TLDR
- Peter DeSantis assumes Amazon’s AI leadership with an emphasis on affordability over rapid deployment.
- The company leverages proprietary Trainium and Inferentia chips to reduce AI development expenses.
- Amazon’s Nova model shows room for improvement in benchmarks, while Nova 2 promises enhanced capabilities.
- AMZN shares have fallen approximately 8% since January amid worries over $200 billion infrastructure investments.
- David Luan, leader of Amazon’s AGI Lab, revealed his exit on Tuesday.
Amazon’s approach to artificial intelligence centers on affordability as a competitive advantage.
Peter DeSantis, who recently assumed leadership of Amazon’s AI division, champions a cost-focused methodology. His fundamental thesis addresses the economic barriers preventing widespread AI adoption.
“AI has a cost problem,” DeSantis said. “If we ultimately want AI to transform everything, the costs have to be different.”
DeSantis inherited AI responsibilities this past December following the departure of former chief AI scientist Rohit Prasad. His 28-year tenure at Amazon includes instrumental roles in building AWS and developing the company’s semiconductor operations.
Amazon’s market valuation has declined approximately 8% since the start of January. Wall Street expresses concern regarding the company’s ambitious $200 billion capital expenditure plan for this year — predominantly allocated to AI infrastructure — while financial analysts forecast roughly $9 billion in cash consumption during Q1.
DeSantis faces considerable pressure to deliver results.
Amazon’s Chip Strategy
The strategy revolves around Amazon’s proprietary semiconductor technology: Trainium chips for model training and Inferentia chips for deployment. The company reports these chips deliver up to 50% cost savings compared to competing solutions.
“If we can build our models on our chips, we can build them at a fraction of the cost of a pure-play AI model provider,” DeSantis said.
This economic advantage has begun winning over clients. Nimbus Therapeutics, a Boston-based pharmaceutical research company, discovered Amazon’s Nova model delivered accuracy comparable to Anthropic’s Claude while costing one-tenth the price.
Amazon provides Nova Forge as well, enabling businesses to develop customized AI solutions rather than purchasing premium alternatives like ChatGPT or Gemini.
Amazon’s primary Nova model has underperformed competitors in independent testing. The organization maintains that Nova 2 demonstrates superior performance, though independent benchmark verification remains unavailable.
Amazon’s response to generative AI came later than competitors. Following ChatGPT’s debut in late 2022, Amazon convened urgent strategy sessions to formulate its response.
“Amazon was slower to realize the importance of generative AI,” said Lloyd Walmsley, senior analyst at Mizuho.
Talent and Competition
Amazon confronts human capital obstacles as well. Compensation for software engineers and research scientists falls below rates offered by Meta, OpenAI, Apple, and Anthropic, based on Levels.fyi data. The organization eliminated approximately 30,000 white-collar positions through two separate workforce reductions.
David Luan, who directed Amazon’s AGI Lab, confirmed his departure on Tuesday. The laboratory will maintain operations under DeSantis’s supervision.
DeSantis says he’s not chasing splashy model releases like OpenAI and Anthropic. He called frequent launches “kind of how you stay in the news” but said they don’t necessarily move the needle for customers.
Amazon reports that Nova model variants now process over 70% of Alexa queries. The Rufus shopping assistant chatbot served more than 300 million users throughout 2025.
DeSantis addressed investor apprehension about capital spending, referencing historical skepticism Amazon encountered during its initial investments in physical retail infrastructure and AWS data center expansion.
Amazon’s AGI Lab, dedicated to developing AI agents, remains within DeSantis’s organizational structure following Luan’s announcement.
