. According to Tilt, the new capital brings the company’s total funding raised to over $50 million.

Founded in 2021 by former Revolut employees Abhi Thanendran and Neil Shah, Tilt is building a live auction app where sellers showcase products in real time and buyers can bid, interact and purchase through livestreams. The company positions itself as a European answer to the rise of entertainment-driven commerce, combining video, community, auctions and AI-powered discovery into one shopping experience.

Why Tilt’s funding matters for European e-commerce

For years, European e-commerce has been defined by search, static listings, price comparison and increasingly efficient checkout flows. Tilt’s model points in a different direction: one where online shopping becomes more interactive, social and content-led.

That shift is already visible across the wider market. TikTok Shop has been expanding across Europe, joining existing markets including France, Germany, Ireland, Italy, Spain and the UK. TikTok has also reported that more than 100,000 European businesses had already joined TikTok Shop across France, Germany, Italy, Spain and Ireland, with strong growth in daily gross merchandise value across those markets.

Tilt operates in the same broader movement, but with a more focused live auction and community-commerce model. Instead of simply adding shoppable videos to a social network, Tilt is building a dedicated marketplace around real-time selling. This makes the company especially relevant for fashion, sneakers, collectibles, second-hand items and categories where scarcity, trust, personal presentation and community can influence conversion.

From static listings to live interaction

Traditional e-commerce is built around product pages. Tilt is built around live interaction.

Sellers go live, show items on camera, answer questions and create a sense of urgency through auctions. Buyers do not only browse; they watch, interact, bid and return to sellers or communities they trust. This is why live commerce can create a very different type of retention from conventional marketplace shopping.

Tilt says it has grown 8x since its 2024 Series A. The company also reports that buyers spend over an hour a day on the app, that 70% return week-on-week, and that 70% of monthly GMV comes from repeat buyers. The platform is currently live across the UK, Italy, Spain and Poland.

These figures matter because they suggest Tilt is not only acquiring occasional buyers. It is trying to build a repeat shopping habit around live discovery, seller personality and community engagement.

AI is becoming central to Tilt’s live commerce model

Tilt’s funding round is not just a live shopping story. It is also an AI e-commerce story.

The company has developed a set of AI tools designed to reduce friction for sellers and improve product discovery for buyers. These include Snap, a tool that can turn an item shown on camera into a listing within seconds; a real-time AI copilot that supports sellers with pricing and responses during livestreams; natural-language search across live sessions; and AI-generated clips that can repurpose livestream content for social media distribution.

This is strategically important. One of the challenges of live commerce is operational intensity. Sellers need to source products, present them, answer questions, price items, manage auctions, create content and maintain community engagement. AI can help reduce that workload and make live selling more scalable for smaller merchants.

For buyers, AI also helps solve a discovery problem. In a live shopping environment, inventory is fluid and time-sensitive. Traditional search is less effective when the best product may be inside a livestream happening right now. Tilt’s AI-powered matching and search tools are designed to connect users with relevant live sessions in real time.

Vinted Ventures’ participation signals a recommerce opportunity

The participation of Vinted Ventures is one of the most significant elements of the round. Vinted is one of Europe’s best-known second-hand marketplaces, and its venture arm’s investment in Tilt suggests that live commerce is increasingly relevant to the future of recommerce.

Second-hand and resale categories are particularly well suited to live selling. Items are often unique, condition varies, trust matters, and buyers may want to see products from different angles before purchasing. A livestream can make resale feel more transparent and more engaging than a static listing.

Tilt itself framed Vinted’s backing as a strategic signal. In the company’s announcement, CEO and co-founder Abhi Thanendran said that the next generation will not browse static listings in the same way, but will discover and buy through video, conversation and live interaction. Vinted’s Martijn van Heeswijk also highlighted Tilt’s differentiated experience in live selling, with potential in fashion and beyond.

For European recommerce, this could be an important development. If resale marketplaces increasingly move toward video-led discovery, the competitive landscape could shift from who has the largest catalogue to who can create the strongest buyer-seller interaction.

Europe is becoming a live commerce battleground

Live commerce has already become a major retail format in Asia. Tilt says the category is a $370 billion-plus industry in Asia, while Western markets are still at an earlier stage of adoption.

