Industry
Vol. 1·Wednesday, March 18, 2026
Microsoft Bets on Model Diversity, Bringing Claude Into the Heart of Copilot
Discussion
Sign in to join the discussion.
Sign in to join the discussion.
A federal judge blocked Perplexity's Comet agent from Amazon's site on March 10. Two days later, the company unveiled Personal Computer, a persistent AI agent running locally on a Mac mini. The two events are not coincidental - they define the strategic dilemma at the center of the agentic web.
For nearly seven years, Microsoft's AI strategy had a simple answer to almost every question: OpenAI. The two companies built what Microsoft itself called "one of the most consequential collaborations in technology" - a partnership that seeded Copilot into 90 percent of the Fortune 500 and reshaped the enterprise software market. The arrival of Claude inside mainline Copilot chat, announced March 9 as part of Wave 3 of Microsoft 365 Copilot, signals that the era of that exclusive dependence is over.
This is not a minor product update. Claude is now available directly in Copilot's primary chat interface via the Frontier Program, alongside OpenAI's latest models.[1] Enterprise users no longer need to navigate to a specialized tool or toggle a separate mode. The model simply shows up where work happens.
The path was methodical. In September 2025, Microsoft quietly opened the door by making Claude Sonnet 4 and Claude Opus 4.1 available as options inside the Researcher agent and Copilot Studio - useful, but clearly secondary to the default OpenAI models.[2] The financial architecture behind that move came into focus in November, when Anthropic agreed to purchase $30 billion of Azure computing capacity in exchange for a $5 billion investment from Microsoft and up to $10 billion from Nvidia.[3] Claude had already become the only frontier model available across all three of the world's leading cloud providers - a status the November deal deepened - and the new Azure spending commitment gave Microsoft a compelling commercial reason to feature it prominently.
Wave 3 is the product payoff on that financial groundwork. Judson Althoff, CEO of Microsoft Commercial Business, described Copilot as "model diverse by design" - operating as "an open, heterogeneous environment" that draws on "leading models from OpenAI and Anthropic, operating openly across clouds and data services without locking customers in."[1] That framing is pointed: it positions Microsoft, not any individual AI lab, as the stable platform enterprises should trust.
Beyond mainline chat, Claude plays a structural role in Copilot Cowork, the most significant new capability in Wave 3. Built in close collaboration with Anthropic, Cowork brings the technology powering Claude's own long-horizon task execution directly into Microsoft 365.[4] The design premise is a departure from the prompt-and-response model that has defined enterprise AI tools to date: instead of generating a single artifact, Cowork allows users to delegate multi-step work that unfolds over minutes or hours, reasoning across files, tools, and organizational context, with visible progress and opportunities to intervene.
Cowork is currently in research preview with a limited set of customers, with broader Frontier Program availability expected by the end of March.[4] It will sit alongside Agent 365, Microsoft's new control plane for enterprise AI agents, which reaches general availability on May 1 at $15 per user per month.[1] IDC projects 1.3 billion agents in circulation by 2028; Agent 365 is Microsoft's bid to govern that proliferation before it becomes a security liability.[4]
The OpenAI partnership has not collapsed - it has been redefined. A restructured agreement signed in October 2025 preserved Microsoft's exclusive IP rights and Azure API exclusivity for OpenAI models until AGI is declared, with those rights now extended through 2032.[5] OpenAI remains Copilot's primary model partner, and its latest generation of models continues to roll out in parallel with Claude's arrival. But the new agreement also granted both parties significant new freedoms: Microsoft can now independently pursue AGI with third parties, and OpenAI can jointly develop some products with external partners - with API products exclusive to Azure, and non-API products free to run on any cloud.[5]
The practical effect is a controlled decoupling - close enough to remain mutually beneficial, loose enough that neither company is structurally dependent on the other. Claude's elevation in Copilot is the clearest expression yet of how Microsoft intends to use that new latitude.
"Real differentiation comes from intelligence - deep work context, embedded in the tools people already use."
Judson Althoff, CEO, Microsoft Commercial Business
Microsoft's multi-model pivot reflects a hard-won lesson from the enterprise software market: customers resist lock-in, and the AI race is moving too fast for any single vendor bet to feel safe. Copilot paid seats grew more than 160 percent year over year in Microsoft's most recent quarter, with daily active usage up tenfold.[1] That momentum gives Microsoft leverage to set the terms of the emerging enterprise AI stack rather than simply distribute whatever OpenAI produces.
The new Microsoft 365 E7 "Frontier Suite," launching May 1 at $99 per user per month, bundles Copilot, Agent 365, Microsoft Entra Suite, and the full E5 security stack into a single SKU.[1] The pricing is calibrated to be cheaper than purchasing the components separately - and it locks enterprise buyers into a Microsoft-managed environment where model choice is a feature Microsoft controls, not a decision customers make by switching platforms.
That is perhaps the most consequential detail in Wave 3. By making multi-model flexibility a core Copilot capability, Microsoft transforms a potential vulnerability - the risk of being disrupted by a better model from a competitor - into a selling point. Claude inside Copilot is not evidence that Anthropic won. It is evidence that Microsoft has engineered a position where it profits regardless of who does.