Can Microsoft's AI Offerings Elevate Anyone to Stardom?

Written by | Landong Business Zhao Weiwei

Who among America's AI titans holds the least sway? The likely answer is Mustafa Suleyman, CEO of Microsoft AI.

Take social media platform X as an example—Sam Altman of OpenAI undoubtedly commands the largest following, with over 5 million fans. He leverages X as a free marketing platform for ChatGPT and OpenAI, with posts amassing tens of millions of views. A single post like "Codex has surpassed 1 million active users" can spark tens of thousands of shares and responses.

Google's Demis Hassabis comes in second, with over a million followers. Yet, he maintains a low profile on social media, posting infrequently. His last high-profile post was the 2023 release of Gemini 1.0. People follow and anticipate his updates primarily due to his authoritative stance in AI academia and DeepMind's advancements.

Dario Amodei of Anthropic is equally understated. He doesn't even use X as a social media outlet, opting instead to post a few updates directing readers to his official blog. Despite his modest social media presence of just 375,000 followers, this hasn't hindered him from leading Anthropic to become the world's most valuable AI startup.

In contrast, fourth-ranked Mustafa Suleyman reflects Microsoft's current standing in the AI landscape.

Suleyman is the most active on social media platform X, yet the reach and impact of his content remain notably average. Two years ago, as one of DeepMind's co-founders, Suleyman joined Microsoft as CEO of Microsoft AI, assuming full responsibility for Microsoft's consumer-facing AI products and reporting directly to Microsoft CEO Satya Nadella.

In April of this year, Microsoft and OpenAI formally transitioned to a non-exclusive partnership after their split. Forty days prior to this move, Suleyman stepped back from his comprehensive leadership role at Microsoft AI to focus on forming the internal MAI superintelligence team, dedicating more energy to independent cutting-edge model development.

Suleyman can be viewed as a key catalyst for Microsoft's "de-OpenAI-ization" and, under Nadella's influence, the future torchbearer of Microsoft's AI strategy.

"Our goal is to establish ourselves among the world's top four. The truly significant labs are three: Google DeepMind, OpenAI, and Anthropic. We're not one of them yet, and that's always been my aim," Suleyman remarked after the recent Microsoft Build conference.

To ascend to the fourth position in the industry, a corresponding flagbearer is necessary.

To be a flagbearer in the AI industry, one must possess a strong foundational belief, be shaped by product and technological achievements, and prove oneself over time and through refinement. For Suleyman, the deeper question is whether Microsoft's AI products are sufficient to elevate anyone to stardom.

At Microsoft Build 2026, Microsoft unveiled a suite of products, including super apps, internal reasoning models, cybersecurity tools, and AI assistants akin to OpenClaw. Notably, in terms of models, it introduced a flagship reasoning model, MAI-Thinking-1, along with six new models focused on image, speech, transcription, and coding.

MAI-Thinking-1 is Microsoft's first fully independently developed reasoning model, boasting 3.5 billion active parameters, an MoE architecture, a 256K context window, and trained from scratch without any data distillation from OpenAI.

A month earlier, Microsoft and OpenAI parted ways. Microsoft relinquished its exclusive license to OpenAI's intellectual property and terminated its revenue-sharing obligations to OpenAI. In return, it secured IP usage rights extending until 2032 and OpenAI's continued commitment to the Azure platform.

Now, Microsoft must stand independently. As head of Microsoft AI, Suleyman emphasized when introducing Microsoft's independently developed reasoning model that it was "trained from scratch, designed for rigorous mathematical operations, coding, and practical enterprise deployment."

In its product introduction, MAI-Thinking-1 ranks among the top in its class: it performs on par with leading models in key software engineering benchmark tests and achieves human preference levels comparable to Sonnet 4.6 in blind comparison evaluations.

Even though Microsoft underscores the model's advantages in performance and cost, it still trails behind peers like OpenAI and Anthropic. Interestingly, Suleyman acknowledges this gap, publicly stating, "We closed a significant gap within six months."

The most notable aspect of the Microsoft Build conference was the sprint toward the enterprise market dominated by Anthropic.

Scout, an online office intelligent agent, is a smart assistant built on the open-source AI platform OpenClaw. It can be integrated with Microsoft's suite of office tools to assist businesses in managing schedules, processing expense reports, writing emails, and more.

Suleyman publicly informed the media that Microsoft is more focused on the enterprise developer market championed by Anthropic and less concerned with the more consumer-oriented routes taken by Google, Meta, and OpenAI.

The label he has crafted for the Microsoft AI brand is "humanistic superintelligence," meaning people-centric. This aims to alleviate external resistance and fear toward Microsoft's AI strategy amid current public sentiment resistant to AI. Under this label, the three application areas Microsoft AI values most in the long term are AI assistants, medical superintelligence, and abundant clean energy.

However, the current competition is more pressing. Microsoft's flagship model is still in beta testing, available to select enterprises on Microsoft Foundry, with a long road ahead before large-scale commercial deployment. This is the fundamental reason Microsoft AI has yet to generate a compelling narrative.

During Suleyman's two-year tenure at Microsoft AI, the weekly active users of its conversational product, Copilot, have long hovered around 20 million, essentially stagnant.

