Microsoft and OpenAI broke up — now they’re ready to fight
‘We have to prove that we can do everything that we need to from the ground up,’ said AI chief Mustafa Suleyman.
AI
Report
Tech
Microsoft and OpenAI broke up — now they’re ready to fight
‘We have to prove that we can do everything that we need to from the ground up,’ said AI chief Mustafa Suleyman.
At Microsoft’s annual Build conference on Tuesday, the company announced a slew of new or expanded AI initiatives, including a super app, in-house reasoning models, a cybersecurity tool, and OpenClaw-esque AI agents. All this news added up to a clear message: Microsoft is positioned to be one of the biggest players in AI, and it’s finally acting like it.
For years, Microsoft’s AI business leaned hard on its early and exclusive partnership with OpenAI. But the drama-filled marriage slowly devolved into a situationship, and the pair effectively separated in late April (though Microsoft is still OpenAI’s primary cloud partner — for now). This year’s Build had the vibe of a freshly single divorcée posting a thirst trap on Instagram. “It’s always fun to be at developer conferences in times of great change,” Microsoft CEO Satya Nadella said onstage Tuesday, adding that events like this are about “coming to grips with the new opportunity.”
AI chief Mustafa Suleyman, in an interview with The Verge, put it even more bluntly.
“The goal is to prove that we can become one of the top four labs in the world,” Suleyman said. “There’s three labs that matter, Google DeepMind, OpenAI, and Anthropic. We are not one of them at the moment, and that’s always been my intention. It’s why I came here. I want to build the very best frontier models in the world, fully multimodal, and in order to do that, we have to prove that we can do everything that we need to from the ground up, and we’re not just going to take from others.”
One of Microsoft’s first steps at Build was indeed to play catch-up on AI models. Suleyman unveiled MAI-Thinking-1, the company’s first reasoning model, along with six other new models focused on image, voice, transcription, and coding. Microsoft said the medium-size MAI-Thinking-1 model, which will likely be marketed to primarily enterprise clients, is “built from scratch for serious math, coding, and real-world enterprise deployment.” Microsoft is years behind both OpenAI and Anthropic here; OpenAI began releasing reasoning models in the fall of 2024. But Suleyman emphasized its performance on benchmarks like coding and its price point, saying it was cheaper than OpenAI equivalents on some tasks — a big deal in the age of the AI money squeeze, which has inspired a lot of complaints with customers.
Microsoft and OpenAI broke up — now they’re ready to fight
‘We have to prove that we can do everything that we need to from the ground up,’ said AI chief Mustafa Suleyman.
AI
Report
Tech
Microsoft and OpenAI broke up — now they’re ready to fight
‘We have to prove that we can do everything that we need to from the ground up,’ said AI chief Mustafa Suleyman.
At Microsoft’s annual Build conference on Tuesday, the company announced a slew of new or expanded AI initiatives, including a super app, in-house reasoning models, a cybersecurity tool, and OpenClaw-esque AI agents. All this news added up to a clear message: Microsoft is positioned to be one of the biggest players in AI, and it’s finally acting like it.
While Microsoft has had years to glean insights from OpenAI, Suleyman made sure to mention that its development did not involve any distillation, meaning that it wasn’t trained using a different company’s AI model. If MAI-Thinking-1 is good, Microsoft clearly doesn’t want people thinking it’s due to the influence of OpenAI.
Suleyman told The Verge that for Microsoft, “the pivotal moment was renegotiating our contract with OpenAI. That meant that we were allowed to train models at a larger scale and explicitly pursue superintelligence entirely with our own IP, with our own data, no distillation, training from scratch.”
Nadella also highlighted Microsoft’s recently launched AI cybersecurity tool MDASH, which he said brings together 100 AI agents to find exploitable bugs “better than any single model.” It was clearly a dig at Claude Mythos Preview, which Anthropic introduced in April to much fear and fanfare — and expanded access to just before Build. OpenAI has its own cybersecurity-focused system as well, and all three companies will likely use their offerings to jockey for position in the government and enterprise markets they desperately need to court.
Microsoft is in a more complex situation with AI agents. The popular open-source platform OpenClaw demonstrated a potential path forward for AI agents, and after OpenAI quickly hired its creator, Peter Steinberger, Microsoft (among other companies) is trying to catch up. One of its key strategies is making OpenClaw work well with Windows. At Build, Nadella said he was very committed to OpenClaw support, and Microsoft employees chatted with developers in the audience about how they were using it.
Steinberger himself made a surprise appearance to great audience reaction, taking the stage to boast about how OpenClaw had bolstered its security and earned user trust. “What I kept hearing was, ‘Peter, I love my Claw, but can I use it at work?’” Steinberger said. “You can totally run OpenClaw inside your company now, and we even made the harness itself a plug-in.” Steinberger said that whether someone trusts Copilot, Codex, or another company’s coding platform, users can now run OpenClaw on top of that via Windows.
