Amazon may sell its AI chips to outside companies, and the business is already heading toward $20 billion

Amazon is considering selling its AI chips to other companies, not just using them inside its own cloud infrastructure. CEO Andy Jassy said the company's in-house semiconductor business is already heading toward more than $20 billion in annualized revenue.

A new direction for Amazon

Amazon has long bet on its own technology inside AWS to better control the cost and performance of its cloud services. Now the company appears to be weighing the next step: turning its internal work into a separate commercial product for the external market.

This matters because it is not a startup's pilot experiment but a possible decision by one of the largest players in the cloud industry, which already has scale, infrastructure and demand from enterprise customers.

Jassy's statement also shows that Amazon's in-house silicon business has stopped being a support function. If the unit really is heading toward more than $20 billion in annualized revenue, the company already treats chips as a notable source of income, not just a way to reduce dependence on third-party suppliers. For the market, it is a signal: Amazon is confident its products can appeal beyond the AWS ecosystem.

What is changing in the market

The AI infrastructure market is quickly shifting from an abstract race of models to a fight over physical resources: compute, memory, power consumption and accelerator availability. Against this backdrop, in-house chips are becoming not just a technical optimization but a strategic asset. If Amazon opens them to other companies, it could compete not only as a cloud provider but also as a supplier of critical hardware for training and running modern models. For Amazon itself, it is also a way to capitalize on years of investment in its own hardware platform.

While most companies discuss AI in terms of models and applications, the largest providers are increasingly fighting for the bottom layer of the stack: the compute base. Whoever controls the chip controls the economics of the service: inference cost, scaling flexibility, supply predictability and the ability to launch new offerings faster for enterprise customers worldwide.

• For customers, potentially more choice in AI infrastructure.
• For Amazon, a new source of revenue beyond classic cloud services.
• For competitors, pressure on prices and a need to develop their own chips faster.
• For the market, another sign that control over chips is becoming a central part of AI strategy.

Why the number matters

The estimate of more than $20 billion matters because it changes the perception of scale. Internal hardware initiatives at large tech companies usually look like investments in the future for a long time: expensive, complex and not always clear from the outside. But this level of annualized revenue means the business is already close to a large standalone company.

For investors and enterprise buyers, this reduces the sense of an experiment and increases confidence in the long-term nature of the strategy.

At the same time, Jassy's words imply something else: Amazon is so far only considering selling chips to other companies, not announcing a full launch. That is an important distinction. The decision may still depend on demand, manufacturing constraints, supply economics and how willing the company is to separate its hardware advantages from the AWS cloud offering. But the very fact of public discussion shows that Amazon already sees such a scenario as realistic.

What this means

Amazon is gradually turning its AI chips from an internal tool into a potential market product. If the company really opens sales to outside buyers, the fight over AI infrastructure will get tougher: not only models and clouds will compete, but the compute platforms themselves. This is no longer a question of prestige but of control over margins.