Microsoft Is Training Its Sales Team to Undercut OpenAI and Anthropic

Microsoft told its salespeople to actively position its own AI products against OpenAI, Google, and Anthropic at an internal strategy meeting on July 15, 2026. The meeting was reported by Bloomberg. Executives outlined how sales staff should highlight weaknesses in competitor products while pitching Microsoft’s in-house MAI models as the cheaper and faster alternative.

This is the same company that gave OpenAI billions of dollars to exist. Things have changed.

What Microsoft Said in That Meeting

Microsoft Executive Vice President Jay Parikh was direct with the room: “Everyone else is selling parts. We’re selling the full end-to-end system. That’s the story that we all need to get out there and tell in FY27.”

Copilot EVP Jacob Andreou went further. He delivered a presentation comparing Copilot directly to Anthropic’s Claude and said Claude was “slower and less accurate, and lacked the proper security integrations” inside Microsoft’s Office apps.

That is a pointed statement about a model Microsoft was actively paying to use until recently.

What the MAI Models Actually Are

At Build 2026 in early June, Microsoft launched seven new in-house models under the MAI banner, short for Microsoft AI.

The flagship is MAI-Thinking-1, running on 35 billion parameters with a 256K context window. Microsoft claims it can deliver up to 10 times better cost efficiency in tuned enterprise workloads compared to competing models. The company also says MAI-Thinking-1 matches or outperforms GPT-5.5 and Claude Opus 4.6 on key benchmarks, particularly in coding tasks on SWE-Bench Pro.

Those are bold claims. They are also exactly the kind of claims that enterprise sales teams need to walk into customer meetings.

Microsoft Already Started Replacing OpenAI and Anthropic

This sales strategy did not come out of nowhere.

Starting July 7, Microsoft began migrating selected Microsoft 365 applications, specifically Excel and Outlook, away from OpenAI and Anthropic models onto its own MAI stack. A person familiar with the matter confirmed that tens of thousands of AI prompts in those applications are now being completed each week using Microsoft’s internally built models.

AI model chief Mustafa Suleiman said in June the company was actively trying to reduce spending on Anthropic by using more MAI models. This sales push is the external version of an internal cost-cutting strategy that was already underway.

How the Microsoft-OpenAI Relationship Got Here

The history matters here.

Microsoft and OpenAI originally entered a unique agreement where Microsoft provided capital and compute to OpenAI in exchange for exclusive access to OpenAI’s API and models. That arrangement made Microsoft the only company that could build products on GPT technology.

That exclusivity ended in April 2026 when the two companies revised their agreement. OpenAI can now sell to any company, including Microsoft’s direct competitors. Microsoft, no longer locked in, is now free to compete.

The relationship did not collapse. Microsoft still holds a stake in OpenAI valued at roughly $135 billion as of late 2025. But the commercial incentives have shifted significantly, and the sales training reflects that shift in plain terms.

What This Means for OpenAI and Anthropic

For OpenAI, losing Microsoft’s distribution as a reliable closed channel is significant. Microsoft Azure was the primary route for enterprise customers to access GPT models. If Azure salespeople are now actively steering those customers toward MAI models instead, OpenAI loses one of its most effective sales channels without losing the investment relationship.

For Anthropic, the pressure is more direct. Claude models are central to Anthropic’s enterprise revenue story. If MAI-Thinking-1 is outperforming Claude Sonnet 4.6 on independent benchmarks and being actively sold to Microsoft’s enterprise customer base, that is a direct challenge to exactly the market segment Anthropic has been targeting most aggressively.

Meta is moving in the same direction. CEO Mark Zuckerberg signaled plans to price Meta’s new Model API roughly 25 percent below OpenAI and Anthropic, according to a JPMorgan analyst note from the same week.

The enterprise AI market is getting competitive on price in a way it has not been before.

The Investor Pressure Behind All of This

Microsoft faces significant investor scrutiny over its AI spending. The company has committed billions to infrastructure, model development, and the OpenAI partnership, and investors want to see returns.

Promoting its own AI products over third-party models it has to pay for is one of the clearest paths to improving AI unit economics. Every enterprise workload that runs on MAI instead of GPT or Claude is a workload where Microsoft keeps both the revenue and avoids the API cost.

The sales training is a business decision as much as it is a competitive one.