Microsoft Shifts Office AI Workloads to Cut OpenAI and Anthropic Costs

Microsoft (MSFT ) shifts some Office AI workloads to MAI models, cutting OpenAI and Anthropic costs while keeping Copilot on a mixed model strategy.

Microsoft has started moving selected Office AI workloads to its own MAI models to reduce outside model costs. 

The shift affects some tasks in Excel and Outlook, while OpenAI and Anthropic still handle most Copilot traffic. 

The move shows Microsoft is turning AI deployment costs into a larger business priority.

Microsoft Routes More Office Tasks to MAI Models

Microsoft now uses internally built MAI models for tens of thousands of prompts each week in Office applications. 

The workloads cover selected tasks in Excel and Outlook, according to Bloomberg. However, the company still relies on OpenAI and Anthropic for many complex production requests.

The change does not mark a full break from external AI partners. Instead, Microsoft is using a mixed model strategy across its productivity software. 

The company can route simpler tasks to cheaper systems and reserve stronger models for heavier work.

That approach matters because Copilot consumes large amounts of computing power. 

Each user request uses tokens, GPU capacity, memory, networking, and safety systems. As adoption grows, small savings per prompt can reduce large operating costs.

MAI Push Follows Build Model Launch

Microsoft presented seven MAI models at its Build developer conference in June. The lineup included models for coding, reasoning, transcription, image generation, and voice work. 

The company positioned those systems as lower-cost options for selected enterprise workloads.

AI chief Mustafa Suleyman has pushed the company to reduce reliance on Anthropic models. He said Microsoft wanted to cut and eventually remove that cost over time. That message placed model economics at the center of the company’s AI strategy.

Microsoft has also made MAI models available in GitHub Copilot. In addition, its transcription model will support Teams and other products in future deployments. 

These moves show a broader effort to place internal AI across daily workplace tools.

According to a recent tweet from Mario Nawfal, Microsoft is also cutting around 4,800 jobs. The tweet said the reductions affect about 2% of the company’s global workforce. It also said sales teams and Xbox face some of the largest cuts.

Cost Control Becomes an AI Advantage

Microsoft says the layoffs do not come from AI directly replacing jobs. However, the company has acknowledged that AI continues to change how work gets done. At the same time, it keeps spending heavily on AI infrastructure.

The Office shift gives Microsoft more control over AI costs and product delivery. It can run its own models on Azure and reduce payments to outside providers. 

It can also choose models based on cost, speed, data needs, and task difficulty.

This strategy keeps OpenAI and Anthropic inside the product stack, but it changes their role. Microsoft now appears to treat external models as part of a wider portfolio. That portfolio gives the company more room to manage margins as AI usage grows.

For enterprise customers, the change points to a broader AI market shift. Companies now want systems that deliver the right performance at a sustainable price. Microsoft is betting that Office AI success will depend on efficient deployment, not only model power.