Amazon (AMZN) Stock: New AI Chief Plans to Utilize In-House Chips to Cut Costs

Posted on by

TLDR

  • Amazon’s new AI leader Peter DeSantis is leaning into a low-cost approach to compete in the AI race.
  • Amazon intends to utilize its in-house Trainium and Inferentia chips to develop AI models at a lower cost than competitors.
  • The company’s Nova model has fallen behind rivals in benchmark tests, though Nova 2 is reported to be more competitive.
  • Amazon’s stock has declined approximately 8% since January due to investor worries about $200 billion in capital expenditures.
  • David Luan, head of Amazon’s AGI Lab, revealed his resignation on Tuesday.

Amazon has a fresh AI strategy: outpace competitors on price, not just performance.

AMZN Stock Card

Peter DeSantis, Amazon’s recently named AI chief, is advancing a cost-focused strategy. His key point is straightforward: AI is too costly, and Amazon can alter that.

“AI has a cost problem,” DeSantis said. “If we ultimately want AI to transform everything, the costs have to be different.”

DeSantis assumed AI duties in December following the departure of former chief AI scientist Rohit Prasad. He’s a 28-year Amazon veteran who assisted in building AWS and its chip businesses.

Amazon’s stock is down about 8% since January. Investors are concerned about the company’s plan to spend $200 billion on capital expenditures this year — primarily on AI infrastructure — while analysts estimate Amazon will use up roughly $9 billion in cash during Q1 alone.

The pressure on DeSantis is significant.

Amazon’s Chip Strategy

The strategy revolves around Amazon’s homemade chips: Trainium for training models and Inferentia for deploying them. Amazon asserts these chips are up to 50% less expensive than similar products 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.

That cost benefit is already drawing in some clients. Boston-based drug discovery company Nimbus Therapeutics discovered Amazon’s Nova model delivered results as accurate as at one-tenth the cost.

Amazon also provides Nova Forge, which allows enterprise clients to create tailored AI models instead of paying for premium tools like ChatGPT or Gemini.

Amazon’s flagship Nova model has trailed competitors in independent benchmarks. The company states Nova 2 performs better, though it hasn’t published third-party benchmark data to support that claim.

Amazon was also late to embrace generative AI. When ChatGPT launched in late 2022, Amazon rushed to react, holding emergency meetings to develop a strategy.

“Amazon was slower to realize the importance of generative AI,” said Lloyd Walmsley, senior analyst at Mizuho.

Talent and Competition

Amazon also faces a talent issue. Its base salaries for software engineers and research scientists lag behind , OpenAI, Apple, and Anthropic, per Levels.fyi. The company also recently laid off approximately 30,000 white-collar employees across two rounds of cuts.

On Tuesday, David Luan, head of Amazon’s AGI Lab, stated he was departing. The lab will keep operating under DeSantis.

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 over 70% of Alexa queries are now processed by versions of its Nova model. More than 300 million people used its Rufus shopping chatbot in 2025.

DeSantis recognized investor worries about spending but pushed back, comparing it to the doubt Amazon faced when it first invested in physical retail and later AWS data centers.

Amazon’s AGI Lab, which focuses on AI agents, will still report to DeSantis after Luan’s departure.