In Europe, the format is now becoming more competitive. TikTok Shop is expanding its European footprint. Whatnot has established itself as a major live auction player globally. At the same time, local platforms such as Tilt are trying to build a European model adapted to regional consumer behaviour, seller communities and recommerce habits.

The wider e-commerce context supports this shift. Eurostat data shows that in 2024, EU enterprises generated 19.49% of total turnover from e-sales, with websites and apps accounting for 8.39% and EDI-type sales accounting for 11.07%. Web sales remain the dominant e-sales method among enterprises, but the next layer of competition is increasingly about discovery, attention and conversion rather than simply having an online sales channel.

This is where live commerce becomes strategically relevant. As more businesses sell online, standing out through static listings becomes harder. Content-led and community-led formats offer a new way to reach buyers, especially younger audiences already accustomed to discovering products through short video and creator recommendations.

What Tilt will use the funding for

Tilt says the new funding will be used to accelerate its AI roadmap, expand its seller base across new markets and continue scaling the team. The company currently has a team of around 60 people.

The expansion strategy appears to focus on three areas.

First, Tilt needs to continue improving the seller experience. Live commerce only works at scale if sellers can list products quickly, manage livestreams efficiently and reach relevant audiences without excessive manual work.

Second, the company needs to expand geographically. Tilt is already active in the UK, Italy, Spain and Poland, but Europe remains fragmented by language, payment habits, delivery expectations and local marketplace preferences.

Third, Tilt needs to defend its position as larger players move into the same space. TikTok Shop has enormous distribution. Whatnot has significant funding and category experience. Traditional marketplaces could also add more livestreaming and auction features over time.

Why retailers and marketplaces should pay attention

Tilt’s raise is relevant beyond the startup itself. It reflects a broader direction of travel in e-commerce.

Retailers, brands and marketplaces are under pressure to reduce customer acquisition costs, increase retention and create more engaging shopping experiences. Live commerce offers a possible answer by combining product discovery, entertainment, community and conversion in the same environment.

For fashion and second-hand sellers, the opportunity is particularly clear. Products can be demonstrated in real time, condition can be shown more transparently, and scarcity can create urgency. For niche categories such as sneakers, collectibles or limited drops, live auctions can turn shopping into an event rather than a transaction.

However, the model also creates new operational requirements. Sellers need presentation skills, inventory discipline, fast fulfilment, clear returns policies and strong customer service. Platforms need trust and safety systems, payment infrastructure, moderation, buyer protection and logistics partnerships. In that sense, live commerce is not simply a marketing format. It is a full e-commerce operating model.

The bigger picture: AI, video and recommerce are converging

Tilt’s $26 million funding round sits at the intersection of three major e-commerce trends.

The first is AI automation. E-commerce companies are increasingly using AI to support product listing, discovery, search, content generation and customer interaction.

The second is video-led shopping. TikTok Shop’s expansion across Europe shows that major platforms believe content-driven commerce can become a mainstream sales channel.

The third is recommerce. Vinted Ventures’ participation highlights the relevance of live selling for second-hand and resale categories, where trust, uniqueness and product condition are especially important.

Together, these trends suggest that the future of online shopping in Europe may not be defined only by faster delivery or lower prices. It may also be defined by richer interaction, stronger communities and smarter discovery.

Conclusion: Tilt’s round is a signal for Europe’s next e-commerce phase

Tilt’s $26 million raise shows that investor interest in European e-commerce is moving beyond traditional marketplace models. The new capital, combined with Vinted Ventures’ participation, positions Tilt as one of the most closely watched European startups in live commerce.

The company still faces major challenges. Live shopping adoption in Europe remains less mature than in Asia, competition from TikTok Shop and Whatnot is intensifying, and scaling across Europe requires strong localization. But Tilt’s growth metrics, AI tooling and recommerce relevance make the company an important indicator of where the market may be heading.

For European e-commerce, the message is clear: the next wave of competition will not only be about who sells online. It will be about who can turn shopping into discovery, interaction and repeat engagement.

The post Tilt raises $26M as Europe’s live commerce race accelerates appeared first on Cross-Border Magazine.

Artificial intelligence is often presented as a software revolution. But the numbers tell a more physical story.

