Microsoft won’t stop buying AI chips from Nvidia, AMD, even after launching its own, Nadella says

This week, Microsoft announced the deployment of its first batch of proprietary AI chips, known as the Maia 200, at one of its data centers. The company has plans to expand this rollout in the months ahead. Designed as an 'AI inference powerhouse,' the Maia 200 is specifically optimized for the demanding tasks of running AI models in live environments. Microsoft has touted the Maia's processing capabilities, claiming it surpasses Amazon's latest Trainium chips and Google's Tensor Processing Units (TPU). The push towards developing their custom AI chips comes as cloud service giants face challenges in securing the latest offerings from Nvidia, amid ongoing supply constraints. Despite the introduction of its own high-performance chips, CEO Satya Nadella emphasized that Microsoft will continue to purchase chips from established partners like Nvidia and AMD. He stated, “We have a great partnership with Nvidia, with AMD. They are innovating. We are innovating.” Nadella noted that while vertical integration is a strategy for the company, it does not preclude collaboration with external vendors. The Maia 200 chip will be utilized by Microsoft's Superintelligence team, which is focused on developing the company's advanced AI models. Mustafa Suleyman, the former co-founder of Google DeepMind and current leader of the team, revealed that the Maia 200 will also facilitate the operation of OpenAI's models on Microsoft’s Azure cloud platform. However, the competition for cutting-edge AI hardware remains fierce, impacting both paying customers and internal operations. Suleyman celebrated the launch on social media, expressing excitement that his team would be the first to leverage the capabilities of the Maia 200 as they work on their forward-looking AI initiatives.