Microsoft is overhauling how it charges for Copilot, shifting the AI assistant to a usage-based pricing model. The company is also evaluating DeepSeek's architecture for enterprise deployments. The changes could make Copilot more scalable, but they bring new worries about cost predictability and where customer data ends up.
How the pricing model changes
Until now, Microsoft offered Copilot under flat-rate subscription tiers. Under the new usage-based model, customers will pay per query or per compute unit — a structure common in cloud services but novel for a productivity AI tool. Microsoft says the move lets businesses scale Copilot usage up or down without committing to a fixed number of seats. But it also means monthly bills could swing wildly if usage spikes, especially in large organizations where employees rely on the assistant for everything from drafting emails to summarizing meetings.
Why DeepSeek is part of the conversation
At the same time, Microsoft is looking at DeepSeek’s open-weight large language model as a potential foundation for enterprise Copilot instances. DeepSeek, developed in China, has gained attention for its efficiency and strong performance on benchmarks. Bringing it into Microsoft’s stack would mark a departure from the company's heavy reliance on OpenAI's GPT models and give enterprise customers another option. But the model's origin raises questions about compliance with data residency rules and export controls — issues Microsoft's legal and security teams are likely weighing.
What data sovereignty worries mean
Enterprise customers in Europe and Asia have long pushed for AI tools that keep data within national borders. Usage-based pricing adds a new layer: when every query is metered, the location of the server processing that query matters. If Microsoft routes a request through a data center outside the customer’s home region to handle peak load, the company could run afoul of local laws. Microsoft hasn't detailed how it will guarantee data stays in-region under the new pricing model. That uncertainty could slow adoption among regulated industries like banking and healthcare, where data sovereignty isn't optional.
The company is expected to share more details at its annual Build conference this spring. For now, enterprise customers are left to decide whether the flexibility of usage-based billing is worth the unpredictability — and whether DeepSeek's efficiency is worth the compliance headache.