AI does not run in the abstract. It runs inside data centers. Those data centers require electricity, water, cooling systems, chips, land, substations, transformers, fiber connections, backup power and skilled workers. As AI moves from experimentation to global implementation, these physical requirements are becoming one of the technology’s biggest constraints.

The issue is not whether AI can scale in theory. The issue is whether the world can, in practice, build enough infrastructure to support it.

According to the International Energy Agency, global data center electricity consumption was around 415 TWh in 2024, equal to roughly 1.5% of total global electricity consumption. In the IEA’s base case, this figure is projected to more than double to around 945 TWh by 2030, reaching just under 3% of total global electricity demand. That means data center electricity use is expected to grow by around 15% per year between 2024 and 2030, more than four times faster than growth in overall electricity consumption from other sectors.

These figures reveal the central problem: AI is not only a digital technology. It is becoming a major physical infrastructure challenge.

415 TWh in 2024: The Starting Point of the AI Infrastructure Problem

The first number to understand is 415 TWh.

That is the IEA’s estimate for global data center electricity consumption in 2024. To put this in perspective, data centers already consumed around 1.5% of the world’s electricity before the full global implementation of AI.

This matters because the AI boom is arriving atop an existing digital infrastructure. Cloud computing, streaming, enterprise software, financial systems, e-commerce, online advertising, and social media already require large data center networks. AI adds a new and more intensive layer of demand.

Traditional data center workloads were already significant, but AI workloads are different. They require high-performance GPUs, dense server clusters, advanced networking, and more powerful cooling systems. This means that AI does not simply increase the number of servers. It changes the type of infrastructure required.

The key point is that AI begins from a high baseline. The world is not moving from zero to AI infrastructure. It is adding AI to an already energy-intensive digital economy.

945 TWh by 2030: Data Center Electricity Demand Could More Than Double

The IEA projects global data center electricity consumption to reach around 945 TWh by 2030 in its base case. That is more than double the 2024 level.

This is one of the most important numbers in the AI infrastructure debate.

A rise from 415 TWh to 945 TWh means an increase of roughly 530 TWh in just six years. That additional demand exceeds the total annual electricity consumption of many countries.

The number also explains why governments, utilities, and energy companies are becoming increasingly involved in AI strategy. AI expansion is no longer only a question for technology companies. It is also a question for power markets, grid planners, and national energy policy.

The challenge is not only total consumption. It is also speed. A 15% annual growth rate in data center electricity consumption is extremely fast for an infrastructure-heavy sector. Electricity grids, substations, transmission lines, and power generation assets are not built at software speed.

This creates one of the main physical walls of AI: demand can grow faster than infrastructure can respond.

Just Under 3% of Global Electricity Demand: Why the Global Figure Can Be Misleading

By 2030, data centers could account for just under 3% of global electricity consumption, according to the IEA base case.

At first glance, 3% may not sound overwhelming. But this global percentage can be misleading for three reasons.

First, data center demand is highly concentrated. AI infrastructure is not evenly distributed across the world. It tends to cluster around major cloud regions, fiber routes, energy hubs, financial centers, and technology markets.

Second, data centers require reliable, continuous power. A large AI facility cannot depend on an intermittent or uncertain electricity supply. It needs firm capacity, high uptime, and stable grid connections.

Third, local grids can become stressed even if the global electricity share remains modest. A national or global average hides local bottlenecks. One region may have enough electricity on paper but lack the transmission capacity, substations or transformers needed to connect a new AI data center campus.

This is why the data center energy debate should not focus only on the global percentage. The more important question is whether the right amount of power can be delivered to the right locations at the right time.

176 TWh in the United States in 2023: The US Shows the Scale of the Challenge

The United States provides one of the clearest examples of how quickly data center electricity demand is growing.

A Lawrence Berkeley National Laboratory report, backed by the US Department of Energy, estimated that US data centers consumed 176 TWh of electricity in 2023. That represented around 4.4% of total US electricity consumption.

The report also estimated that US data center electricity use could rise to between 325 TWh and 580 TWh by 2028. That would represent between 6.7% and 12.0% of total US electricity consumption.

This range is important because it shows how uncertain the future still is. The lower and upper estimates depend on factors such as AI adoption, server shipments, hardware efficiency, cooling choices, and utilization rates.

