Microsoft is quietly shifting away from OpenAI's technology in some of its most widely used productivity tools. According to a report from Bloomberg, the company's Excel and Outlook applications are increasingly handling user prompts with Microsoft's own in-house AI models, known as MAI, rather than relying on OpenAI's systems.
The change is part of a broader effort to rein in the soaring costs associated with artificial intelligence. Microsoft has invested heavily in OpenAI, integrating its models into products like Copilot for Microsoft 365. But as AI usage scales, the expense of running those models on the backend has become a significant line item.
Why Microsoft Is Building Its Own AI
Microsoft's relationship with OpenAI has been one of the defining partnerships of the AI boom. The software giant has poured billions into the startup and integrated its GPT models into everything from Word to Azure cloud services. But relying on a third party for core technology comes with both financial and strategic costs.
By developing and deploying its own MAI models, Microsoft gains more control over performance, pricing, and data privacy. It also reduces the licensing fees it pays to OpenAI. For everyday users, the switch may be invisible—the prompts and responses in Excel and Outlook look and feel the same. But behind the scenes, Microsoft is aiming to lower the per-query cost of AI assistance.
This shift comes at a time when Microsoft is under pressure to show that its massive AI investments are paying off. The company has been spending heavily on data centers and chips to support AI workloads, a theme that surfaced in recent reports about job cuts in its Xbox division. As noted in our coverage of Microsoft cutting 4,800 jobs in an Xbox restructure, the company is reallocating resources toward AI while trimming elsewhere.
What It Means for Investors
For investors, this development signals that Microsoft is serious about managing the cost side of the AI equation. While the revenue potential from AI-powered features is enormous—from higher subscription tiers to new enterprise tools—the expenses have been climbing just as fast.
Swapping out OpenAI's models for its own could improve profit margins on AI products over time. It also reduces dependency on a single partner, which is a strategic hedge. If OpenAI were to raise prices or change its terms, Microsoft would have more flexibility to adapt.
That said, building competitive AI models is not cheap. Microsoft is investing in its own research and infrastructure, which adds to capital expenditures. The company's recent decision to cut 4,800 jobs, partly to offset AI data center costs squeezing cash flow, shows that the financial trade-offs are real.
Investors should watch for signs that Microsoft's in-house models can match or exceed OpenAI's performance in productivity tasks. If they do, the cost savings could be substantial. If they fall short, the company may need to maintain dual systems, which would limit the financial benefit.
The Bigger Picture for AI Spending
Microsoft is not alone in trying to bring AI costs under control. Across the tech industry, companies are realizing that the initial excitement around generative AI is giving way to a more sober assessment of its economics. Running large language models at scale requires enormous computing power, and the bills add up quickly.
For everyday investors, the key takeaway is that AI is still in its early innings, and the business models around it are evolving. Companies that can deliver AI features without blowing up their cost structures will be better positioned in the long run. Microsoft's move to develop its own models is a step in that direction, but it will take time to see the full impact on earnings.
As the AI landscape shifts, Microsoft's strategy of building in-house capabilities while maintaining ties with OpenAI gives it options. For now, the company is betting that its own MAI models can handle the workload—and do it more cheaply.
