Amazon has announced a major shift in its AI hardware strategy by moving to directly sell its in-house Trainium chips to third-party customers. The decision marks a deeper push into the AI infrastructure market and signals Amazon’s ambition to challenge NVIDIA’s dominance in AI accelerators.
CEO Andy Jassy said the move could increase the company’s chip division revenue from $20 billion annually to $50 billion. That target underscores Amazon’s confidence in the commercial potential of its custom silicon as demand for AI computing continues to accelerate across cloud platforms and enterprise applications.
Targeting Cost and Supply Constraints
According to the disclosed details, Trainium2 delivers a 30% better cost-performance ratio than competing products. Amazon also said the upcoming Trainium3 is already nearly fully subscribed ahead of shipment, suggesting strong demand for alternative AI compute options. With global AI compute capacity still constrained, chips that offer better economics are increasingly attractive to customers.
By expanding external sales of Trainium, Amazon is also working to reduce its reliance on NVIDIA GPUs. In recent years, the rapid rise of generative AI has intensified pressure on GPU supply and driven up procurement costs across the industry. Amazon’s decision to open up its in-house chips to outside buyers could help broaden customer access to computing power while strengthening its own vertically integrated stack.
Building a Full-Stack AI Ecosystem
The initiative is part of Amazon’s broader effort to build a full-stack AI ecosystem spanning chips, training, inference, and cloud deployment. If Trainium gains broader adoption beyond Amazon’s internal platforms, the company could strengthen its position in AI hardware and reshape competitive dynamics in the sector.
More broadly, Amazon’s move reflects a growing trend among major technology companies to develop proprietary hardware and compete more aggressively in the core AI compute layer. As the global shortage of AI computing resources remains a key industry challenge, new commercial alternatives could intensify competition across the AI chip market.