Amazon’s Cost-First AI Strategy: Can In-House Chips Counter OpenAI’s Dominance?

Key Takeaways

• Peter DeSantis, Amazon’s newly named AI leader, is prioritizing affordability over cutting-edge performance in the company’s AI approach.
• The tech giant intends to leverage proprietary Trainium and Inferentia processors to develop AI systems at lower costs than competitors.
• While Amazon’s Nova model trails in performance benchmarks, the upcoming Nova 2 promises improved results.
• AMZN shares have declined approximately 8% since the start of January amid worries over planned $200 billion capital investments.
• David Luan, who led Amazon’s AGI Lab, revealed his resignation this week.

Amazon is charting a distinct path in artificial intelligence: winning through affordability rather than pure technical superiority.

Amazon.com, Inc., AMZN

Peter DeSantis, recently elevated to lead Amazon’s AI initiatives, is championing an approach centered on economics. His fundamental premise is straightforward — artificial intelligence carries prohibitive costs, and Amazon has the solution.

DeSantis assumed control of AI operations last December following the exit of former chief AI scientist Rohit Prasad. A 28-year company veteran, he played instrumental roles in developing AWS and Amazon’s semiconductor division.

Amazon’s shares have declined about 8% since the new year began. Market participants are growing anxious over the company’s blueprint to allocate $200 billion toward capital expenditures throughout the year — predominantly for AI infrastructure — as Wall Street analysts forecast Amazon will consume approximately $9 billion in cash during Q1 alone.

The stakes for DeSantis couldn’t be higher.

The Semiconductor Gambit

The strategy revolves around Amazon’s proprietary silicon: Trainium processors for model training and Inferentia chips for inference workloads. Amazon asserts these components deliver up to 50% cost savings compared to equivalent alternatives from competitors.

This economic edge is already drawing interest from clients. Nimbus Therapeutics, a Boston-headquartered pharmaceutical discovery company, discovered Amazon’s Nova model delivered results matching Anthropic’s Claude in accuracy while costing just one-tenth the price.

Amazon additionally provides Nova Forge, enabling business clients to construct tailored AI systems instead of subscribing to premium solutions like ChatGPT or Gemini.

Amazon’s primary Nova offering has underperformed competitors in independent performance evaluations. The company maintains that Nova 2 delivers superior results, although publicly available third-party benchmark validation remains unavailable.

Amazon also reacted slowly to the generative AI revolution. Following ChatGPT’s debut in late 2022, Amazon rushed to formulate a response, convening urgent strategic sessions.

Workforce Challenges and Market Rivals

Amazon confronts significant talent retention obstacles. Compensation packages for software engineers and research scientists fall behind Meta, OpenAI, Apple, and Anthropic, based on Levels.fyi data. The organization also eliminated roughly 30,000 white-collar positions through two separate workforce reductions.

This Tuesday, David Luan, who headed Amazon’s AGI Lab, confirmed his departure. The laboratory will maintain operations under DeSantis’s leadership.

Amazon reports that Nova model variations now power more than 70% of Alexa interactions. The company’s Rufus shopping assistant chatbot served over 300 million users throughout 2025.

DeSantis recognized investor apprehension regarding expenditures but defended the strategy, referencing similar doubt Amazon encountered during its initial physical retail expansion and subsequent AWS data center investments.

Amazon’s AGI Lab, dedicated to developing AI agents, remains under DeSantis’s oversight following Luan’s announcement.

The post Amazon’s Cost-First AI Strategy: Can In-House Chips Counter OpenAI’s Dominance? appeared first on Blockonomi.