Microsft Might Add Deepseek To Copiliot
Microsoft is exploring a lower-cost AI model strategy for its Copilot Cowork product as rising usage costs push enterprise customers to their limits. According to a recent report, the company is testing a fine-tuned version of DeepSeek V4 as a potential budget-friendly alternative to the more expensive models currently powering Cowork. This move comes as Microsoft officially rolls out the tool globally and shifts to a usage-based pricing system.
The change in pricing reflects a deeper issue: AI agents are expensive to run. Unlike simple chatbots, Copilot Cowork handles complex, multi-step workflowsnretrieving data, calling tools, and iterating on tasks. Each of these steps consumes tokens and compute power, quickly driving up costs for businesses running hundreds of tasks per week. Currently, Cowork relies heavily on models from Anthropic, with OpenAI’s latest models available in premium tiers. But Microsoft is signaling that model diversity is key to making AI economically viable at scale. A cheaper model tier, potentially powered by DeepSeek or Microsoft’s own in-house “Cowork 1” model, is expected soon.
To address security concerns, especially around using a Chinese-developed model, Microsoft plans to host any DeepSeek-based offering on Azure. This ensures enterprise-grade protections, including data isolation, compliance controls, and tenant-level security boundaries. At the same time, Microsoft is introducing “Copilot Credits,” a usage-based system that charges customers based on how much AI they actually use. Admins can set spending limits across users and teams, giving organizations more control, but also making costs more visible. The broader takeaway is clear: while AI agents are powerful, their cost remains a major barrier to widespread adoption. Microsoft’s strategy suggests the future of enterprise AI won’t be about a single best model, but a mix of models optimized for performance, cost, and use case. God bless and Tech Talk To You Later!!
