Microsoft Copilot Cowork: usage-based billing and Deepseek V4
Microsoft will move Copilot Cowork from flat-rate plans to usage-based pricing and may offer a self-hosted, fine-tuned Deepseek V4 on Azure.
TL;DR
- 01Microsoft will move Copilot Cowork from flat-rate plans to usage-based pricing and may offer a self-hosted, fine-tuned Deepseek V4 on Azure.
- 02Microsoft is weighing a self-hosted, fine-tuned version of Deepseek V4 as a cheaper model option for Copilot Cowork while shifting Cowork from flat-rate plans to usage-based billing.
- 03The company says a final decision is expected in the coming weeks and frames the change as part of a broader model-choice strategy.
Microsoft is weighing a self-hosted, fine-tuned version of Deepseek V4 as a cheaper model option for Copilot Cowork while shifting Cowork from flat-rate plans to usage-based billing. The company says a final decision is expected in the coming weeks and frames the change as part of a broader model-choice strategy.
What is changing with Copilot Cowork?
Microsoft plans to move Copilot Cowork from flat-rate pricing to usage-based billing because heavy users are driving costs up quickly. Copilot EVP Charles Lamanna told Axios that flat-rate pricing "isn't sustainable because of 'users who do hundreds of tasks a week,'" and Microsoft already moved GitHub Copilot to usage-based billing for the same reason.
The product currently adapts Anthropic's Claude technology. The source notes Claude "leans heavily on agentic reasoning and burns through tokens fast," a behavior Microsoft cites as a driver of rising operational costs under flat-rate plans.
How would Deepseek V4 fit into the mix?
Microsoft is considering offering a self-hosted, fine-tuned Deepseek V4 as an optional, lower-cost model choice, hosted fully on Azure so customer data remains in Microsoft’s cloud. The company says the Deepseek option would be optional and fully hosted on Azure, and that the model has been customized with safeguards against bias.
The primary source frames Deepseek V4 as a cheaper model alternative to Claude-based Cowork, and it notes potential sensitivity because Deepseek is a Chinese AI model. Microsoft expects that choice to draw criticism, especially in the US, and emphasizes that customers would retain control because the model would run on Azure and be optional.
Why Microsoft is pushing a model ecosystem now
Satya Nadella has been arguing for an ecosystem of AI models companies can pick and tune for specific use cases and costs. In a blog post cited in the source, Nadella described AI as a consumption business and said he wants "intense users and intense usage."
The move toward usage-based billing and optional model choices fits that frame: usage-based pricing allocates cost to the heaviest consumers, while multiple model options let customers trade off price, capability, and hosting preferences. Microsoft is positioning model choice and pricing as levers to manage both cost and customer needs.
Why it matters
Shifting Copilot Cowork to usage-based pricing alters the economics for teams that run many agentic tasks. Heavy users who perform large volumes of automated tasks could see costs climb in flat-rate plans, which Microsoft says was the unsustainable outcome. Offering a self-hosted Deepseek V4 could lower per-task model costs for some customers and shift control of data to Azure-hosted instances, but the move also raises geopolitical and trust debates because the model originates from China.
What to watch
Microsoft expects to decide in the coming weeks whether to include a Deepseek V4 option. Watch for the formal pricing announcement for Copilot Cowork and any details on how usage will be metered, plus the exact guardrails Microsoft applies to the Deepseek deployment if it is approved.
Source attribution: Primary source material includes reporting that cites Axios and statements attributed to Copilot EVP Charles Lamanna and CEO Satya Nadella. The primary source is dated Jun 16, 2026.
Written by The Brieftide · Source: The Decoder
The Brieftide Daily · 06:00
Briefs like this one, in your inbox every morning.
Continue reading
More in Enterprise AI AdoptionOpenAI Partner Network launch: $150M fund to scale enterprise AI
OpenAI commits $150M to a Partner Network to help global partners accelerate enterprise AI adoption, deployment and transformation.
OpenAI Academy launches three courses for practical AI work
Three new Academy courses teach practical AI skills, repeatable workflows, and how to apply agents in everyday work.
BBVA and OpenAI: 100,000 employees using ChatGPT Enterprise
BBVA embedded ChatGPT Enterprise bank-wide, saving ~3 hours per employee per week and spawning more than 20,000 internal GPTs.
OpenAI: LSEG scales trusted AI, empowers 4,000 staff
LSEG uses OpenAI to scale trusted AI across its global business, accelerating insights, shrinking release cycles and empowering 4.