Amazon Is in Talks to Sell Its Trainium AI Chips to Outside Data Centers, Expanding Beyond AWS to Challenge Nvidia

Key Takeaways

• Trainium third‑gen sold out; fourth‑gen attracting early interest
• Amazon aims to double chip revenue to ~$50 B via external sales
• European sovereign‑cloud demand drives push for locally owned AI chips
• Amazon’s move mirrors Google’s TPU outreach to non‑cloud users
• External sales unlikely to erode AWS, given current AI under‑consumption

Pulse Analysis

Amazon’s decision to market its Trainium AI processors beyond its own cloud reflects a broader industry trend of chip makers seeking diversified revenue streams. Historically, Amazon kept Trainium confined to AWS, using it to differentiate its cloud offering from rivals like Microsoft and Google. By opening the hardware to third‑party data centers, Amazon can monetize excess capacity and tap into enterprises that require on‑premise AI compute for latency, data‑privacy, or regulatory reasons. This strategy also signals confidence in Trainium’s performance, which has already outpaced demand for its third generation, prompting a rapid rollout of a fourth‑gen version.

The European market is a key catalyst for the outreach. EU regulators are pushing for sovereign cloud infrastructure, encouraging local control over AI workloads to mitigate geopolitical risk. Trainium’s availability to non‑AWS operators aligns with these policy goals, offering a home‑grown alternative to Nvidia’s GPUs and Intel’s Habana chips. Google’s recent decision to sell its TPU accelerators to external data centers set a precedent, suggesting that major cloud providers see hardware as a strategic asset that can be leveraged across ecosystems rather than a proprietary lock‑in. For European firms, the prospect of sourcing Amazon‑designed silicon could reduce reliance on US‑based GPU vendors while maintaining compatibility with existing cloud‑native tools.

Financially, the move could be transformative for Amazon’s chip business. Andy Jassy’s projection of roughly $50 billion in annualized chip revenue—potentially doubled by external sales—places Amazon among the top AI hardware vendors, directly challenging Nvidia’s market dominance. The additional revenue would diversify Amazon’s earnings beyond its core retail and cloud services, enhancing resilience amid fluctuating cloud spending. Moreover, broader adoption of Trainium could accelerate software ecosystem development, driving more AI startups to build on Amazon’s stack and further entrenching its position in the AI supply chain. In a market where compute demand is outpacing supply, Amazon’s expanded hardware footprint may reshape competitive dynamics and spur faster innovation across the AI hardware landscape.