Microsoft is aggressively pursuing self-reliance in the artificial intelligence domain through substantial investments in computing capacity and infrastructure expansion. This strategic shift aims to reduce dependence on external AI model providers and secure a leading position in the rapidly evolving tech landscape.
To power its AI ambitions, Microsoft is making "significant investments" in physical infrastructure. This includes expanding its own AI chip clusters to boost the computing power needed to build its own models. Microsoft AI CEO Mustafa Suleyman emphasized the importance of self-sufficiency in AI for a company of Microsoft's size and diversity. Suleyman noted that the company's previously released preview models were trained on a "tiny cluster" of GPUs, specifically 15,000 Nvidia H100 GPUs. In comparison, OpenAI CEO Sam Altman anticipates having over 1 million GPUs online by the end of 2025. Microsoft's goal is to have the capacity to build world-class frontier models in-house, while remaining pragmatic and utilizing other models when necessary.
Illustrating this commitment, Microsoft has entered into a five-year agreement with Nebius Group to secure AI computing infrastructure valued at $17.4 billion, potentially reaching $19.4 billion if options for additional services are exercised. This deal will provide Microsoft with advanced GPU-powered computing infrastructure, with deliveries expected to begin in late 2025. The agreement enables Microsoft to expand its cloud capacity without incurring additional capital expenditures, as Nebius will finance data center construction through cash flow from the contract and debt secured against it. The new data center will be located in Vineland, New Jersey, and will provide dedicated GPU infrastructure capacity.
This infrastructure expansion is crucial for training and running AI models, which demand massive amounts of high-performance GPUs. By choosing Nebius as a supplier, Microsoft aims to ensure it has the necessary resources to compete with Amazon and Google in the cloud and AI markets. The deal also benefits Nebius, transforming it into a major player in the AI infrastructure market and guaranteeing billions in revenue.
Microsoft's strategy involves both building and leasing computing capacity. While the company is investing heavily in its own infrastructure, it also plans to lease capacity from other providers. This approach allows for faster scaling and capital efficiency, enabling Microsoft to focus investments on software, applications, and customer services.
The AI infrastructure market is experiencing rapid growth, with global spending on AI chips and cloud capacity projected to exceed $200 billion by 2030. Data centers powering AI workloads already account for a significant portion of global electricity use, and this figure is expected to rise as AI adoption increases. Microsoft is therefore under pressure to balance growth with environmental responsibility, investing in renewable energy and improving data center efficiency. Microsoft researchers have also developed a prototype analog optical computer (AOC) that uses light rather than digital switches for calculations, potentially reducing the energy demands of AI by a significant margin. The AOC is estimated to improve energy efficiency by a factor of one hundred.
Microsoft's commitment to AI extends beyond infrastructure. The company is investing heavily in research and development, creating its own AI models and integrating AI capabilities into its existing cloud services. This comprehensive approach positions Microsoft to capitalize on the growing demand for AI solutions and maintain its competitive edge in the tech industry.
