Microsoft plans to release its first AI chip "Athena" to reduce reliance on NVIDIA GPU.

According to a new report, Microsoft plans to launch its own AI chip for the first time next month, which is a strategic move. The chip, codenamed "Athena," may reduce Microsoft's reliance on NVIDIA-designed GPUs due to the shortage of GPU supply caused by increased demand. The chip's release is likely to take place at the Microsoft Ignite conference starting on November 14th. Athena is expected to compete with NVIDIA's flagship model, the H100 GPU, in AI acceleration in data centers. This custom chip has already undergone secret testing within small groups at Microsoft and its partner, OpenAI. Microsoft started developing Athena in 2019 with the aim of reducing costs and establishing a closer relationship with NVIDIA. Currently, Azure relies on NVIDIA GPUs for AI capabilities, which are used by Microsoft, OpenAI, and cloud customers. However, with Athena, Microsoft can follow in the footsteps of competitors Amazon Web Services and Google, leading the way in providing in-house developed AI chips for cloud users. Performance details are not yet clear, but Microsoft hopes Athena can match the highly demanded H100. Despite many companies promoting superior hardware and cost advantages, NVIDIA GPUs remain the preferred choice for AI developers due to NVIDIA's CUDA platform. Attracting users to adopt new hardware and software will be crucial for Microsoft. This plan may also reduce dependence on NVIDIA due to the tight GPU supply. After closely collaborating with OpenAI, Microsoft reportedly ordered hundreds of thousands of NVIDIA chips to support OpenAI's products and research needs. Significant cost savings could be achieved by using their own chips. OpenAI may also be considering reducing its dependence on Microsoft and NVIDIA chips. Reuters reported earlier today that the AI research lab is considering manufacturing its own AI chips. Recent job postings on the OpenAI website also indicate the company's efforts to recruit talent for evaluating and co-developing AI hardware. While Microsoft and other cloud service providers have not immediately halted their plans to purchase GPUs from NVIDIA, convincing their cloud customers to choose their own chips over NVIDIA's GPU servers may have long-term economic advantages. Microsoft is also closely collaborating with AMD to develop the upcoming AI chip, MI300X. With the surge in AI workloads, diversification provides multiple options. Cloud competitors are adopting similar strategies to avoid vendor dependence. Amazon and Google have strategically integrated their own AI chips into activities promoting their cloud businesses. Amazon's financial support for competitor Anthropic comes with the condition that Anthropic must use Amazon's AI chips, Trainium and Inferentia. Meanwhile, Google Cloud promotes cases of customers such as AI image developer Midjourney and Character AI using the company's tensor processing units. As AI chips become crucial components in data centers, the situation becomes increasingly challenging. With this development, Microsoft will join the ranks of its competitors, vying for a share of the rapidly growing AI silicon market. With Athena, Microsoft can provide more choices for cloud customers while charting a more independent direction in the next generation of AI infrastructure.