Amazon Explores Selling Trainium AI Chips to Third Parties
AWS is in early talks to market its custom silicon directly to data center operators, potentially creating a $50 billion competitor to Nvidia.

Amazon Web Services is exploring a significant strategic shift that could position it as a direct competitor to Nvidia in the AI chip market. AWS AI chief Peter DeSantis confirmed to Bloomberg that the company is in early-stage discussions to sell its Trainium AI processors to other companies for use in their data centers, though he declined to name potential buyers.
The move follows comments from Amazon CEO Andy Jassy in his April shareholder letter, where he revealed that if AWS operated its chip business as a standalone entity and sold to third parties, it would generate approximately $50 billion in annual revenue based on current production volumes. That figure would make it comparable in scale to Intel's annual revenues, though still well below Nvidia's current $326 billion revenue run rate.
The capacity constraint challenge
AWS faces a fundamental tension in pursuing this strategy. The company has consistently reported that Trainium chip capacity sells out almost immediately to its cloud customers. Jassy noted in April that even Trainium4—which won't launch for more than a year—has already sold out its initial capacity allocation.
Selling chips externally would require AWS to either expand production significantly or divert supply from existing cloud customers who are already on waiting lists. Manufacturing expansion presents its own obstacles, as AWS relies on partners like TSMC, where Nvidia has recently become the largest customer, surpassing even Apple.
Why it matters
This potential move represents a fundamental shift in AWS's business model. Currently, AWS monetizes its custom chips through a layered approach: charging for AI compute tokens while also capturing revenue from adjacent services like storage, security, networking, and monitoring that customers need for complete AI applications. Selling chips directly would sacrifice that multiplier effect for immediate hardware revenue. The decision signals how competitive pressure and market opportunity may be forcing cloud providers to reconsider vertical integration strategies that have served them well. If AWS proceeds, it would mark one of the most significant challenges yet to Nvidia's position in the AI infrastructure market.
The broader chip ambitions
AWS spokesperson Doron Aronson confirmed the company is considering direct chip sales, noting that while AWS has historically declined such requests, future sales to third parties are "quite possible." The timing is notable as Nvidia CEO Jensen Huang recently announced the company's expansion into a new $200 billion CPU market for AI workloads, moving into territory traditionally held by Intel and AMD.
The discussions remain in early stages, according to TechCrunch, and any actual product availability would depend on resolving the manufacturing and capacity allocation challenges. AWS added OpenAI to its model offerings after Jassy's April letter, further intensifying internal demand for its chip capacity.
These details were first reported by Bloomberg and confirmed to TechCrunch by AWS.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
