An interview with AWS CEO Adam Selipsky on generative AI, the Microsoft-OpenAI deal, competition among cloud providers offering generative AI services, and more
Richard Waters / Financial Times :
Context & Ripple Effects
This interview lands mid-campaign for AWS. In April, the company opened its Bedrock platform to models from Anthropic, Stability AI, AI21 Labs, and its own — an explicit bid to be the neutral host for generative AI rather than betting on one lab. Product VP Matt Wood had already framed the pitch around curating the best models for AWS customers, and Selipsky followed with a Verge Q&A covering Nvidia dependence and the custom Trainium and Inferentia chips.
What changed by this Financial Times conversation is that Selipsky is now answering for the competitive structure head-on: Microsoft's OpenAI deal gave a rival cloud an exclusive flagship model, and Selipsky's counterargument is that neutrality plus a model library beats a single-lab lock-in. The same themes resurface months later when he sits down again ahead of re:Invent, making this interview an early statement of the position AWS would carry through the year.
First-order effects
- Selipsky puts AWS on record against the Microsoft-OpenAI pairing, telling enterprise buyers that Bedrock's multi-model catalog — Anthropic, Stability AI, AI21 Labs, and Amazon's own models — is the safer procurement path than a single-vendor exclusive.
Second-order effects
- Microsoft is forced to defend its OpenAI exclusivity as a feature rather than a constraint, while Anthropic, Stability AI, and AI21 Labs gain leverage from having a second major cloud distribution channel beyond any one partner.
Third-order effects
- If the pattern holds, cloud competition shifts from raw compute pricing to model portfolios and in-house silicon — Selipsky's Trainium/Inferentia push against Nvidia dependence points toward each hyperscaler bundling its own chips with its own AI stack.
The trend: Cloud providers are splitting into two camps — vertically integrated model owners like Microsoft and neutral multi-model platforms like AWS — with custom AI silicon becoming the next differentiator.