Amazon’s cloud unit announced the launch of the new artificial intelligence chip “Trainium 2” and the general-purpose processor “Graviton 4” at the Reinvent conference in Las Vegas. The company explained that it provides access to Nvidia’s “H200” graphics processing units.
Amazon aims to excel as a cloud provider with a wide range of cost-effective options. Amazon’s cloud offers excellent products, which means the highly demanded graphics processing units from Nvidia, the leading company in manufacturing artificial intelligence chips.
This dual approach may put Amazon in a better position to compete with Microsoft. Earlier this month, Microsoft followed a similar approach by unveiling the Maia 100 artificial intelligence chip and hinting at the presence of Nvidia’s H200 graphics processing units in its cloud.
The Graviton 4 processors are based on Arm architecture design and consume less energy compared to Intel or AMD chips. The Graviton 4 offers up to 30% better performance compared to the current Graviton 3 chip.
More than 50,000 of Amazon’s customers use Graviton chips, and the company mentioned that Antherbiq and Databricks plan to build models using the new Trainium 2 chip, which boasts four times better performance than the original model.
Amazon has become the leading company in cloud services offering Nvidia’s GH200 Grace Hopper Superchips with NVLink technology across the cloud.
Amazon and Nvidia collaborated on the Siba project to design a giant fast artificial intelligence computer powered by graphics processing units, hosted by Amazon in the cloud for Nvidia’s research and development team.
This computer features 16,384 GH200 Grace Hopper Superchips from Nvidia and is capable of processing 65 exaflops of artificial intelligence. Nvidia uses this computer to enhance the new wave of its generative artificial intelligence innovations.
There has been a significant increase in demand for Nvidia’s graphics processing units since the launch of OpenAI’s ChatGPT artificial intelligence chatbot last year. As a result, Nvidia chips ran out, and companies raced to integrate similar generative AI technologies into their products.
Introducing an artificial intelligence chip from a cloud provider may pose a challenge to Nvidia, however, Amazon is simultaneously expanding its partnership with Nvidia.
