Microsoft's cloud, where OpenAI lives and runs. The company that supplies the computers under ChatGPT also owns a stake in OpenAI.
In practice, Azure is a network of data centers around the world, full of processors specialized for AI (chips called GPUs). A company like OpenAI doesn't buy these machines, it pays Microsoft to use them over the network, exactly as much as it needs at any given moment. When you ask ChatGPT something, the request runs through these Azure machines, not through a server owned by OpenAI itself.
Azure exists because Microsoft long ago decided to sell not just software, but the raw computing power behind it too. It's a business like electricity: you pay for what you use instead of building your own power plant. For companies like OpenAI, where training one model costs billions in machine time, this is the only practical path. Building your own data center takes years, Azure gives you the power right away.
But the relationship with OpenAI is more tangled than a simple customer-supplier one. Microsoft has invested billions in OpenAI, and after the restructuring in the fall of 2025 holds a stake of about 27% in the company. That means it sells OpenAI computing time at a price and profits from its success as a co-owner at the same time. If OpenAI does well, Microsoft wins twice - once from the cloud bill, a second time from the value of its stake.
So when you read news about OpenAI, Azure is almost always standing behind the scenes. No Azure. No ChatGPT. That's where the risk analysts talk about comes from: OpenAI is financially intertwined with a single company that is at once its supplier, its investor, and, in the long run, its competitor, because Microsoft is also building its own AI products.