Amazon’s AI Superstructure: DeSantis Takes Command of Models, Chips, and Cloud

On December 17, 2025, Andy Jassy announced one of Amazon’s most significant organizational changes in years: Peter DeSantis, a 27-year Amazon veteran, will lead a new unified organization combining AI models (Nova and AGI), the economics of AI compute infrastructure — /”>custom silicon (Graviton, Trainium, Nitro), and quantum computing – all reporting directly to the CEO.

This Isn’t Just a Reorg

It’s Amazon’s declaration that the next chapter of AI will be won not through model innovation alone, but through hardware-software co-design at massive scale. While OpenAI pursues research breakthroughs and Google consolidates under DeepMind, Amazon is betting that infrastructure mastery – optimizing across models, chips, and cloud together – will be the decisive advantage.

The Leadership Changes

Peter DeSantis becomes the AI Czar, reporting directly to Jassy. His mandate: optimize across the full stack – models, chips, and cloud infrastructure. DeSantis isn’t a research scientist. He’s an infrastructure builder with what Jassy calls “unusual technical depth” and a “track record of solving problems at the edge of what’s technically possible.”

Rohit Prasad departs after building Alexa from idea to hundreds of millions of users, then leading Nova’s creation. His exit signals a pivot from research-first to execution-first leadership.

Pieter Abbeel (Covariant co-founder, robotics AI pioneer) leads frontier model — as explored in the intelligence factory race between AI labs — research within AGI, continuing robotics work alongside – signaling Amazon’s interest in embodied AI.

Why DeSantis Is The Signal

The DeSantis timeline tells the story: he built EC2 (the foundation of cloud computing), acquired Annapurna Labs (custom silicon), led AWS Infrastructure, and created Amazon’s chip portfolio. The throughline is execution, not research.

Model, chip, and quantum under one roof allows for end-to-end optimization. Amazon is saying the next chapter of the AI race will not just be about raw model prowess, but the hard engineering problem of making AI fast, cheap, and reliable at global scale.

Key Takeaway

When Prasad (the researcher) leaves and DeSantis (the operator) takes over, the organizational culture shifts. As enterprise AI transforms from software to substrate, Amazon is betting execution beats innovation.

Source: Amazon’s AI Superstructure on The Business Engineer