In another wild season of AI chips, Meta tokens sell millions of Amazon AI CPUs

Amazon just made a big splash with Meta thanks, again, to Amazon’s home chips. Meta has signed an agreement to use millions of a The gravity of AWS chips to support the growth of AI, Amazon announced on Friday.

Note that AWS Graviton is an ARM-based CPU, (central processing unit, a chip that handles computing tasks) and not a GPU (graphics unit).

While GPUs remain the tool for training large models, once those models are trained, the AI ​​assistants built on top of them are making changes to the chip model. Agents perform a number of tasks such as real-time reasoning, coding, analysis, and coordination involved in managing agents through multiple activities. AWS’ latest version of Graviton is designed to address AI-related needs, the company says.

The deal brings more Meta revenue to AWS than competitors like Google Cloud. last august, Meta signed a six-year, $10 billion deal with Google Cloudalthough Meta was, until then, primarily an AWS client that also used Microsoft Azure.

We couldn’t help but notice that AWS had timed the announcement of the partnership when the Google Cloud Next conference was concluded, as a joke to its cloud partner. Google, of course, also makes their own AI chips and announced their new products at the show.

Indeed, Amazon also produces its own AI GPU: Trainium, which, despite its name, is used for training and control – a stage that takes place after the training of the model, when it is ready quickly.

But Anthropic had already stepped in with the previously announced deal this month which commandee many chips for years to come. Developer Claude agreed to spend $100 billion over 10 years to run most of his services on AWS – focusing on Trainium – while Amazon agreed to invest $5 billion (bringing its total investment to $13 billion) to Anthropic in return.

Ultimately, the Meta deal allows Amazon to showcase its main AI client as a proving ground for its in-house CPUs. These are chips that compete with Nvidia’s new Vera CPU, which is also ARM-based and designed for AI tasks. The difference, of course, is that Nvidia sells its chips and AI systems to businesses and cloud providers (including AWS). AWS only sells access to the chips through its cloud service.

Earlier this month Amazon CEO Andy Jassy he took a look at Nvidia and Intel in his annual letter, saying that businesses want to control prices for AI, and that they want to win for this reason. This also means that the pressure cannot be greater on Amazon’s internal chip-building team to deliver, the team itself we visited last month for a special tour of their lab.