But even the lower end of the projection is significant. Moving from 176 TWh in 2023 to 325 TWh in 2028 would mean a major increase in just five years. The upper estimate of 580 TWh would imply a much more dramatic transformation of US electricity demand.

This is why AI infrastructure has become a strategic issue in the US. It affects grid planning, power generation, corporate energy procurement, environmental policy, and regional development.

325 TWh to 580 TWh by 2028: Why Forecast Ranges Matter

The US forecast range of 325 TWh to 580 TWh by 2028 should not be treated as a single guaranteed outcome. It is a scenario range.

This distinction matters.

AI energy demand is difficult to forecast because several forces act simultaneously. Demand for AI services may grow faster than expected. Hardware may become more efficient. Models may become smaller and more specialized. Inference may become cheaper. Or, on the other hand, AI may be embedded in almost every digital workflow, creating much higher demand.

The wide range indicates that the future of AI infrastructure remains open. But it also shows that even conservative scenarios require serious planning.

If the lower scenario materializes, the US still needs major investment in power infrastructure. If the upper scenario materializes, the country faces a much larger energy planning challenge.

The safe conclusion is not that one exact number will happen. The safe conclusion is that AI has made data center electricity demand too large to ignore.

11x Growth in AI Server Power Density: The Rack-Level Problem

AI’s physical challenge is not only about total electricity use. It is also about power density.

The IEA reports that the power density of AI servers increased by a factor of 11 between 2020 and 2025. By 2027, it could rise by another four times.

This is one of the most important technical constraints in AI infrastructure.

A traditional data center may have enough physical space but not enough power and cooling capacity per rack. AI servers concentrate huge amounts of computing power in a small area. That creates intense local heat and requires more advanced electrical and cooling systems.

The IEA also notes that an individual server rack in an advanced data center could, by 2027, have peak power demand equivalent to that of 65 households.

This number illustrates why AI data centers are different from older facilities. The challenge is not only building more square meters of data center space. It is building facilities capable of supporting extremely dense computing environments.

In practical terms, this means more pressure on transformers, power electronics, cooling systems, backup power, and facility design.

65 Households per Rack: Why Cooling Becomes a Core Constraint

The image of an AI rack drawing peak power equivalent to that of 65 households helps explain the cooling problem.

Every watt of power a server uses eventually becomes heat. The more power concentrated in each rack, the more difficult it becomes to remove that heat safely and efficiently.

Cooling already accounts for a significant share of data center energy use. The IEA notes that cooling can account for around 7% of electricity use in efficient hyperscale data centers, but more than 30% in less efficient enterprise data centers.

AI increases the importance of cooling by altering the facility's thermal profile. High-density GPU clusters often require liquid cooling, direct-to-chip cooling or other advanced thermal management systems.

This creates several consequences:

• Older data centers may not be able to host advanced AI workloads without expensive retrofits.
• New AI data centers must be designed around cooling from the beginning.
• Operators need specialized expertise to manage liquid cooling systems.
• Cooling choices can shift pressure between electricity and water consumption.

The more AI scales, the more the data centers become a thermal engineering challenge.

7% to Over 30%: Cooling Efficiency Can Change the Energy Equation

The range from 7% to over 30% of electricity used for cooling shows how much facility design matters.

Efficient hyperscale data centers can reduce the cooling share through advanced design, better airflow, liquid cooling, climate-aware siting and operational optimization. Less efficient facilities may spend a much larger share of their electricity simply removing heat.

This is critical for AI because cooling inefficiency multiplies the energy burden. If compute demand rises quickly and cooling systems are inefficient, total electricity consumption grows faster.

However, cooling optimization is not straightforward. A system that reduces electricity use may require more water. A system that reduces water use may consume more electricity. A facility in a cooler climate may need less cooling but may still face grid or connectivity constraints.

This is why AI sustainability cannot be judged by one number alone. Electricity, water, carbon intensity, and location all matter.

Water Consumption: The Less Visible AI Resource Problem

Water is becoming one of the most politically sensitive physical constraints for AI data centers.

The reason is simple: data centers often need water for cooling, especially when evaporative systems are used. Water consumption varies widely depending on location, design, climate, and cooling technology.

