Microsoft bets $33B on NVIDIA AI chips to fuel ChatGPT
Microsoft has inked over $33 billion in deals with neocloud providers—including a massive $19.4 billion arrangement with Nebius Group NV—to get access to more than 100,000 NVIDIA GB300 AI chips.
The goal? To power its own AI projects and meet surging demand from both its internal teams and OpenAI's ChatGPT.
Neoclouds are the new tech heroes
Neoclouds are small, specialized companies that lease out AI-focused computing power, helping Microsoft speed up training for new models.
By teaming up with them, Microsoft can let its main servers focus on things like large language models and smart assistants.
The company is also boosting capacity at its Racine, Wisconsin data center to nearly 900 megawatts—think of it as leveling up their tech base.
The AI arms race
The explosion in demand comes from both Microsoft's own products and the wild popularity of ChatGPT (which hundreds of millions use).
By locking in deals with Nebius, CoreWeave, Nscale, and Lambda, Microsoft is making sure it has the muscle to keep pace as everyone wants smarter AI tools.