What's Happening?
Amazon plans to start selling its Trainium AI chips directly to other companies by late 2027, a departure from its current model of only offering their computing capacity through Amazon Web Services (AWS). This strategic shift was announced by CEO Andy
Jassy during the company's October 29, 2026, earnings call. Previously, Trainium chips were exclusively used within AWS, where customers rented their processing power. The new approach will allow external entities to purchase the physical chips outright. This move introduces a second revenue stream for Amazon's AI chip operations, alongside the existing AWS cloud service. The company anticipates this external sales business to become significant, though specific details regarding buyers, pricing, sales volume, or revenue targets have not yet been disclosed. Amazon will annually reassess the allocation between external sales and AWS deployment, acknowledging that manufacturing supply is a key constraint.
Why It's Important?
This decision by Amazon marks a significant evolution in its AI chip strategy, potentially intensifying competition in the rapidly growing AI hardware market. By selling Trainium chips directly, Amazon aims to capture a share of the market currently dominated by companies like NVIDIA. This could provide Amazon with upfront revenue from chip sales, which is crucial given its substantial capital expenditures, including an estimated $220 billion in 2026 and a projected increase in 2027. For customers, this offers an alternative to renting computing power, providing more direct control over their AI infrastructure. However, the success of this venture will depend on Amazon's ability to secure manufacturing allocations and attract buyers, as each chip sold externally is one less available for its own AWS cloud services. The move also highlights Amazon's commitment to expanding its presence in the AI ecosystem beyond cloud services.
What's Next?
Amazon is set to begin external sales of its Trainium AI chips in late 2027. The company has indicated that it will review the balance between external sales and AWS usage annually, suggesting that the availability of chips for outside buyers could fluctuate based on internal demand and manufacturing capacity. Key information such as initial buyers, pricing structures, sales volumes, and specific chip generations remains undisclosed. The market will be watching for these details, as they will provide a clearer picture of Amazon's competitive positioning and revenue potential in the AI chip sector. The impact on competitors, particularly NVIDIA, will also be a significant point of interest, especially as Amazon has open-sourced TorchNeuron to allow developers to run existing PyTorch code on Trainium, potentially challenging NVIDIA's CUDA ecosystem.
Beyond the Headlines
Amazon's decision to sell its AI chips externally could have broader implications for the cloud computing and AI industries. It signifies a potential shift in how major tech companies monetize their proprietary hardware, moving beyond a purely service-based model. This could lead to increased vertical integration within the tech sector, where companies not only offer cloud services but also the underlying hardware. The move also underscores the strategic importance of AI chips as a critical component of the digital economy, prompting companies to explore diverse monetization strategies. Furthermore, by offering its chips directly, Amazon could foster a more diverse and competitive AI hardware landscape, potentially driving innovation and reducing reliance on a few dominant suppliers. The long-term success of this strategy will depend on Amazon's ability to balance its internal AWS needs with external market demand and navigate the complexities of chip manufacturing and supply chain management.
