Microsoft to spend heavily to build its own AI chip cluster and become 'self-sufficient,' AI CEO says in leaked meeting (www.businessinsider.com)

Microsoft is ramping up investments in its own AI chip cluster to achieve self-sufficiency in AI development, signaling a strategic shift away from heavy reliance on OpenAI, according to Microsoft AI CEO Mustafa Suleyman. While the two companies remain partners, with Microsoft continuing to support OpenAI via Azure cloud services, Microsoft aims to independently build cutting-edge AI models using a combination of open-source technologies, collaborations with other AI developers, and internally developed models. This move underscores Microsoft’s desire for greater control and flexibility in AI innovation amid ongoing contract renegotiations with OpenAI. Key technical details reveal that Microsoft’s first fully in-house foundation model, MAI-1-preview, was trained on a relatively small cluster of 15,000 Nvidia H100 GPUs—far smaller than the massive clusters used by rivals like Google and Meta. Suleyman emphasized the need for expanding Microsoft’s AI training infrastructure through significant investments in proprietary AI chip clusters. This expansion is critical for scaling and competing with leading AI labs, enabling Microsoft to build “world-class frontier models” while pragmatically integrating external models when beneficial. CEO Satya Nadella reaffirmed the company’s dual approach of fostering its partnership with OpenAI alongside aggressive development of internal AI capabilities, positioning Microsoft for a more autonomous future in the rapidly evolving AI landscape.