Microsoft's Internal AI Push
Microsoft Integrates In-House MAI Across Copilot and Office, Reducing Costs and External Dependence
Key Points
Microsoft unveiled seven first-party MAI models at its Build developer conference, signaling a broad, production-ready AI lineup beyond the initial MAI offerings.
Microsoft’s in-house coding AI is described as capable of matching or beating leading coding assistants while using less power, illustrating efficiency gains in internal tooling.
AI chief Mustafa Suleyman has publicly said Microsoft is trying to cut spending on Anthropic by using more of its own systems, underscoring the cost-avoidance motive behind the MAI push.
Microsoft is winding down most internal Claude Code licenses as part of its broader cost-optimization, with mid-May 2026 cited as the timeline for license wind-downs.
Microsoft is replacing OpenAI and Anthropic models with its own in-house AI — called MAI — inside core apps like Excel and Outlook, according to Crypto Briefing. The move is aimed at cutting costs and giving Microsoft more control over its AI systems.
Tens of thousands of prompts per week are already being routed to MAI models, Lapaas Voice reported. Microsoft AI chief Mustafa Suleyman has been direct about the motive: spend less on Anthropic by using more of Microsoft's own tech.
Microsoft Builds Its Own AI to Cut Outside Costs
Microsoft unveiled seven first-party MAI models at its Build developer conference. These models are designed for efficiency — doing more work at a lower cost per token than external models. One internal coding AI reportedly matches or beats leading coding tools while using less power.
The strategy is simple: keep routine, high-volume tasks in-house. Save the more complex work for outside providers like OpenAI when needed. This tiered approach lets Microsoft reduce its biggest AI bills without giving up top-end performance, according to Lapaas Voice.
Claude Licenses Are Being Wound Down Inside Microsoft
Microsoft is winding down most of its internal Claude Code licenses. The timeline for those wind-downs is mid-May 2026, according to Head Topics. Claude Code is Anthropic's AI coding assistant, and it has been widely used inside Microsoft's engineering teams.
Suleyman has publicly said Microsoft is actively trying to cut Anthropic spending by shifting to its own systems. That is a blunt signal from Microsoft's top AI executive — and it puts real pressure on Anthropic's business with its biggest customer.
A 2025 Deal Unlocked Microsoft's Ability to Compete
Microsoft and OpenAI renegotiated their deal in 2025. The new terms loosened the exclusivity rules that had tied Microsoft to OpenAI's models. That change gave Microsoft the legal room to build and deploy its own competing AI — which is exactly what it is now doing, Crypto Briefing reported.
Microsoft still has a broad partnership with OpenAI. But the renegotiated deal means the two companies are now, in part, competitors inside Microsoft's own product stack. That is a notable shift in a relationship that once made OpenAI the default engine for nearly all of Microsoft's AI products.
Anthropic's Valuation and AI Pricing Now Under Pressure
Market observers say the MAI push could hurt Anthropic's valuation. If Microsoft — one of Anthropic's biggest customers — is actively replacing Claude with its own models, that shrinks a key revenue stream. It also puts pressure on AI pricing across the industry, according to Head Topics.
The broader trend is clear: big tech companies are moving toward first-party AI to control costs. Microsoft is leading that shift. Analysts expect more of its Office products — beyond Excel and Outlook — to run on MAI as the rollout expands, Digitimes noted.
Sentiment
Mostly PositivePublishers
34
Articles
82
Reach
116
