Amazon says its AI services on AWS exceed a $15 billion annual run rate

What happened

Amazon CEO Andy Jassy said AWS's AI services had reached an annual revenue run rate of more than $15 billion, and also highlighted the growth of the company's own chips, such as Trainium and Graviton.

The figure confirms that AI is now a significant business line within AWS, and that developing in-house silicon is part of its strategy to reduce dependence on external chip suppliers.

What it means for a mid-sized business

The growth of in-house chips such as Trainium and Graviton could translate, in the medium term, into pricing changes and new deployment options for those using AI services on AWS. Such changes are not always transparent or automatic.

For a business already using or evaluating cloud AI services, the practical lesson is to avoid total lock-in to a single hardware or vendor option, and to design applications with standards that allow adapting to new alternatives without rebuilding the whole system.

What to review

Before committing to a long-term AI architecture, it is worth checking:

• Whether the application depends on a specific chip or instance type and how much effort migrating it would take.
• Whether the design uses abstraction layers that make it easier to switch compute provider.
• What cost impact a pricing change from the contracted AI provider would have.
• Whether a documented exit plan exists for critical AI services.

Frequently Asked Questions

What are Trainium and Graviton?

They are Amazon's own chips: Trainium is designed for AI workloads and Graviton is a general-purpose processor for AWS instances.

Why does Amazon develop its own chips?

To reduce its dependence on external semiconductor suppliers and better control the cost and availability of the compute capacity it offers.

Should this concern a business using AWS?

Not directly, but applications should be designed with standards that allow taking advantage of pricing or hardware changes without depending on a single option.