The ** Microsoft AI cost-cutting** strategy has officially begun, and it’s shaking up the balance of power among the world’s biggest AI providers. According to a Tuesday report from Bloomberg, Microsoft has started swapping out OpenAI and Anthropic models for its own in-house MAI models inside two of its most popular productivity tools: Excel and Outlook.
Why Microsoft Is Cutting AI Costs Now
Tens of thousands of prompts in Excel and Outlook are now being handled every week by Microsoft’s proprietary MAI models instead of third-party systems. This Microsoft AI cost-cutting push comes as the financial burden of running AI at scale keeps climbing across the entire tech industry. Every AI response carries a compute cost, and as Copilot features spread across hundreds of millions of Office seats, that bill adds up fast.
Microsoft isn’t alone here. Amazon, Uber, Meta, and Accenture have all made moves recently to rein in AI spending, and some companies are even eyeing cheaper Chinese models as alternatives — despite data security concerns.
The MAI Model Lineup Behind the Shift
This Microsoft AI cost-cutting effort didn’t come out of nowhere. At its Build conference last month, Microsoft unveiled seven new MAI models, including:
MAI-Thinking-1— a reasoning model built for coding, math, and multi-step problem-solving** MAI-Code-1-Flash**— a low-cost, inference-efficient coding model** MAI-Image-2.5**— an image generation and editing model** MAI-Voice-2and MAI-Transcribe-1.5**— speech generation and transcription models
Microsoft claims its coding model can match the performance of Anthropic’s Opus 4.6, though independent benchmarks released around Build suggested MAI-Thinking-1 still trails behind OpenAI and Anthropic’s top offerings.
What This Means for OpenAI and Anthropic
Microsoft isn’t cutting ties with its AI partners. OpenAI and Anthropic still handle the majority of Copilot traffic, and Microsoft’s multibillion-dollar stakes in both companies remain intact. But Microsoft AI CEO Mustafa Suleyman has been blunt about the long-term goal, saying the company wants to “reduce and ultimately eliminate” what it pays Anthropic.
The freedom to build competing models traces back to a 2025 renegotiation of Microsoft’s OpenAI contract, which ended Microsoft’s exclusivity requirement. That deal also let OpenAI sell through rivals, opening the door for AWS to offer OpenAI’s models too.
A Bigger Bet Behind the Scenes
CEO Satya Nadella has reportedly worried about Microsoft becoming “the next IBM” if it leaned too heavily on a single AI partner. Owning models — not just renting them — is central to avoiding that outcome. Nadella has also framed the shift as bigger than cost savings, saying Microsoft wants to move from simply consuming frontier models to fully participating in the frontier AI ecosystem.
For everyday users, the interface won’t change. But under the hood, this Microsoft AI cost-cutting strategy could eventually mean customers pay the same price for AI features while Microsoft pockets the difference — with premium third-party models like GPT or Claude potentially becoming paid add-ons down the line.