Jassy taps 27-year Amazon veteran to run AGI org, which is now definitely a thing that exists
Summary
Amazon has formalised an “AGI” organisation that reports directly to CEO Andy Jassy. Peter DeSantis, a 27-year Amazon veteran and formerly SVP of Utility Computing, will lead the new unit. He brings with him teams including the reworked Annapurna Labs silicon group (Graviton, Nitro, Trainium, Inferentia) and Amazon’s quantum computing effort. Pieter Abbeel has been named head of the frontier model research team within AGI. The move signals Amazon’s strategic bet on vertical integration—optimising models, chips and cloud software together—and positions its huge robotics and fulfilment footprint as a potential testbed for embodied AI.
Key Points
- Andy Jassy has created a central AGI organisation that reports to him directly.
- Peter DeSantis will lead the new AGI organisation, taking teams from AWS and other areas.
- Annapurna Labs (Amazon’s silicon teams) and the quantum computing group are folded into AGI.
- Pieter Abbeel is named head of the frontier model research team, signalling serious research hires and capability building.
- The strategy emphasises end-to-end optimisation across models, chips and cloud infrastructure—vertical integration akin to Apple’s playbook for AI.
- Amazon’s vast deployed robotics fleet and operational scale could provide a unique advantage if embodied AI becomes important.
Context and relevance
This reorganisation is Amazon signalling that advanced AI is a company-wide strategic priority, not just an AWS feature set. By bringing silicon, model research and long-term bets like quantum under one leader, Amazon aims to compete more effectively with Microsoft (which relies on third-party GPUs) and Google (which has separate TPU and model teams).
Including quantum in the same organisation suggests a willingness to invest patiently and strategically rather than chase only near-term results. Naming Pieter Abbeel to lead frontier model work highlights Amazon’s intent to build serious base-model capability, while its robotics and logistics businesses offer an embodied AI test environment few rivals can match.
Why should I read this?
Short version: Amazon just made AGI an actual organisation, put a hardware and long-term-tech heavy-weight in charge, and handed model leadership to a respected researcher. If you follow cloud, AI supply chains, or enterprise AI strategy, this changes the competitive map — and could mean Amazon is about to try to control the stack from silicon to service. Worth five minutes of your time, at least.