OpenAI misses targets, AWS widens access

OpenAI has two stories colliding at once. One is about demand — growth has reportedly cooled enough that internal revenue and user targets were missed in recent months. The other is about distribution — on April 28, OpenAI and Amazon Web Services widened their partnership so AWS customers can buy and run OpenAI models inside Amazon’s cloud tools. Put simply, OpenAI is trying to prove it can keep growing fast enough to justify enormous infrastructure spending, while also making itself easier to buy. (money.usnews.com) ### What actually changed on AWS? AWS added OpenAI models to Amazon Bedrock in limited preview, alongside Codex and a new Bedrock Managed Agents offering powered by OpenAI. That means companies already standardized on AWS can use OpenAI through familiar procurement, security, identity, and governance systems instead of building a separate buying path. AWS also said this usage can count toward existing cloud commitments — a very practical detail for enterprise buyers. (aws.amazon.com) ### Why is that a big deal? Because cloud distribution is often the sale. A lot of large companies do not want a new vendor workflow, a new compliance review, or a new billing relationship just to try a model. Bedrock already acts as a marketplace layer for model access, so OpenAI showing up there lowers friction fast. The catch is that this is also a power shift — OpenAI is no longer as tightly bound to Microsoft as its main route into enterprises. (aws.amazon.com)s spending. Reuters, citing the Wall Street Journal, said OpenAI fell short of its own goals for new users and revenue, which stirred internal concern about whether future data-center commitments can be supported if growth stays softer. CNBC’s writeup says the same shortfall raised questions around funding massive compute plans. That does not mean OpenAI is shrinking — it means the curve may not be rising as steeply as the company expected. (money.usn([aws.amazon.com)2026-04-27/openai-falls-short-of-revenue-and-user-targets-as-it-races-toward-ipo-wsj-reports)) ### Why do data centers matter so much here? Frontier AI is brutally expensive to run. Training costs money, but inference — serving millions of user requests and agent tasks — also eats hardware, power, and networking. OpenAI has already signed very large infrastructure deals with AWS, including a previously announced multi-year partnership framed by Amazon as a $38 billion commitment for compute capacity. If revenue growth slips even a bit, those obligations look heavier. (aboutamazon.com) ### Is this just about Microsoft losing exclusivity? Not just that, but it is part of the story. OpenAI told customers and staff earlier this month that Microsoft had limited its ability to reach some enterprise clients, and it pitched Amazon as a way to widen market access. AP also noted the AWS expansion came right after OpenAI loosened ties with Microsoft. Basically, OpenAI wants more than one major pipe into the market. (cnbc.com)nswer to Google, Anthropic, and Microsoft in enterprise AI. Bedrock works best when customers can compare multiple top-end models without leaving AWS. Adding OpenAI makes Bedrock more complete, and managed agents give Amazon a way to sell not just models but higher-level automation products. That is stickier — once workflows and internal tools are built around them, customers are less likely to move. (aboutamazon.com)onomics. The AWS deal helps on reach immediately — more buyers, easier procurement, fewer Microsoft bottlenecks. But wider access does not magically fix unit economics if model serving remains expensive and customer growth undershoots plan. Multi-cloud distribution can boost sales. It can also expose whether demand is truly deep enough to carry the infrastructure bill. (money.usnews.com))) ### Bottom line? This is OpenAI trying to solve two problems with one move. AWS widens the funnel. But the harder question is still whether revenue can catch up to the cost of building and running frontier AI at industrial scale. (money.usnews.com)