Amazon AWS's customized strategy has achieved results, targeting AI chip giant NVIDIA

Zhitong
2025.06.18 01:17
portai
I'm PortAI, I can summarize articles.

Amazon AWS will update the Graviton4 chip, achieving a network bandwidth of 600 Gbps, making it the highest bandwidth in the public cloud. This update is a success of Amazon's customization strategy aimed at competing with traditional semiconductor manufacturers like NVIDIA, especially in the field of artificial intelligence infrastructure. AWS has built an AI supercomputer, Project Rainier, for the startup Anthropic and invested $8 billion to support its development. AWS stated that the Trainium3 chip will be launched this year, with performance doubling that of Trainium2 and saving 50% energy

According to Zhitong Finance APP, Amazon's (AMZN.US) cloud computing service platform AWS will announce an update to its Graviton4 chip, which has a network bandwidth of 600 gigabits per second, the highest bandwidth in the public cloud, according to the company.

AWS engineer Ali Saidi compared this speed to a machine reading 100 music CDs per second.

Graviton4 is a central processing unit (CPU) and is one of many chip products launched by Amazon's Annapurna Labs located in Austin, Texas. This chip is a significant victory for the company's custom strategy, enabling it to compete with traditional semiconductor manufacturers like Intel (INTC.US) and AMD (AMD.US).

However, the real battle is in challenging NVIDIA (NVDA.US) in the artificial intelligence infrastructure space.

At the AWS re:Invent 2024 conference held last December, the company announced the AI supercomputer Project Rainier built for the startup Anthropic. AWS has invested $8 billion to support Anthropic.

Gadi Hutt, Senior Director of AWS Customer and Product Engineering, stated that Amazon is seeking to reduce the cost of AI training and provide an alternative to NVIDIA's expensive graphics processing units (GPUs).

According to AWS, Anthropic's Claude Opus 4 AI model was launched on the Trainium2 GPU, while Project Rainier is powered by over 500,000 chips—an order that would traditionally go to NVIDIA.

Hutt noted that while NVIDIA's Blackwell chip outperforms Trainium2, AWS chips offer better cost performance.

He stated, "Trainium3 will be launched this year, with performance doubling that of Trainium2 and saving 50% of energy."

Rami Sinno, Engineering Director at AWS's Annapurna Labs, mentioned that the demand for these chips has already exceeded supply.

From upgrading Graviton4 to launching the Rainier project's Trainium chips, Amazon is showcasing its broader ambition to control the entire AI infrastructure stack from networking to training to inference.

As more mainstream AI models (like Claude 4) prove they can be successfully trained on non-NVIDIA hardware, the question is no longer whether AWS can compete with this chip giant, but how much market share it can capture.

An AWS spokesperson stated that the timeline for the Graviton4 update will be announced by the end of June

Source: Zhitong The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.