According to Bloomberg on July 7, Microsoft has begun gradually introducing its internally developed MAI series models into core Office products such as Excel and Outlook, replacing its previous heavy reliance on OpenAI and Anthropic models. Currently, tens of thousands of AI prompts per week in Excel and Outlook are already being handled by MAI models. This shift marks a strategic adjustment in Microsoft's AI infrastructure, taking a key step from 'partner dependence' toward a 'self‑built closed loop'.
Strategic background: cost pressure and the drive for autonomy
At the core of Microsoft's move is the need to control rapidly rising AI costs. As a major investor and partner of OpenAI, Microsoft spends heavily each year on AI infrastructure. Introducing in‑house MAI models can significantly reduce reliance on external APIs and directly lower cloud service invocation costs. At the same time, this reflects Microsoft's strategic consideration of AI technology sovereignty: by developing models internally it can master core technologies and mitigate supply‑chain risks.
The MAI series models are the result of Microsoft's AI teams building from the ground up, covering multimodal capabilities such as reasoning, coding, and image generation. They are trained and optimized for enterprise scenarios and offer relatively high reliability and cost efficiency. For routine office tasks like Excel data analysis and Outlook email drafting, MAI models can already meet most needs.
Technical approach: implementing a model routing mechanism
Microsoft has not completely abandoned external models, instead adopting an intelligent model routing (Model Routing) strategy:
- Simple, routine tasks → MAI in‑house models (lower cost, faster response)
- Complex, creative tasks → still call flagship models such as OpenAI/Anthropic
This hybrid architecture both ensures performance and achieves cost optimization. The MAI models' handling of tens of thousands of prompts in production demonstrates their usability in real office scenarios.
Far‑reaching impact: supply‑chain restructuring and ecosystem competition
1. Impact on Microsoft itself
- Cost structure optimization: Significantly reduce the marginal cost of AI features like Copilot, improving overall gross margins.
- Improved product competitiveness: In‑house models better fit the Office ecosystem, enabling deeper integration and more customized experiences.
- Enhanced technical autonomy: Reduce external dependence and gain control of core AI capabilities, laying a foundation for long‑term competition.
- Positive capital market reaction: After the news, Microsoft's share price rose slightly as investors recognized its cost‑control ability.
2. Impact on OpenAI and Anthropic
- Revenue pressure: As Microsoft is an important customer for both companies, diverted traffic will directly affect their API revenues.
- Changing competitive landscape: Accelerates the trend of cloud giants developing in‑house AI, forcing a redefinition of partner relationships.
- Strategic reflection: OpenAI and others may need to strengthen partnerships with more enterprises or accelerate their own commercialization paths.
3. Impact on the broader AI industry
- Accelerated decentralization: More companies may prefer building or adopting hybrid AI stacks, reducing dependence on a few large model providers.
- Layered technological iteration: Competition in base models will intensify, and mid‑tier models for office scenarios will see dedicated optimization waves.
- Ecosystem restructuring: Drives AI from 'general large models' toward 'vertically integrated platforms', accelerating enterprise AI adoption.
4. Impact on ordinary users and enterprises
- Upside: Lowered costs for Office AI features could help stabilize subscription prices or enable more free features.
- Potential challenges: In‑house models may initially lag flagship models in performance and will require iterative improvement.
- Privacy and compliance advantages: Processing enterprise data on internal models can help improve data security and regulatory compliance.
Industry takeaway: AI strategies must balance cost and innovation
Microsoft's move reaffirms that AI is not merely a technical race but also a contest of business models. With compute costs high, efficiently using owned resources and building closed‑loop ecosystems has become a critical task for tech giants.
For Chinese AI companies, this episode offers an important lesson: while relying on international cutting‑edge models, they should accelerate in‑house development to build a controllable AI technology stack. Meanwhile, the AI‑ization of productivity tools like Office also provides a catching‑up opportunity for domestic office software.
Microsoft's shift from 'dependence' to 'self‑reliance' is one sign of the AI industry's maturation. Going forward, more companies may follow, jointly driving AI from 'cash‑burning experiments' toward a new phase of sustainable commercialization.
▶️ Enjoy this article? Creating content is hard work!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
More AI news at: https://www.airosetta.com