Amazon considers selling its self-developed AI chips to external customers, with annual revenue potentially reaching $50 billion.

Amazon considers selling its self-developed AI chips to external customers, with annual revenue potentially reaching $50 billion.

```

Amazon is considering expanding its in-house chip business to a broader market.

On Thursday, April 9, as mentioned by WallstreetCN, Amazon CEO Andy Jassy released the annual shareholder letter, disclosing that the company is considering selling its self-developed chips to third parties.

In the annual letter to shareholders, Jassy stated that Amazon's internal chip division currently generates more than $20 billion in annualized revenue; if it operates as an independent business and sells to both AWS clients and external third parties, annualized revenue could reach $50 billion.

Jassy wrote in the letter:

The demand for our chips in the market is extremely strong, and in the future, we are very likely to sell full rack chips to third parties.

Amazon has been cultivating the chip sector for many years, but its business model has been entirely different from that of Nvidia or Intel. Previously, Amazon did not directly “sell” physical chips to third parties, but rather “rented” the computing power of these chips to enterprises through its cloud service AWS.

Scale of In-house Chip Business Revealed for the First Time

Amazon’s chip division mainly produces three types of products: general-purpose compute chips, AI accelerators, and specialized chips to improve server operating efficiency.

Currently, these hardware resources are provided to AWS cloud computing customers in rental form, not sold directly to external parties.

Jassy likened the evolution of the current AI chip landscape to the structural migration previously seen in the CPU field. Amazon launched its in-house CPU chip, Graviton, in 2018, which delivered up to 40% better price-performance ratio compared to x86 processors, and is now widely adopted by 98% of the top 1,000 EC2 customers.

Amazon EC2 (Elastic Compute Cloud) is one of the core cloud computing services provided by AWS, allowing users to rent virtual servers on demand to run applications in the cloud.

In the AI chip field, Amazon is replicating this approach. The second-generation AI training chip Trainium2 offers about 30% better price-performance than similar GPUs and is now essentially sold out.

The Trainium3, which will start shipping in early 2026, will offer a further 30-40% performance improvement over Trainium2, and the subscription rate is already close to full. For the Trainium4, which is about 18 months away from official launch, a large portion of its capacity has already been pre-ordered.

The shareholder letter specifically pointed out that Amazon’s main inference service, Amazon Bedrock, now runs most of its inference workloads on Trainium chips, with demand remaining strong. Jassy also revealed that in the future, selling racks of Trainium chips to third-party customers is not out of the question.

AI Computing Power Shortage Drives Overflow in Demand

The immense computing power required for AI model training has kept market supplies persistently tight, pushing companies to actively look for alternatives to Nvidia GPUs.

It is against this backdrop that the external sales potential of Amazon’s self-developed chips has drawn growing attention.

Jassy disclosed in the letter that two large AWS clients have requested to purchase all Graviton capacity for 2026, but Amazon could not agree as it needs to balance the needs of other customers. This detail highlights the current severe shortage in computing supply.

Jassy admitted that AWS currently faces capacity bottlenecks leading to unmet demand, and expects that by the end of 2027, total power capacity will have doubled from its current level.

In 2025, AWS will add 3.9 GW of new power capacity, with Q4 2025 revenue up 24% year-over-year and annualized revenue reaching $142 billion. He characterizes the pressure on short-term free cash flow as a necessary cost, believes AI is a “once-in-a-lifetime opportunity,” and says he “will not take a conservative approach.”

Andy Jassy has been at the helm of Amazon for nearly five years. In this shareholder letter, he reiterated the growth potential of the chip business, reflecting Amazon’s deeper strategy in cloud computing and AI infrastructure.

With the potential for external sales now raised, this previously behind-the-scenes business may become Amazon’s next growth story for the capital markets.

Risk Warning and DisclaimerThe market involves risks, and investments should be made cautiously. This article does not constitute personal investment advice, nor does it take into account the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investments made accordingly are at one’s own risk. ```