Key Takeaways
- Peter DeSantis, Amazon’s new AI leader, is pursuing a cost-focused approach to win the artificial intelligence competition.
- The company is leveraging proprietary Trainium and Inferentia chips to develop AI systems at lower costs than competitors.
- While the Nova model hasn’t matched rivals in performance tests, Amazon claims Nova 2 shows significant improvements.
- Shares of Amazon have declined approximately 8% since the start of the year amid concerns over $200 billion in infrastructure investments.
- David Luan, who led Amazon’s AGI Lab, confirmed his exit from the company this week.
Amazon is charting a distinct course in artificial intelligence: winning through affordability rather than pure technological superiority.
Under the leadership of Peter DeSantis, Amazon’s recently named AI head, the tech giant is embracing a price-centric approach. His thesis is straightforward — artificial intelligence remains prohibitively expensive, and Amazon has the capability to fix that.
“AI has a cost problem,” DeSantis said. “If we ultimately want AI to transform everything, the costs have to be different.”
DeSantis assumed control of AI initiatives this past December after Rohit Prasad, the former chief AI scientist, left the company. With 28 years at Amazon, DeSantis played instrumental roles in developing AWS and the company’s semiconductor division.
Amazon’s shares have fallen about 8% since the beginning of January. Wall Street is growing anxious over the company’s ambitious $200 billion capital investment plan for this year — primarily allocated to AI infrastructure — with analysts estimating around $9 billion in cash burn during the first quarter alone.
The stakes for DeSantis couldn’t be higher.
The Custom Silicon Approach
Amazon’s strategy revolves around its internally developed semiconductors: Trainium chips designed for model training and Inferentia chips optimized for inference operations. According to Amazon, these proprietary chips deliver up to 50% cost savings compared to equivalent solutions from competitors.
“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 pricing advantage is already resonating with certain clients. Nimbus Therapeutics, a Boston-based pharmaceutical research company, discovered that Amazon’s Nova model delivered results matching Anthropic’s Claude in accuracy while costing just one-tenth the price.
Additionally, Amazon provides Nova Forge, enabling corporate clients to develop tailored AI solutions instead of subscribing to premium services like ChatGPT or Gemini.
Despite these advantages, Amazon’s flagship Nova model has underperformed competitors in independent performance evaluations. While the company maintains that Nova 2 delivers superior results, it hasn’t yet provided third-party benchmark verification.
Amazon was also caught flat-footed when generative AI exploded onto the scene. Following ChatGPT’s debut in late 2022, Amazon conducted urgent internal meetings to formulate a response strategy.
“Amazon was slower to realize the importance of generative AI,” said Lloyd Walmsley, senior analyst at Mizuho.
The War for Talent
Recruiting and retaining top-tier talent presents another obstacle for Amazon. Compensation packages for software engineers and research scientists at Amazon fall below those offered by Meta, OpenAI, Apple, and Anthropic, per Levels.fyi data. The company also eliminated approximately 30,000 corporate positions through two separate workforce reductions.
This Tuesday brought news of David Luan’s departure, who headed Amazon’s AGI Lab. The division will remain operational under DeSantis’s oversight.
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.
According to Amazon, variations of its Nova model now process over 70% of all Alexa voice queries. The company’s Rufus shopping assistant chatbot attracted more than 300 million users throughout 2025.
While acknowledging Wall Street’s apprehension about massive capital expenditures, DeSantis defended the investment strategy, referencing similar skepticism Amazon encountered when it first expanded into brick-and-mortar retail and subsequently built out AWS infrastructure.
Following Luan’s departure, Amazon’s AGI Lab, which concentrates on developing AI agents, remains under DeSantis’s leadership.