United Nations University researchers have warned that AI infrastructure creates pressure not only on electricity systems but also on water, land and e-waste systems. A recent Reuters report on UNU research stated that data centers could double their power and water consumption by 2030 due to growing AI demand.

The water issue is highly local. A data center in a water-rich region does not create the same risk as a facility in a drought-prone or water-stressed region. This is why public opposition can emerge even when a company claims to have made global sustainability progress.

For communities, the question is not only whether the data center is efficient. The question is whether local water resources are being used to support global AI services.

This is one of the clearest examples of AI's physical nature. A digital service can create very real local resource conflicts.

2.5 Million Tonnes of E-Waste per Year: The Hardware Lifecycle Problem

AI infrastructure also creates an e-waste challenge.

United Nations University researchers have warned that AI infrastructure could generate up to 2.5 million tonnes of e-waste per year by 2030.

This matters because AI hardware evolves quickly. GPUs, servers, networking equipment, batteries, cooling systems, and power components may be replaced frequently as newer generations become more efficient or more powerful.

The environmental impact of AI is therefore not limited to electricity use during operation. It also includes the manufacturing, transportation, replacement, and disposal of hardware.

This creates a lifecycle problem.

A model may become more efficient over time, but if the infrastructure supporting it requires rapid hardware turnover, the environmental burden does not disappear. It shifts into supply chains, mining, manufacturing, and waste processing.

Responsible AI infrastructure must therefore include recycling, reuse, material recovery and longer hardware lifecycle planning.

More Than 800 Operators Surveyed: The Industry Itself Sees the Bottlenecks

The physical constraints are being raised not only by environmental groups or external critics. Data center operators themselves are reporting them.

Uptime Institute’s 2025 Global Data Center Survey, based on responses from more than 800 data center owners and operators, highlights rising costs, worsening power constraints, AI density challenges, staffing issues, supply chain delays and technological uncertainty.

This is important because it shows that the bottlenecks are operational rather than theoretical.

Operators are facing real constraints in securing power, upgrading facilities, hiring staff, managing supply chains and adapting to AI workloads. The industry is resilient, but it is also under pressure.

This is another reason why AI implementation may be slower or more uneven than the hype suggests. Demand for AI can grow quickly, but the infrastructure required to deliver it depends on slower physical systems.

165% by 2030: Market Forecasts Point in the Same Direction

Financial sector forecasts also point to significant growth in data center power demand.

Goldman Sachs Research has forecast that global data center power demand could increase by around 165% from 2023 levels by 2030.

This type of forecast should be treated differently from IEA or DOE-backed estimates. It is a market outlook, not a public-agency baseline. But it remains useful because it shows that energy demand growth is now central to investor expectations for AI.

The direction of travel is consistent across multiple sources: AI is increasing demand for data center capacity, and that capacity requires large amounts of power.

The exact number may vary by methodology. The trend is clear.

The Chip Constraint: AI Hardware Cannot Scale Instantly

The physical walls of AI are not limited to energy and water. AI also depends on specialized hardware.

Advanced AI systems require GPUs, AI accelerators, high-bandwidth memory, networking equipment, optical components, power electronics, and advanced packaging. These supply chains are complex and concentrated.

A shortage in any one of these components can slow deployment.

This is why AI infrastructure is vulnerable to semiconductor supply constraints, export controls, manufacturing bottlenecks, geopolitical tensions and competition for advanced memory and packaging capacity.

The key issue is that software demand can scale almost instantly, but chip supply cannot. Semiconductor manufacturing requires years of planning, enormous capital investment, and highly specialized production capacity.

AI’s global implementation therefore depends not only on model development but on the physical ability to manufacture and deliver the hardware required to run those models.

The Land Constraint: Data Centers Need Space, Permits and Local Acceptance

AI also requires land. Large data center campuses need server halls, cooling systems, substations, backup generators, batteries, security perimeters, access roads, and sometimes adjacent energy infrastructure.

This creates pressure in regions where suitable land, grid access, and connectivity overlap. The best data center locations are not just empty plots. They must combine power availability, fiber access, regulatory approval, water or cooling resources, construction capacity, and political acceptance.

This is where local resistance can become a barrier. Communities may question whether data centers consume too much electricity or water, whether they create enough permanent jobs, whether they increase local energy costs, or whether they place too much pressure on public infrastructure.

