OpenAI sees 'vertical wall' demand

AI demand is no longer the question. Capacity is. That was the real signal in Sarah Friar’s latest defense of OpenAI’s business — not just that people want the products, but that the company seems more constrained by infrastructure than by customers. At the same time, retail is showing the other half of the AI story. People will happily use AI to research what to buy, but they still get weird about letting it spend money for them. ### What did OpenAI actually say? Friar, OpenAI’s CFO, pushed back on reports that the company had missed internal revenue and user targets. Her point was basically that headline misses can be misleading inside a young company with moving forecasts. She said OpenAI is still beating its plan “at the highest level” and described demand as a “vertical wall” — meaning the problem is not finding users, it’s serving them. (bloomberg.com) ### Why does that matter? Because it changes how you read the AI market. If OpenAI were struggling to convert interest into paid usage, that would suggest the generative AI boom is cooling. But Friar’s framing points the other way — the appetite is there, and the harder problem is enough comput(bloomberg.com)re partners and chip suppliers after reports about softer internal targets surfaced. (bloomberg.com) ### So is AI commerce booming too? Yes — but in a narrower way than the hype suggests. Consumers are increasingly using AI as a shopping layer: to compare products, summarize reviews, and narrow choices. One recent retail snapshot said 35% of U.S. shoppers now use AI assistants, up from 12% a year earlier. Another said 45% used AI tools during the last holiday season. That is real behavior change. (adyen.com) ### Then where’s the ceiling? Payments. Trust. Handing over the final click. That’s the catch. The current pattern is that people like AI for discovery, but hesitate when the tool starts acting like an agent with permission to buy. MarTech’s write-up on the space put it plainly: research behavior is rising, but payment hesitation is still the line man(adyen.com)n — shoppers are much more comfortable with AI helping than fully automating the whole purchase flow. (martech.org) ### What is Amazon trying? Amazon is pushing right into that gap. This week it expanded its AI-generated shopping “experts” so users can ask follow-up questions inside the app, instead of just hearing a fixed audio summary. The feature builds on “Hear the highlights” and turns it into more of a conversational product guide(martech.org) distinction matters. (mobilesyrup.com) ### Why not let the agent just buy? Because buying is not just a technical step. It is a trust step. Once an AI starts choosing sellers, handling payment credentials, or making trade-offs on price versus quality, the user needs to believe the system is aligned with them — not just fluent. That is much harde(mobilesyrup.com) the last mile is social and behavioral, not just product design. (martech.org) ### Is there one story tying this together? Basically, yes. AI has crossed the line from novelty to demand engine in some categories. OpenAI is saying the bottleneck is supply. Retailers are learning the bottleneck is trust. Those are different constraints, but they point to the same phase of the market: the easy part was (martech.org)ough to let it act. (bloomberg.com) ### Bottom line The AI boom looks real. But it is maturing into two harder businesses — infrastructure on one side, trust on the other. OpenAI’s “vertical wall” line captures the first problem. AI shopping captures the second. Whoever solves both gets more than usage — they get durable control of how people work and buy. (bloomberg.com)