• AWS’s Graviton CPU chips are used by more than 90% of its top 1,000 data center customers.
• Graviton chips, launched in 2018, are becoming an important part of AWS’ overall business.
• AWS’ custom silicon strategy includes AI chips like Inferentia and Trainium.

In 2018, Amazon Web Services has launched Graviton, its own line of in-house central processing unit chips for data center servers. Just six years later, the vast majority of AWS’s largest server customers have become Graviton users.

Rahul Kulkarni, director of compute and AI/ML at AWS, told Business Insider that more than 90% of the top 1,000 elastic compute cloud (EC2) customers use Graviton chips. Customers can only access Graviton chips through AWS EC2 servers.

“This continues to show how Graviton is gaining traction,” Kulkarni said.

The new data point is the latest indication of Graviton’s growing success. AWS previously said that more than 50,000 customers, including the top 100 EC2 users, operate Graviton-based servers.

Graviton is a key part of AWS’s data center business. Designing its own chips allows AWS to reduce data center operating costs because it no longer needs to purchase processors from other vendors, such as Intel or AMD. Graviton also uses Arm-based designs, which are more expensive and more energy efficient than conventional x86-powered chips.

Amazon does not disclose the amount of revenue from its custom silicon business. The company is expected to announce its third-quarter results on Thursday.

Kulkarni said Graviton’s value for money, energy efficiency and overall performance are leading to greater customer adoption. Large companies, such as Epic Games, Databricks and Pinterest, are all important Graviton customers, he said.

“We are very committed to custom silicon, and this is an area that we will continue to invest in at a very aggressive pace going forward,” Kulkarni said.

AI Inference

AWS began considering designing its own custom chips after its senior vice president and distinguished engineer, James Hamilton, wrote a 6-page internal strategy document in 2013. In 2015, AWS doubled down on its efforts. this plan by acquiring Annapurna Labs, a chip designer based in Israel.

Bernstein Research wrote last year that Amazon is the “most successful” Arm server chip designer, supplying more than 50% of the world’s Arm server chips. He estimated that Graviton accounted for about 20% of AWS’ CPU usage by mid-2022. AWS offers access to other processors, including Intel or AMD chips.

AWS’ Kulkarni said Graviton is still primarily used for general computing purposes, but a growing number of customers are using it for AI inference and machine learning frameworks. He said Graviton, now in its fourth generation, has added new features and capabilities that enable unique use cases, like AI inference, that don’t necessarily require dedicated machine learning processors.

“This is a new revenue channel that goes beyond what AWS intended to use Graviton for,” Kulkarni said.

Over the past two years, Amazon has been reduce costs across the company, closing dozens of projects or scaling back unprofitable units. AWS’ chip design business appears immune to these cuts.

On a call with analysts in August, Jassy said Graviton had been “very successful” for AWS. He added that Custom AI chips from AWSInferentia and Trainium are expected to follow a similar growth trajectory, although they see mixed results due to Nvidia’s early lead, as BI previously reported.

“This is one of the most strategic areas for us,” Kulkarni said. “We will absolutely continue to drive innovation in custom silicon, as we have for over 10 years.”

Do you work at Amazon? Do you have any advice?

Contact journalist Eugene Kim via encrypted messaging apps Signal or Telegram (+1-650-942-3061) or by email ([email protected]). Contact us using a non-professional device. Discover this one from Business Insider source guide for more tips on sharing information securely.