Microsoft apparently collaborating with AMD on AI processors to challenge Nvidia

In April, a report stated that Microsoft is working on artificial intelligence (AI) chips to use them to train large language models (LLMs). The report, however, did not disclose any partners with whom the business was working. According to a recent rumor, Microsoft is cooperating with Advanced Micro Devices (AMD) for the development of AI CPUs.

Citing individuals with knowledge of the issue, a report by Bloomberg indicated that Microsoft would assist improve AMD’s supply of in-demand chips. It also emphasized that this alliance is part of a multi-pronged attempt to get more of the highly wanted components.

An alternative to Nvidia

According to the article, the persons also claimed that the firms would provide an alternative to Nvidia, which leads the market for AI-capable processors (GPUs). It has to be mentioned that Microsoft constructed a supercomputer with OpenAI and its design contains tens of thousands of Nvidia A100 graphics chips connected together to give power to train AI models.

“We built a system architecture that could operate and be reliable at a very large scale. That's what led in ChatGPT becoming feasible. That’s one model that came out of it. There’s going to be many, many others,” Nidhi Chappell, Microsoft head of product for Azure high-performance computing and AI, said at the time.

How Microsoft can aid AMD

The Bloomberg article added that Microsoft would give help, including technical personnel and collaborating with the chipmaker on a native Microsoft processor for AI workloads, code-named Athena, to bolster AMD’s efforts.

Google’s AI chip

Google said last year that it built an AI device dubbed the Tensor Processing Unit (TPU). The CPU has been particularly created for machine learning activities and is said to do billions of operations per second and, at the same time, utilize minimal watts of electricity.

The TPU is to be used with Google’s TensorFlow software, Google's open-source software framework for machine learning.