Microsoft AI Chief Mustafa Suleyman: 'Our Goal Is to Eliminate What We Pay Anthropic'

Sources indicate that each developer using external tools was costing Microsoft between $500 and $2,000 a month . At that scale, the economics of relying on an outside provider become difficult to justify, especially when internal alternatives are maturing rapidly.

Seven in-house models to replace the spend

Just days before the Bloomberg interview, Microsoft's AI Superintelligence Team unveiled a family of seven models at the Build 2026 developer conference in San Francisco. The new MAI lineup includes:

• MAI Thinking-1, a reasoning model
• MAI Voice-2 and MAI Voice-2 Flash
• MAI Image-2.5 and MAI Image-2.5 Flash
• MAI Transcribe-1.5
• MAI Code-1-Flash

This release is Microsoft's most aggressive push yet to create credible in-house alternatives to models from both Anthropic and OpenAI . Suleyman described the effort in a blog post as being "about long term self-sufficiency for Microsoft and our partners. It's about models you can trust" .

On performance, Microsoft claims that after tuning its models for consulting firm McKinsey, the company outperformed OpenAI's GPT-5.5 on cost efficiency by a factor of ten . Separately, the company says its MAI-Thinking-1 reasoning model matches Anthropic's Claude Opus 4.6 on a widely used coding benchmark—but at a lower price point .

The broader independence push

Suleyman's comments about Anthropic fit into a pattern of statements he has made over the past year. He has described his "personal mission at Microsoft" as building superintelligence and achieving "true AI self-sufficiency"—a goal that applies to the company's relationship with OpenAI as much as with Anthropic .

In a separate interview with Semafor, Suleyman acknowledged that Microsoft is in a game of catchup, building frontier AI models, custom accelerator chips, and high-performance "harnesses" to compete with tools from Cursor, Anthropic, OpenAI, and Google. "We got here in six months, which is itself a remarkable achievement," he said. "We're now neck and neck with essentially what was state of the art just a few months ago" .

The financial commitment behind this rhetoric is enormous. Microsoft is investing more than $80 billion in AI data-center infrastructure, including custom accelerator chips and the Majorana 2 quantum chip, to support running its own models at scale instead of paying external providers . Earlier in 2026, Microsoft also released a speech transcription model priced at 36 cents per hour, underlining the company's focus on driving down costs across its AI portfolio .

A careful balancing act

For all the tough language, Microsoft's relationships with its AI partners are not being severed overnight. The company maintains a multibillion-dollar Foundry partnership with Anthropic that gives its customers access to Claude models, and its internal directive to switch from Claude Code to GitHub Copilot CLI does not affect that broader commercial arrangement .

Similarly, while Microsoft pushes for independence from OpenAI, the two companies remain contractually tied. CEO Satya Nadella has previously admitted, "We are over-dependent on OpenAI" . The new directive from Suleyman drives that point home operationally, but structurally, the partnerships continue.

What has changed is the direction of travel. Microsoft is no longer content to serve as the cloud platform through which other companies' AI models are delivered. It now intends to compete directly on the frontier—and its AI chief has made clear that the company views Anthropic's pricing as an expense line to be eliminated, not a cost of doing business.