Unpacking Commerce | March 2026
Emerging trends, analysis, and more.
Hello, it’s Alexandre from Spring Invest, a European investment fund that is shaping the future of commerce. Welcome to the latest edition of Unpacking Commerce, our newsletter about emerging trends in retail, brands, and new commerce.
🫧 The mirage collapses: OpenAI retreats from checkout, and agentic commerce reveals its true shape
Six months ago, OpenAI launched Instant Checkout inside ChatGPT with the ambition of turning the world’s most popular AI chatbot into a transactional commerce platform. Last week, The Information reported that OpenAI is scaling back the feature. Purchases will now be routed through third-party retailer apps - Instacart, Target, Expedia - rather than completed natively in chat. The numbers are damning: out of Shopify’s millions of merchants, roughly a dozen had actually gone live with ChatGPT checkouts.
The timing matters as much as the decision itself. This retreat arrives less than one week after OpenAI announced a $50 billion strategic partnership with Amazon, part of a record $110 billion funding round at a $730 billion pre-money valuation. The deal includes custom AI models for Amazon’s consumer products, 2 gigawatts of Trainium compute capacity, and AWS as the exclusive third-party cloud distributor for OpenAI Frontier. Eric Seufert, whose Agentic commerce is a mirage analysis has been the sharpest structural critique of this space, argues that the Amazon partnership likely sealed the coffin. If OpenAI is aligning with Amazon’s advertising stack, an affiliate checkout model becomes strategically redundant.
The economic problems with Instant Checkout were structural, not incidental. OpenAI’s 4% transaction fee was destructive for merchants running on 3–8% net margins - typical for mid-market Shopify CPG sellers. Unlike advertising, where a merchant controls the margin they forfeit through a bid, a flat commission is an exogenous cost imposed on an unpredictable volume of sales. Advertising allocates demand ex ante through pricing mechanisms; affiliate models capture it ex post through static percentages. One scales; the other breaks.
The market’s reaction was immediate. Booking Holdings and Expedia shares jumped roughly 10% on the news. Mizuho moved its top internet pick to Booking, explicitly citing the ChatGPT pullback. Travel - which seemed like the most natural vertical for conversational checkout - turned out to illustrate the problem most clearly: real-time pricing, complex inventory, regulatory compliance, and entrenched consumer trust in existing checkout flows all conspired against native AI transactions.
Two parallel developments reveal where value is actually accruing. On March 2, Criteo announced it is the first ad-tech partner integrating with OpenAI’s advertising pilot in ChatGPT, activating $4 billion in annual media spend across 17,000 advertisers. In parallel, Amazon is reportedly exploring technology to help third-party apps sell ads within chatbot interfaces, extending its $68.6 billion ad business into conversational AI surfaces. The direction is unmistakable: the monetization of AI-driven commerce is converging on advertising, not transactions.
Amazon’s own numbers frame the gravitational pull. Ad revenue grew 22% in Q4 2025 to $21.3 billion. Rufus, its on-platform shopping agent, was used by 300 million customers last year, with users 60% more likely to complete purchases. Amazon has no incentive to let third-party agents disintermediate this flywheel - and every incentive to push its advertising infrastructure outward into the chatbot ecosystem it now partly owns.
Morgan Stanley estimates agentic commerce could represent $190–385 billion in U.S. e-commerce spending by 2030, capturing 10–20% of market share. But who captures the economics? If the transactional model is failing and the advertising model favors incumbents with first-party data, the investable surface area may be narrower than the headline suggests.
➡️For European commerce founders and investors, the question is worth sitting with: where in this stack does value accrue outside the walled gardens - structured product data, discovery optimization, retail media enablement? Or is the entire value chain destined to consolidate around a handful of platforms?
🏗️ Shopify’s quiet platform shift: from storefront to commerce protocol layer
While the checkout retreat grabbed headlines, the more consequential structural move may be Shopify’s. In January, Shopify and Google co-launched the Universal Commerce Protocol (UCP) at NRF 2026 - an open standard that allows AI agents to discover products, negotiate checkout flows, and complete purchases across any merchant. The protocol is already endorsed by over 20 partners, including Walmart, Target, Visa, Mastercard, and Zalando.