This means AI companies need more than technical capacity. They need a social license to build.

The Core Equation: AI Growth Is Now Physical Growth

The numbers point to one conclusion: AI growth is becoming physical growth.

• The global data center sector consumed around 415 TWh in 2024.
• It could reach around 945 TWh by 2030.
• Data center electricity demand could grow around 15% per year.
• US data centers used around 176 TWh in 2023.
• US data center demand could rise to 325 TWh to 580 TWh by 2028.
• AI server power density increased 11 times between 2020 and 2025.

By 2027, an advanced AI rack could draw peak power equivalent to 65 households.

Cooling can represent around 7% of electricity use in efficient hyperscale facilities but more than 30% in less efficient enterprise data centers.

AI infrastructure could generate up to 2.5 million tonnes of e-waste per year by 2030.

These numbers show that AI is no longer only a model race. It is an infrastructure race.

What the Numbers Mean for Global AI Implementation

The data suggests that AI’s global implementation will be shaped by six physical bottlenecks. The first is electricity. AI requires large volumes of reliable power.

The second is grid capacity. Electricity must be delivered to specific locations, not just generated somewhere in the system. The third is cooling. Higher power density creates more heat and requires more advanced thermal management.

The fourth is water. Cooling choices can create local water stress and political opposition. The fifth is hardware. AI depends on complex semiconductor and component supply chains.

The sixth is land and permitting. Data centers need physical space, public acceptance, and regulatory approval. These bottlenecks do not mean AI will stop growing. They mean AI growth will be more uneven, more expensive, and more dependent on infrastructure planning than many early expectations suggested.

AI Is Digital in Experience, Industrial in Reality

AI feels digital because users experience it through software. But its foundations are industrial. It runs on electricity. It depends on the chips. It produces heat. It may consume water. It requires land. It generates hardware waste. It needs power grids, cooling systems, fiber networks, transformers, substations, and skilled workers.

The central question for AI is therefore changing. The first question was: Can AI become intelligent enough to transform business and society? The next question is: can the physical world support AI at a global scale?

The answer will depend not only on better models, but on energy systems, data center design, chip manufacturing, cooling technology, water management, regulation, and public trust. 

AI may be the defining digital technology of the decade, but its limits are increasingly physical.

The post The Physical Walls of AI: The Numbers Behind the Energy, Water, and Infrastructure Challenge appeared first on Cross-Border Magazine.

Return Helper has raised US$4 million in Series A funding to expand its AI-powered cross-border e-commerce returns infrastructure. The round highlights the growing importance of reverse logistics in international online retail, where returns are no longer seen as a simple operational cost but as a strategic part of profitability, customer experience, and inventory recovery.

The funding round brought together a mix of financial and strategic investors, including Cathay Venture, MLC Ventures, Jun Yue Investment, Pegatron Venture Capital, and returning investor Colopl Next. MLC Ventures, the corporate venture arm of Mitsubishi Logistics Corporation, is especially relevant to the company’s next phase, as Return Helper looks to deepen its logistics capabilities and expand further across Japan and other international markets.

Founded as a technology-first provider of global cross-border e-commerce returns solutions, Return Helper supports merchants selling through major platforms such as Shopify, Amazon, TikTok and eBay. The company operates more than 20 overseas warehouses, works with over 30 carriers and logistics partners, and has offices across Asia, including China, Hong Kong, Japan, Singapore, and Taiwan.

The new capital will support three areas: international expansion, AI development and recommerce. Together, these priorities point to a larger shift in e-commerce logistics. As cross-border selling becomes more common, merchants increasingly need infrastructure that can manage not only outbound fulfillment but also returns, resale, recycling, refurbishment, and local recovery of returned inventory.

Why cross-border returns are becoming a strategic issue

Returns are among the most complex aspects of e-commerce. In domestic retail, they already pose challenges for customer service, warehouse operations, refunds, fraud prevention, stock visibility, and resale. In cross-border e-commerce, those problems multiply.

International returns can involve customs declarations, local regulations, fragmented carrier networks, long transit times, high shipping costs, inspection delays, and uncertain resale options. A returned product may be worth less than the cost of shipping it back to the original warehouse. For many merchants, especially those selling low-margin products internationally, the financial logic of cross-border returns can quickly break down.

