The global technology sector is currently navigating a $3 to $4 trillion artificial intelligence infrastructure buildout, forcing hyperscalers and sovereign entities to confront severe capital expenditure constraints. Faced with the extreme costs of compute power, Amazon is accelerating the development of proprietary hardware to meet international processing demands. CEO Andy Jassy confirmed that Amazon’s custom silicon division, encompassing Trainium, Graviton, and Nitro processors, has now surpassed a $20 billion annual revenue run rate.
To capitalize on the shortage of affordable computing resources, Amazon is actively exploring the direct sale of Trainium chips to independent and sovereign data centers as of mid-June 2026. This strategy moves the company beyond its traditional AWS cloud subscription model and places it in direct hardware competition with Nvidia. Financial monitors now project that Amazon’s stock trajectory could mirror Nvidia’s historic expansion if its custom division maintains this momentum.
Amazon’s hardware push has already secured commitments from major developers like Apple and Anthropic. Jassy noted that Anthropic engineers directly collaborated on chip architecture, actively designing solutions to bypass the strictures of closed hardware ecosystems.
The Trainium Rollout Sequence
Amazon is scaling its custom silicon operations through an aggressive release schedule designed to undercut existing GPU pricing:
- Trainium2 Deployment: The second-generation chips deliver an estimated 30% price-performance improvement over standard GPUs for training large language models.
- Trainium3 Shipments: Amazon initiated shipments of its third-generation architecture in early 2026, with heavy subscription rates from primary AI developers.
- External Sales Exploration: In June 2026, the company began evaluating direct hardware sales to external entities requiring independent infrastructure.
How Amazon’s Hardware Push Threatens Nvidia’s Market Share
The current hardware cycle closely tracks Amazon’s disruption of Intel’s CPU monopoly in 2018. By aggressively migrating heavy cloud workloads to its proprietary Graviton processors, Amazon successfully broke its dependence on third-party data center silicon. The company is now executing that exact playbook in the AI accelerator market.
Nvidia’s dominance in the AI hardware sector is actively declining. After holding an 87% market share in 2024, Nvidia is projected to control roughly 75% of the market by the end of 2026. As Amazon and Google force their own proprietary alternatives into the broader supply chain, the pricing power and market concentration of the traditional GPU manufacturers face immediate contraction.