But Microsoft is also promoting its own separate Copilot “super app” that integrates OpenClaw-esque agents. A super app is a major focal point for OpenAI right now — president Greg Brockman is leading development of one that will tie together ChatGPT, the Codex coding platform, and the Atlas web browser. Microsoft’s strategy is similar, bringing together a variety of existing Copilot AI assistants. Its agents, called “Autopilots,” are designed to act as a helpful user interface. Cassidy Williams, GitHub’s senior director of developer advocacy, called Copilot “your home base for development and operations on your computer,” demonstrating how multiple agents could perform tasks like app-building. (In an extra flourish, Williams demonstrated how she could approve or deny code changes by flashing her computer camera a thumbs-up or thumbs-down.)
Autopilots are designed specifically to appeal to business customers — Nadella called them “autonomous, long-running agents with full enterprise compliance.” The first one Microsoft will offer is “Scout,” billed as “your always-on personal agent,” but clients can build and personalize their own. The Autopilot agents should be able to look through an email inbox, join group chats in Teams, check a calendar, and send daily briefings, among other things. Accordingly, employees on stage at Build repeatedly emphasized Copilot’s security tools and guardrails — obviously aiming to calm enterprise clients who may have heard horror stories about tools like OpenClaw.
Suleyman made sure to emphasize, again and again, Microsoft’s “humanist superintelligence” as an “AI that prioritizes humanity first” — part of AI companies’ recent rebrand of AGI to make it sound less frightening in an era when people are pushing back against AI more than ever before.
Nvidia CEO Jensen Huang, another speaker known for working closely with OpenAI, appeared via video call to tout how Nvidia’s RTX Spark chip is fueling Microsoft’s AI agent goals. “I could be traveling and I’m on the phone and I can text my PC … and it would fire up the tools on the PC,” Huang said. “The idea that the PC evolved from a personal computer to a personal AI is just really exciting.”
Microsoft spent years betting on OpenAI, and in some ways, that’s left it behind in the AI race. But as OpenAI and other competitors turn to enterprise to finally make money, it’s got some obvious advantages. Microsoft already has a substantial client base and, compared with other AI companies, a reputation for safety and security. And like Google, it also has deep pockets, considerable computing resources, and a diversified revenue stream, meaning it can take big bets without a ton of risk.
Suleyman told The Verge, “There’s a lot of people who are either like chasing startup valuations or about to IPO, so we can operate with a little bit more humility and a little bit more long-term optimization.” He added, “We’ve got the money to be able to buy Anthropic [models] when we need to. We’ve got the optionality in Azure with 11,000 models, so people can use literally whatever they want whenever they want, but that buys us the time to do it right from the start.”
At the same time, there are a lot of unanswered questions here. Microsoft called out a lot of benchmark wins and advancements for its seven new models, but that doesn’t always translate to real-world adoption, and even a new model that pulls ahead for a week or two can quickly fall behind. AI super apps are a mostly yet-untested idea. And Microsoft is entering a crowded yet still largely underwhelming AI agent marketplace with a product that we haven’t seen in action. There’s still plenty of room for its promises to fall flat.
Hayden Field
Tom Warren
AI
Analysis
Microsoft
Microsoft Build
OpenAI
Report
Tech
More in: Microsoft Build 2026: All the news about Windows, AI, RTX Spark, and more
For years, Microsoft’s AI business leaned hard on its early and exclusive partnership with OpenAI. But the drama-filled marriage slowly devolved into a situationship, and the pair effectively separated in late April (though Microsoft is still OpenAI’s primary cloud partner — for now). This year’s Build had the vibe of a freshly single divorcée posting a thirst trap on Instagram. “It’s always fun to be at developer conferences in times of great change,” Microsoft CEO Satya Nadella said onstage Tuesday, adding that events like this are about “coming to grips with the new opportunity.”
AI chief Mustafa Suleyman, in an interview with The Verge, put it even more bluntly.
“The goal is to prove that we can become one of the top four labs in the world,” Suleyman said. “There’s three labs that matter, Google DeepMind, OpenAI, and Anthropic. We are not one of them at the moment, and that’s always been my intention. It’s why I came here. I want to build the very best frontier models in the world, fully multimodal, and in order to do that, we have to prove that we can do everything that we need to from the ground up, and we’re not just going to take from others.”
One of Microsoft’s first steps at Build was indeed to play catch-up on AI models. Suleyman unveiled MAI-Thinking-1, the company’s first reasoning model, along with six other new models focused on image, voice, transcription, and coding. Microsoft said the medium-size MAI-Thinking-1 model, which will likely be marketed to primarily enterprise clients, is “built from scratch for serious math, coding, and real-world enterprise deployment.” Microsoft is years behind both OpenAI and Anthropic here; OpenAI began releasing reasoning models in the fall of 2024. But Suleyman emphasized its performance on benchmarks like coding and its price point, saying it was cheaper than OpenAI equivalents on some tasks — a big deal in the age of the AI money squeeze, which has inspired a lot of complaints with customers.
