Microsoft and OpenAI broke up — now they’re ready to fight
Microsoft used Build to show off in-house AI models, agents, and security tools as it moves beyond OpenAI dependence.
Intelligence analysis by GPT-5.4 Mini
Microsoft is signaling that its AI strategy is no longer built around OpenAI alone. At Build, it unveiled new models, agents, and a super app push while framing the moment as proof it can build frontier AI on its own.
Microsoft used to lean on OpenAI like a kid borrowing a friend’s bike. Now it is building its own bike, with its own parts, and showing it off at Build so it can race on its own.
Analysis
Microsoft’s new posture
At Build, Microsoft presented a broad AI slate: a new reasoning model called MAI-Thinking-1, six additional models for image, voice, transcription, and coding, a cybersecurity tool called MDASH, and AI agents tied to Windows and its Copilot ecosystem. The article frames this as a clear sign that Microsoft wants to be seen as a top-tier AI company in its own right, not just OpenAI’s partner.
From dependence to competition
For years, Microsoft leaned heavily on its relationship with OpenAI. That relationship has now effectively split, even though Microsoft remains OpenAI’s primary cloud partner for the moment. Mustafa Suleyman said Microsoft’s goal is to become “one of the top four labs in the world” and to build frontier models from the ground up with its own IP and data. He also said the company renegotiated its contract with OpenAI, which gave it room to train models at a larger scale and pursue superintelligence independently.
What Microsoft is emphasizing
Microsoft is stressing two things: capability and control. MAI-Thinking-1 is described as being built from scratch for math, coding, and enterprise deployment, and Suleyman said it does not rely on distillation from another company’s model. Microsoft also highlighted price and performance, arguing that its model can be cheaper than OpenAI equivalents on some tasks. That matters in a market where customers are increasingly sensitive to AI costs.
Agents, security, and the super app race
The company is also pushing AI agents as a practical product layer. It wants OpenClaw-style agents to work well on Windows and is promoting a Copilot super app that ties together agents and other Microsoft products. At the same time, Microsoft is selling MDASH as a way to coordinate 100 AI agents for cybersecurity work, directly positioning itself against rivals like Anthropic and OpenAI in enterprise and government markets.
Key points
- Microsoft used Build to signal that it wants to compete as a serious AI lab, not just OpenAI’s partner.
- The company introduced MAI-Thinking-1, its first reasoning model, plus six other new models for different AI tasks.
- Mustafa Suleyman said Microsoft wants to prove it can build frontier models from scratch with its own IP and data.
- Microsoft also highlighted MDASH, a cybersecurity tool that coordinates 100 AI agents.
- The company is pushing OpenClaw-style agents and a Copilot super app as part of its broader AI strategy.
If Microsoft’s new models and agents perform well, it could become a stronger independent player in AI and offer customers more choice. The company’s focus on lower-cost enterprise tools and Windows integration could make AI easier to adopt at work.
Microsoft is still behind OpenAI and Anthropic in reasoning models, so its new efforts may not close the gap quickly. If the products do not match the quality or cost claims, the company could end up with expensive in-house AI that still relies on rivals for the most advanced capabilities.