Shopify’s layering strategy is deliberate. It activated a default MCP endpoint on every store on its platform; launched Agentic Storefronts in December to manage presence across ChatGPT, Perplexity, Copilot, and Google AI Mode from a single admin toggle; and critically, opened its product catalog to non-Shopify merchants through a new Agentic plan. Brands on WooCommerce, Magento, or custom stacks can now list products in the Shopify Catalog and sell through AI channels - without running a Shopify store.
This is a repositioning from storefront provider to commerce infrastructure rail. The Catalog, powered by specialized LLMs that categorize and enrich product data, is the moat. According to data shared at Shopify’s agentic commerce webinar, orders originating from AI searches increased 15x between January 2025 and January 2026, with higher average order values than direct site traffic.
The contrast with Amazon is instructive. Amazon restricts agentic access at the point of discovery - it wants to retain advertising primacy. Shopify restricts at the point of purchase - it wants to retain transaction primacy through Shop Pay. Both protect their core economics, but Shopify’s approach creates a more porous surface for third-party agents, which is why it has emerged as the default infrastructure partner for ChatGPT, Copilot, and Gemini commerce integrations.
➡️The practical implication is that an AI-mediated discovery layer is forming fast, and the tooling between a brand’s catalog and an agent’s recommendation engine - data enrichment, protocol compliance, attribution - is where new infrastructure gets built. Restructuring product data specifically for LLMs (enriching feeds with FAQs, sales arguments, and visual variations) is one illustration.
📢 ChatGPT ads go live: the predictable arc from affiliate to advertising
On January 16, OpenAI officially launched advertising in ChatGPT for its Free and Go tiers. At the same time, it walked back in-chat checkout. The juxtaposition tells the story: OpenAI is deprioritizing commerce transactions precisely as it ramps up commerce advertising.
The early signals are worth tracking. Criteo’s analysis of 500 U.S. retailers in February 2026 found that users referred from LLM platforms convert at approximately 1.5x the rate of other referral channels. Early advertiser sightings include Expedia, Best Buy, Target, Ford, and Adobe. An Adthena analysis of over 500 prompts found ads in approximately 0.8% of responses - a deliberately conservative density, presumably to protect user trust while the format matures.
Criteo has moved first as an ad-tech intermediary. But Amazon’s reported exploration of powering chatbot ads via its Publisher Services suggests the real competition may pit established ad infrastructure providers against each other, not AI startups against incumbents. OpenAI itself projects $46 billion in revenue from non-paying users by 2030 through ads and commissions - a target that only makes sense if the advertising model works at scale, where the affiliate model could not.
The 1.5x conversion premium on LLM-referred traffic, if it holds, will draw ad budgets fast. The open question is therefore: does this new surface become an extension of existing performance marketing playbooks, or a fundamentally different channel that rewards different capabilities? And in the ad-tech ecosystem, where does the margin sit - with the AI platform, the demand-side intermediary, or the commerce data layer that makes the recommendations actually work?
What we’re reading
Agentic commerce is a mirage (part 2) - Eric Seufert, Mobile Dev Memo - The sharpest structural analysis of why OpenAI’s Instant Checkout retreat was inevitable, and why Amazon’s advertising flywheel dictates the shape of AI commerce.
Building the Universal Commerce Protocol - Shopify Engineering - The technical architecture behind UCP: layered protocol design, merchant-agent negotiation, and why Shopify modeled it after TCP/IP rather than building a monolithic standard.
Agentic Commerce Impact Could Reach $385 Billion by 2030 - Morgan Stanley Research - The bull case for agentic commerce as a $385B U.S. market by 2030, with grocery and CPG as the first major category to tip.
OpenAI is killing Instant Checkout. Don’t dance on the grave of agentic shopping yet - Kiri Masters, The Drum - Alibaba’s Qwen app already completes purchases at scale inside a single conversational interface because it owns the full stack. The behavior works in China; the question is infrastructure ownership, not consumer appetite.
👍 If you like Unpacking Commerce and want to help it grow, please share this newsletter with your colleagues, followers, and friends. If you hate it, then send it to your enemies. Have a great day, and see you soon!
About us
Spring Invest is a European investment fund dedicated to companies that are shaping the future of commerce. We champion doers who build innovative companies making commerce better, from enabling technologies to new commerce models and everything in between. More info about our investment thesis 👉 here.
