Microsoft’s AI business is being shaped by a striking concentration: according to a Bloomberg report, OpenAI accounts for $24.1 billion during the fiscal year ending in June, roughly 70 percent of Microsoft’s total AI revenue.
That figure gives important context to Microsoft’s public positioning on artificial intelligence. The company benefits heavily from OpenAI’s growth, but it is also talking more openly about model diversity, open-weight models, and the dangers of too much value being captured by a small group of proprietary AI systems.
OpenAI is the center of Microsoft’s AI revenue picture
The Bloomberg report places OpenAI-related revenue at $24.1 billion for the fiscal year ending in June. That is not a side note in Microsoft’s AI story. It is reportedly the majority of the company’s AI revenue, at roughly 70 percent.
CEO Satya Nadella said in late March that Microsoft’s AI business was on track to top $37 billion annually. Read alongside the Bloomberg figure, the message is clear: OpenAI is a central driver of the business Microsoft is building around AI.
The arrangement between the companies explains why the revenue relationship is so large. Under their agreement, OpenAI pays Microsoft for computing power, model development costs, and a revenue share. In practical terms, OpenAI’s demand for infrastructure and model work flows back into Microsoft’s AI revenue line.
That makes the relationship powerful, but it also creates a strategic question. If one partner provides such a large share of reported AI revenue, Microsoft has reason to keep that relationship strong while also reducing the risk of being too dependent on any single model provider.
Microsoft’s messaging looks different in that context
The same company long associated with vendor lock-in has recently been championing open-weight models. It has also warned against a future in which a small number of proprietary AI models capture the value of entire industries.
Those positions can seem surprising if viewed only through Microsoft’s history. They make more sense when viewed through the current AI revenue mix described in the source. Microsoft is earning heavily from OpenAI, yet it is also signaling that the AI market should not be controlled by only a few closed systems.
This is not necessarily a contradiction. A company can profit from a leading proprietary model provider and still want a broader ecosystem that gives customers and developers more options. For Microsoft, model choice can support its AI business even when one partner is currently responsible for most of the reported revenue.
The emphasis on open-weight models also speaks to control. If customers can use more than one class of model, Microsoft can position itself around infrastructure, products, and deployment flexibility rather than only around access to one provider’s models.
The distillation dispute shows the pressure points
Nadella has also criticized AI labs like OpenAI and Anthropic for opposing "distillation," the practice of training models on proprietary model outputs to build competitors with less effort.
That issue matters because it sits directly inside the debate over who gets to benefit from model progress. Proprietary AI labs want to protect the value of their outputs. Competitors want ways to produce capable systems faster and with fewer barriers. Platform companies want a market where useful models can spread across products and workflows.
The source also notes that Chinese manufacturers have reportedly been especially aggressive with this approach. That adds another layer to the strategic tension: model output is not only a technical asset, but also a competitive input that others may use to advance their own systems.
For Microsoft, the issue is delicate. OpenAI is a major revenue source under the reported figures, but Microsoft’s broader AI strategy appears to depend on more than one model path. Nadella’s comments point to a future in which the company wants room for both proprietary systems and models developed through more open or derivative methods.
Office shows Microsoft is building its own path
Microsoft has also been steadily swapping in its own AI models across Office products. That matters because Office is one of the places where AI moves from infrastructure and model development into everyday software use.
The shift does not erase the role OpenAI plays in Microsoft’s AI revenue. Instead, it shows Microsoft working on a more layered approach. OpenAI remains a major financial engine, while Microsoft’s own models can become part of the products customers already use.
This strategy gives Microsoft more flexibility. If its AI revenue is heavily tied to OpenAI, internal model development can help the company shape product behavior, control costs, and decide where outside models are essential versus where its own systems are enough.
The bigger takeaway is that Microsoft’s AI business is not just about one agreement or one set of models. The reported $24.1 billion figure shows how important OpenAI has become. The company’s public support for open-weight models, its warnings about proprietary concentration, its stance on distillation, and its own model work in Office show why Microsoft is also preparing for a market that may not stay centered on a single AI supplier.