While Microsoft has had years to glean insights from OpenAI, Suleyman made sure to mention that its development did not involve any distillation, meaning that it wasn’t trained using a different company’s AI model. If MAI-Thinking-1 is good, Microsoft clearly doesn’t want people thinking it’s due to the influence of OpenAI.
Suleyman told The Verge that for Microsoft, “the pivotal moment was renegotiating our contract with OpenAI. That meant that we were allowed to train models at a larger scale and explicitly pursue superintelligence entirely with our own IP, with our own data, no distillation, training from scratch.”
Nadella also highlighted Microsoft’s recently launched AI cybersecurity tool MDASH, which he said brings together 100 AI agents to find exploitable bugs “better than any single model.” It was clearly a dig at Claude Mythos Preview, which Anthropic introduced in April to much fear and fanfare — and expanded access to just before Build. OpenAI has its own cybersecurity-focused system as well, and all three companies will likely use their offerings to jockey for position in the government and enterprise markets they desperately need to court.
Microsoft is in a more complex situation with AI agents. The popular open-source platform OpenClaw demonstrated a potential path forward for AI agents, and after OpenAI quickly hired its creator, Peter Steinberger, Microsoft (among other companies) is trying to catch up. One of its key strategies is making OpenClaw work well with Windows. At Build, Nadella said he was very committed to OpenClaw support, and Microsoft employees chatted with developers in the audience about how they were using it.
Steinberger himself made a surprise appearance to great audience reaction, taking the stage to boast about how OpenClaw had bolstered its security and earned user trust. “What I kept hearing was, ‘Peter, I love my Claw, but can I use it at work?’” Steinberger said. “You can totally run OpenClaw inside your company now, and we even made the harness itself a plug-in.” Steinberger said that whether someone trusts Copilot, Codex, or another company’s coding platform, users can now run OpenClaw on top of that via Windows.
But Microsoft is also promoting its own separate Copilot “super app” that integrates OpenClaw-esque agents. A super app is a major focal point for OpenAI right now — president Greg Brockman is leading development of one that will tie together ChatGPT, the Codex coding platform, and the Atlas web browser. Microsoft’s strategy is similar, bringing together a variety of existing Copilot AI assistants. Its agents, called “Autopilots,” are designed to act as a helpful user interface. Cassidy Williams, GitHub’s senior director of developer advocacy, called Copilot “your home base for development and operations on your computer,” demonstrating how multiple agents could perform tasks like app-building. (In an extra flourish, Williams demonstrated how she could approve or deny code changes by flashing her computer camera a thumbs-up or thumbs-down.)
Autopilots are designed specifically to appeal to business customers — Nadella called them “autonomous, long-running agents with full enterprise compliance.” The first one Microsoft will offer is “Scout,” billed as “your always-on personal agent,” but clients can build and personalize their own. The Autopilot agents should be able to look through an email inbox, join group chats in Teams, check a calendar, and send daily briefings, among other things. Accordingly, employees on stage at Build repeatedly emphasized Copilot’s security tools and guardrails — obviously aiming to calm enterprise clients who may have heard horror stories about tools like OpenClaw.
Suleyman made sure to emphasize, again and again, Microsoft’s “humanist superintelligence” as an “AI that prioritizes humanity first” — part of AI companies’ recent rebrand of AGI to make it sound less frightening in an era when people are pushing back against AI more than ever before.
Nvidia CEO Jensen Huang, another speaker known for working closely with OpenAI, appeared via video call to tout how Nvidia’s RTX Spark chip is fueling Microsoft’s AI agent goals. “I could be traveling and I’m on the phone and I can text my PC … and it would fire up the tools on the PC,” Huang said. “The idea that the PC evolved from a personal computer to a personal AI is just really exciting.”
Microsoft spent years betting on OpenAI, and in some ways, that’s left it behind in the AI race. But as OpenAI and other competitors turn to enterprise to finally make money, it’s got some obvious advantages. Microsoft already has a substantial client base and, compared with other AI companies, a reputation for safety and security. And like Google, it also has deep pockets, considerable computing resources, and a diversified revenue stream, meaning it can take big bets without a ton of risk.
Suleyman told The Verge, “There’s a lot of people who are either like chasing startup valuations or about to IPO, so we can operate with a little bit more humility and a little bit more long-term optimization.” He added, “We’ve got the money to be able to buy Anthropic [models] when we need to. We’ve got the optionality in Azure with 11,000 models, so people can use literally whatever they want whenever they want, but that buys us the time to do it right from the start.”
At the same time, there are a lot of unanswered questions here. Microsoft called out a lot of benchmark wins and advancements for its seven new models, but that doesn’t always translate to real-world adoption, and even a new model that pulls ahead for a week or two can quickly fall behind. AI super apps are a mostly yet-untested idea. And Microsoft is entering a crowded yet still largely underwhelming AI agent marketplace with a product that we haven’t seen in action. There’s still plenty of room for its promises to fall flat.
Hayden Field
Tom Warren
AI
Analysis
Microsoft
Microsoft Build
OpenAI
Report
Tech
More in: Microsoft Build 2026: All the news about Windows, AI, RTX Spark, and more