That is the problem Return Helper is trying to solve. Instead of treating international returns as a linear process in which products are simply sent back to the point of origin, the company focuses on local handling and asset recovery. Returned items can be inspected, processed, resold, recycled, or redirected through local networks, helping merchants recover more value and reduce waste.

This approach is increasingly important in Europe, where e-commerce growth, sustainability expectations and regulatory pressure are all changing how retailers think about returns. Customers expect easy return options, but merchants need to control costs. Regulators and consumers are also paying more attention to waste, product destruction, and the environmental footprint of online retail. For cross-border sellers, that creates demand for smarter returns infrastructure.

AI as the next layer of returns management

Return Helper’s funding round also reflects a wider trend: the use of AI in operational e-commerce workflows. While much of the public conversation around AI in retail has focused on product recommendations, chatbots and content generation, the next wave of adoption is moving deeper into logistics and back-office operations.

Returns management is well-suited to AI because it involves repetitive yet decision-heavy processes. Every return can require a judgment: should the item be accepted, refunded, exchanged, repaired, resold, recycled, or routed to a specific warehouse? Should the merchant pay for return shipping? Is the product still suitable for resale? Is there a fraud risk? Is local recovery more profitable than international shipment?

AI agents and automated decision systems can help merchants process these decisions faster and more consistently. In Return Helper’s case, the company has stated that part of the Series A funding will go toward developing AI agents to automate return decisions. This suggests that the company is not only building logistics infrastructure but also decision intelligence for cross-border reverse logistics.

For merchants, the value is clear. Faster returns processing can improve customer satisfaction. Better routing decisions can reduce costs. Automated inspection and resale workflows can increase recovery rates. More accurate data can help teams understand why products are being returned and how to reduce future return volumes.

Recommerce moves to the center of the business model

One of the most interesting aspects of Return Helper’s positioning is its emphasis on recommerce. The company is not only helping merchants manage returns; it is helping them turn returned goods back into recoverable value.

Recommerce refers to the resale, refurbishment, or redistribution of returned, used, or excess inventory. In the context of cross-border e-commerce, recommerce can be especially valuable because returning products to the original market is often inefficient. A product sold into Europe from Asia, for example, may be too expensive to ship back. Local resale or recycling may be the better financial and environmental option.

This matters because e-commerce returns have traditionally been treated as a margin drain. A customer returns a product, the merchant loses the sale, shipping costs increase, warehouse work increases and the product may lose resale value. If returns can instead be converted into local resale opportunities, the economics change.

Return Helper’s model fits into this shift. By combining overseas warehouses, carrier partnerships, returns software, warehouse management systems and recommerce operations, the company aims to help merchants recover value from products that might otherwise become stranded, discounted or written off.

For European e-commerce, this is particularly relevant. As marketplaces, direct-to-consumer brands and cross-border sellers compete on convenience, return policies remain a major part of the customer experience. However, generous return policies are only sustainable if the operational model behind them is efficient. Recommerce gives retailers a way to balance customer expectations with profitability and sustainability.

Europe as a key growth opportunity

Although Return Helper has strong roots in Asia, Europe appears to be an increasingly important part of its growth story. The company has reported strong momentum in Europe, including significant revenue growth and a growing base of European merchant partners.

This is not surprising. Europe is one of the world’s most attractive e-commerce regions, but it is also highly fragmented. Different languages, consumer expectations, carrier networks, tax rules, and market structures make cross-border operations complex. For merchants selling into multiple European countries, managing returns through a single centralized workflow can be difficult.

At the same time, European consumers are accustomed to strong consumer protection standards and convenient post-purchase experiences. A poor returns experience can damage trust and reduce repeat purchases. For international sellers, especially those entering Europe from Asia or other regions, returns infrastructure can become a competitive barrier.

Return Helper’s European opportunity lies in helping those sellers localize the post-purchase experience. Instead of forcing every returned product through a lengthy international reverse-logistics chain, merchants can use local processing, local resale, and more flexible warehouse routing. That can reduce cost, shorten processing times, and improve customer satisfaction.

The role of strategic investors

The investor mix in Return Helper’s Series A is also significant. This is not only a software funding story. The participation of logistics-linked and corporate investors suggests that cross-border returns are becoming a priority area for the broader supply chain industry.

