Q&A with AWS Product VP Matt Wood on the generative AI boom, the escalating AI battle between the tech giants, curating the best AI models for AWS, and more
“It really helps if you think of them less as this very loaded term of artificial intelligence, and more just applied statistics.” Twitter: Ben Smith / @semaforben : Amazon makes “a bet that the future of AI is an explosion of highly specialized generative AI models, many of which will be open source” https://www.semafor.com/... Reed Albergotti / @reedalbergotti : Some people think there may only be room for a small handful of foundation models in the AI industry. Amazon thinks there's room for thousands. Here's my interview with AWS product boss Matt Wood. https://www.semafor.com/...
Context & Ripple Effects
This interview is AWS articulating, mid-2023, the platform strategy it had already begun executing: since April, when AWS started offering customers access to LLMs from Anthropic, Stability AI, and AI21 Labs, it has positioned itself as a neutral host for other companies' models rather than a single-model shop like Microsoft with OpenAI. Matt Wood's framing here — an 'explosion' of highly specialized, largely open-source models, with AWS curating the best — is the thesis behind that move.
The bet matters because it defines how AWS chooses to fight the escalating giant-vs-giant AI battle Wood describes: not by owning one frontier model, but by controlling distribution and selection across many. That logic carries through AWS leadership's subsequent public positioning, from Adam Selipsky on the Microsoft-OpenAI deal and cloud competition to Matt Garman drawing parallels between early AWS and AI years later.
First-order effects
- Model partners like Anthropic, Stability AI, and AI21 Labs gain a major enterprise distribution channel through AWS's curation layer, while AWS customers get a vetted menu instead of building direct vendor relationships.
- Wood's open-source-heavy bet puts AWS explicitly at odds with the closed, exclusive-partnership model Microsoft built around OpenAI, sharpening the competitive split among the big three clouds.
Second-order effects
- If specialized open-source models proliferate, rivals are pushed toward the same aggregator posture — competing on which cloud surfaces and fine-tunes the widest model catalog, turning model choice itself into a pricing and lock-in battleground.
- A many-models strategy raises AWS's dependence on underlying compute suppliers like Nvidia even as it hedges with its own chips, making silicon cost and availability a second front in the same battle.
Third-order effects
- If the pattern holds, value in generative AI migrates from whoever trains a frontier model to whoever controls distribution and curation — echoing the dynamic Garman later invoked by comparing AI to early AWS infrastructure.
- An industry organized around thousands of open-source specialists would make open model ecosystems strategic infrastructure, with cloud platforms acting as de facto gatekeepers of what enterprises can reach.
The trend: Cloud providers are shifting from exclusive single-model alliances to competing as curated distribution layers for many specialized, increasingly open-source AI models.