Microsoft is significantly expanding its AI infrastructure to better compete with OpenAI and Anthropic. This expansion involves a major investment in computing clusters specifically designed for AI model training. Microsoft aims to achieve self-sufficiency in AI, acknowledging the increasing tensions as companies launch competing products and expand partnerships.
Microsoft is developing its own AI reasoning models and testing models from xAI, Meta, Anthropic, and DeepSeek. The company recently unveiled its first self-developed large model, trained using 15,000 Nvidia H100 chips. Microsoft's CEO, Satya Nadella, has stated the company will employ a multi-model strategy across its product range, incorporating AI models preferred by customers. Microsoft has also signed a five-year agreement with Nebius Group for AI computing infrastructure worth $17.4 billion, potentially reaching $19.4 billion.
Microsoft is also expanding its data centre infrastructure in the Netherlands to meet the growing hardware demands for AI and data storage. The new data centre is expected to create approximately 350 jobs and will utilise rainwater for cooling.
Related Articles

OpenAI, Microsoft renegotiate key terms
Read more about OpenAI, Microsoft renegotiate key terms →
Microsoft Diversifies AI Sources
Read more about Microsoft Diversifies AI Sources →
Microsoft MAI Models Debut
Read more about Microsoft MAI Models Debut →
Microsoft Tests In-House AI
Read more about Microsoft Tests In-House AI →