MLC Ventures, the corporate venture capital arm of Mitsubishi Logistics Corporation, brings a clear logistics angle to the round. Mitsubishi Logistics has deep experience in warehousing, transportation and supply chain infrastructure, which could support Return Helper’s expansion in Japan and beyond. Pegatron Venture Capital also adds a strategic dimension, given Pegatron’s wider role in technology manufacturing and supply chain ecosystems.

This type of investor base is important because returns management requires more than a software dashboard. It depends on physical infrastructure, carrier relationships, warehouse processes, local market knowledge, and operational execution. By combining software and logistics capabilities, Return Helper is positioning itself in a category where pure SaaS tools may not be enough.

What this means for merchants

For e-commerce merchants, the Return Helper funding round is part of a broader signal: returns are becoming a core part of international growth strategy. Brands that expand across borders without a clear returns model may face rising costs, slower refunds, poor customer reviews, and inventory losses.

A stronger returns strategy can help merchants in several ways. It can reduce unnecessary international shipping, improve refund speed, support exchanges, recover more value from returned items, and provide better data on product quality, sizing, customer expectations, and fraud patterns.

For marketplace sellers, the pressure is even higher. Platforms such as Amazon, TikTok Shop, eBay, and Shopify-based storefronts all create different operational requirements. Sellers that operate across multiple channels need systems that can centralize returns and connect the process with warehouses, carriers and resale channels.

Return Helper’s focus on multi-platform support is therefore important. By serving merchants who sell through Shopify, Amazon, TikTok, eBay, and other channels, the company addresses the reality of modern e-commerce: sellers rarely operate through a single storefront.

The wider market trend: from fulfillment to reverse logistics

For years, the e-commerce logistics conversation focused heavily on fulfillment speed. Retailers competed on faster delivery, better tracking, distributed warehouses, and last-mile convenience. That remains important, but the next competitive layer is reverse logistics.

As online sales volumes grow, return volumes grow with them. In categories such as fashion, footwear, electronics, and home goods, returns can represent a major operational burden. Cross-border returns add even more friction. This is why investors are paying closer attention to companies that can deliver returns more cheaply, smarter, and more profitably.

The Return Helper round shows that reverse logistics is becoming part of the core e-commerce technology stack. Returns management, warehouse software, AI decision-making, carrier orchestration, and recommerce are no longer separate functions. They are converging into a single post-purchase infrastructure layer.

This is also where AI can create measurable value. Unlike more experimental AI use cases, returns management has direct operational metrics: processing time, cost per return, resale recovery rate, refund time, customer satisfaction, fraud reduction, and inventory recovery. That makes it easier for merchants to connect technology investment with business outcomes.

Return Helper’s next challenge

The opportunity is clear, but the challenge is also significant. Cross-border returns are operationally complex, and scaling them across markets requires consistency, local knowledge, and strong execution. Return Helper will need to prove that its model can work across different regions, product categories, merchant sizes, and regulatory environments.

Its ability to integrate AI into real logistics workflows will also be important. AI-powered returns decisions can create value, but only if they are supported by accurate data, reliable warehouse processes and clear merchant rules. Automation in returns cannot be generic; it must account for product condition, category, value, customer history, regional regulation, and resale potential.

If Return Helper can combine AI decision-making with strong physical infrastructure, it could become an important player in the next phase of cross-border e-commerce logistics.

Return Helper’s US$4 million Series A funding round reflects a larger shift in global e-commerce. Cross-border returns are no longer a secondary operational issue. They are becoming a strategic battleground for profitability, customer experience, and sustainability.

By focusing on AI-powered returns management, recommerce and international logistics infrastructure, Return Helper is addressing one of the hardest problems in online retail: how to make global selling work after the sale has already happened.

For European merchants and international sellers entering Europe, the message is clear. Growth is not only about acquiring customers and shipping products across borders. It is also about building the infrastructure to manage what comes back. As return volumes rise and margins remain under pressure, companies that can turn reverse logistics into recoverable value may become increasingly important to the future of cross-border e-commerce.

The post Return Helper Raises US$4M Series A to Scale AI-Powered Cross-Border E-Commerce Returns appeared first on Cross-Border Magazine.