This is not solely his responsibility. In 2024, when Suleyman joined Microsoft, the company brought him and the core team from Inflection AI on board with a $650 million package. Suleyman, in charge of Microsoft AI, took over Copilot, Bing, and Edge's full range of consumer products.

However, a constraint has always been present: Microsoft's binding agreement with OpenAI explicitly prohibited Microsoft from training models exceeding a specific computational threshold, effectively blocking Microsoft's ability to develop its own large models. Later, Suleyman told Fortune magazine that this was a significant limitation, especially for a company of Microsoft's scale.

A year later, Suleyman announced a major upgrade to Copilot: memory function, personalized companion positioning, real-time screen analysis, shopping features, and more, enabling Copilot to complete tasks like booking tickets on behalf of users.

In reality, these features were technologies already released by ChatGPT and Google a year earlier. Microsoft Copilot was merely catching up to what its competitors had already achieved. These homogeneous updates naturally failed to support a narrative of user growth and struggled to impress users.

Six months later, Google Gemini 3.0 and ChatGPT were locked in a fierce competition, forcing OpenAI to raise a red alert.

But Suleyman sparked a public backlash with his social media comments, which drew significant criticism and opposition. He said, "Wow, so many cynics! Every time I hear someone say AI isn't impressive, I find it amusing."

"I grew up playing Snake on a Nokia phone! Now we can have fluid conversations with superintelligent AI and generate any image/video. The fact that some people aren't impressed by this baffles me."

Many users opposed Suleyman's remarks not because they disapproved of AI technology but because Microsoft was integrating immature AI conversational technology into computer systems and office suites. People objected to issues like the inability to opt out, privacy concerns, and data collection.

At the time, some media outlets argued that this exposed a severe disconnect between Microsoft's product philosophy and actual user demands. It revealed Suleyman's deeper dilemma at Microsoft: he believed in the products, but they hadn't yet reached the level he trusted.

During this period, Suleyman had to worry about the business because, two weeks earlier, he had become the head of Microsoft's MAI superintelligence team, proposing the "vision of humanistic superintelligence" in an attempt to find a breakthrough for the stagnant consumer product line.

Looking back, that marked the beginning of Suleyman's role transformation. He was no longer just a strategist for consumer products and set the stage for Microsoft and OpenAI's separation.

In March of this year, Suleyman's authority narrowed. He focused on superintelligence and cutting-edge model research and development, aiming to deliver world-class models for Microsoft within five years. Meanwhile, the consumer and enterprise Copilot teams merged under former Snap executive Jacob Andreou. A month later, Microsoft and OpenAI officially announced their separation, ending the exclusive licensing agreement.

Data released by Sensor Tower shows that in February of this year, the Microsoft Copilot app had 6 million daily active users—a testament to Suleyman's tenure. In comparison, OpenAI's ChatGPT had 440 million daily active users, while Google's Gemini had 82 million.

Microsoft AI currently lacks a true flagbearer, but this is not solely Suleyman's issue.

Microsoft's AI narrative has been deeply intertwined with OpenAI, with OpenAI providing the models and Microsoft serving as the distribution channel. Now, their decoupling has just begun, and Microsoft needs sufficient product and technological achievements to establish its own AI narrative rhythm.

The deeper question is whether Microsoft AI needs a true flagbearer—or whether Microsoft's AI products are sufficient to elevate anyone to stardom?

The answer may be harsh. In the current AI industry chain, Microsoft Cloud is unparalleled—it is the most profitable among cloud providers. In 2025, Microsoft Intelligent Cloud (including Azure) generated $50.5 billion in profit with a 42% profit margin, both metrics surpassing rivals Amazon AWS and Google Cloud. Its revenue grew 34% year-over-year, with approximately 12% of that growth contributed by AI-related businesses.

Cloud providers connect upstream storage chip companies with downstream large model vendors and countless enterprise users. Microsoft has established a highly mature commercialization mechanism: enterprise software licensing, Office 365 subscriptions, Azure infrastructure, and GitHub toolchains. This mature product network allows Microsoft AI to effortlessly penetrate every corner of the internet.

But do users choose Microsoft's AI products because Suleyman excels?

No. The developer penetration rate of GitHub Copilot is credited to Microsoft CTO Kevin Scott, who oversees long-term underlying technology strategy and the developer ecosystem. Enterprise adoption of AI large model services falls under Scott Guthrie's purview—he leads Microsoft's cloud business, currently the most profitable and stable segment of Microsoft AI. Jay Parikh, in charge of the developer platform and agent factory, also reports directly to Nadella.

Nominally, Suleyman is Microsoft AI's leader, but the most profitable and strategically significant areas are controlled by other peer-level executives, making it difficult for him to become Microsoft AI's flagbearer at present.

What about the future? Suleyman must fight with his back against the wall.

Without reliance on OpenAI, the long-term competitiveness of all Microsoft AI products and agents hinges on its own underlying model development. Whether models can achieve high performance, low cost, and controllable deployment directly determines the long-term stability of Microsoft's upper-layer AI products and underlying cloud infrastructure. This is why Suleyman stated after his authority adjustment: "The model is the product."

Will Suleyman become Microsoft AI's flagbearer? The consumer business has already disproven this proposition. Progress at the model layer is now more critical than ever. The only remaining opportunity is for the MAI models he leads to secure a top-four position globally within the next five years, standing alongside Google, Anthropic, and OpenAI.