According to Jassy, Amazon's chip division could have a valuation of $50 billion, and he suggests that there may be plans to sell them to external customers.

      In summary, Andy Jassy’s annual letter to shareholders, released on 9 April 2026, indicates that Amazon's custom chip division, which includes Graviton, Trainium, and Nitro, generates over $20 billion in annual revenue growing at triple-digit rates each year. Jassy suggests that if this business were to be sold on the open market like Nvidia, it could be valued at around $50 billion annually. He also hints that Amazon might start selling these chips directly to third parties and defends the company’s $200 billion capital expenditure plan for 2026 as based on confirmed customer demand rather than speculation.

      "Not on a hunch": the $200 billion wager

      Jassy began the financial section of the letter by directly addressing the skepticism surrounding Amazon's capital commitments. He stated, “We’re not committing approximately $200 billion in capex for 2026 based on a hunch. We won’t be conservative in our approach. We’re investing to be a significant leader, leading to much larger future business, operating income, and free cash flow.” This declaration comes in the context of a company whose free cash flow dropped from $38 billion to $11 billion last year, primarily due to a $50.7 billion rise in capital expenses, most directed toward AI infrastructure.

      Jassy emphasized that substantial portions of the anticipated CapEx for 2026 already have customer support backing, citing OpenAI's commitment of over $100 billion to AWS as an example. This commitment extends an earlier $38 billion seven-year partnership initiated in November 2025, which also includes OpenAI utilizing around two gigawatts of Trainium capacity through AWS infrastructure. SoftBank, which has a majority stake in OpenAI and is funding its infrastructure development through methods like a $40 billion bridge loan, effectively supports the demand that Jassy highlights as validation for his CapEx position.

      A $50 billion chip business in plain sight

      Amazon’s custom silicon program features three product lines. Graviton is a customized CPU that Jassy claims offers over 40% better price-performance compared to similar x86 processors, which are predominantly controlled by Intel and AMD. Currently, 98% of the top 1,000 EC2 customers utilize Graviton, indicating a long-term shift in the economics of cloud computing. Demand has been so strong that two significant AWS clients inquired about acquiring all available Graviton capacity for 2026, but Amazon declined.

      Trainium is the AI training and inference accelerator representing Amazon's direct response to Nvidia. Trainium2, offering around 30% better price-performance than comparable GPUs, is nearly sold out. Trainium3, which started shipping in early 2026 and provides an additional 30 to 40% improvement in price-performance over Trainium2, is also close to being fully reserved, with companies like Uber migrating workloads onto it. Trainium4, which is still around 18 months from widespread availability and features compatibility with Nvidia’s NVLink Fusion interconnect technology, has already seen significant reservations. Nitro, the custom network and security chip that forms the foundation of AWS's virtualization layer, completes the triad of offerings. Collectively, Jassy states these three lines generate over $20 billion in annual revenue, increasing at triple-digit percentages annually. “If we were a standalone chip company,” he notes, “our chips would be producing more than $50 billion in annual revenue.” Currently, this business operates solely within AWS; customers access Trainium and Graviton via EC2 instances rather than purchasing the chips directly.

      Jassy argues that at scale, Trainium will “save us tens of billions in capex annually, and provide several hundred basis points of operating margin advantage compared to depending on hardware from other suppliers for inference.” This assertion is crucial to the investment thesis supporting the $200 billion CapEx program: custom silicon serves not only as a competitive differentiator but also as a structural cost advantage that grows as the ratio of inference to training in AI workloads increases over time.

      The Nvidia relationship and the "new shift"

      Jassy is measured in his portrayal of Amazon's competitive stance against Nvidia. He writes, “We have a solid partnership with NVIDIA; we will always have customers who prefer to utilize NVIDIA products," while asserting, “virtually all AI to date has been executed on NVIDIA chips, but a new shift is underway.” He points out that customers are seeking improved price-performance. Nvidia has reported a revenue of $68.1 billion in the fourth quarter of 2025, showing a 73% year-on-year growth, and entered 2026 maintaining a dominant market position that Amazon's custom silicon is beginning to erode from within its AWS customer base rather than in the broader merchant market. The incorporation of NVLink Fusion in Trainium4 means Amazon is building connections rather than barriers: it enables customers to use Trainium accelerators alongside Nvidia GPUs in a single system, thus maintaining flexibility for enterprises