Microsoft is losing its way in the AI race. Can Copilot get it back on track?
Original Title: Microsoft lost its way in the AI race. Can Copilot get it back on course?
Original Author: Jeremy Kahn
Translation: Peggy, BlockBeats
Editor's Note: Microsoft was once the first heavyweight to bet on OpenAI in the generative AI wave. With investments in OpenAI and an exclusive cloud partnership, Microsoft was seen as the most certain winner of the AI era: Azure benefited from models, Copilot was fully integrated into Office, Bing, GitHub, and enterprise software, and Nadella, like leading Microsoft's shift to the cloud, was expected to once again lead a platform-level migration.
However, two years later, Microsoft's advantage began to unravel. OpenAI was no longer just a technology supplier to Microsoft but also a direct competitor for enterprise customers; models like Claude and Gemini quickly caught up, diminishing the lead brought by GPT's exclusivity; the emergence of AI Agents further challenged Microsoft's long-standing SaaS business model. Stock price retracement, lower-than-expected Copilot adoption rates, GitHub Copilot being surpassed by Cursor and Claude Code, all forced Microsoft to reassess its AI strategy.
What is most noteworthy in this article is not whether Microsoft can still catch up with OpenAI, Anthropic, or Google in terms of model capabilities, but that Microsoft is attempting to redefine its position: it is no longer betting entirely on a single model but is shifting towards a "model-agnostic" enterprise AI platform strategy. In other words, Microsoft aims to become the foundational layer connecting models, data, security, workflows, cloud computing, and enterprise software. Models may come from OpenAI, Anthropic, or even Microsoft's own Superintelligence team in the future, but what truly stays within the Microsoft ecosystem are enterprise customers' work platforms, data assets, development environments, and security frameworks.
This is also the backdrop for Nadella's personal involvement in Copilot product development. For Microsoft, the AI competition is no longer just a model race between labs but a systemic competition about organizational speed, product form, customer relationships, and capital expenditure. Claude Code and Claude Cowork prove that AI Agents could reshape software development and office processes; projects like OpenClaw indicate that an "always-on" AI assistant is transitioning from concept to reality. What Microsoft needs to do is to package these more aggressive AI native experiences into secure, compliant, and governed frameworks that enterprise customers can embrace.
However, the cost of this path is not low. In order to catch up with cutting-edge models and support Agent-based products, Microsoft is pushing the AI competition towards "gigawatt-level" infrastructure investment: more data centers, larger chip clusters, and higher capital expenditure. By 2026, Microsoft expects capital expenditure to reach around $190 billion. In other words, Microsoft in the AI era needs to not only rapidly iterate like a startup but also continuously invest heavily in assets like a cloud computing giant.
The real challenge facing Microsoft is not whether it can still be the sole winner of the AI era, but whether it can continue to hold onto the core entry point of enterprise software in a situation where models are rapidly commercialized and Agents continually disrupt the software business model. For Nadella, this may not be a simple product adjustment, but more like Microsoft's second entrepreneurship in the AI platform migration.
The original text is as follows:
From the top clockwise are: Microsoft CEO Satya Nadella; OpenAI co-founder and CEO Sam Altman; Microsoft AI division CEO Mustafa Suleyman.
In mid-January 2026, Redmond, Washington State. The weather was cold and gloomy, the kind of morning most suitable for hitting the snooze button for a little longer. However, in Building 92 of Microsoft's vast campus, an engineering team had already arrived early.
They were fighting a tough battle, and they were already behind.
This team was developing a new AI product. It was more like a personal assistant that could help users book flights, reply to emails, and even find a reliable local plumber. Team members were well aware that other tech companies were also developing similar products. Just then, Microsoft CEO Satya Nadella arrived on the scene. He wanted to show them something.
Nadella opened his laptop and launched an application. It was a system for commanding and controlling multiple AI Agents, which he called the "Chain of Debate." While demonstrating, Nadella explained to the engineers. The team members exchanged a knowing look, as if seasoned basketball players suddenly realized that a new teammate was actually very good at playing.
Because this application wasn't something Nadella had others do for him; it was something he wrote himself using AI tool "vibe coding."
"This set the tone for how aggressively the entire team should drive the work forward." recalled Jacob Andreou, Microsoft's Executive Vice President in charge of Copilot design. At that time, Nadella stayed in the same room with everyone, almost standing behind the engineers, and participated with his own computer.
Seeing the CEO so excited about building the new product from scratch also energized the team. By late February, this sprint came to an end, and Microsoft launched Copilot Tasks—an AI tool with a personal assistant-like capability. The prototype Nadella had built earlier became a reference model for a feature in Copilot called the "model council" and other components.
However, Nadella's frequent deep dive into the AI product team and even hands-on prototype construction itself speaks volumes about Microsoft's current situation. After all, this is a $3 trillion tech giant, not a grassroots startup where the CEO frequently participates in sprint planning sessions and codes alongside developers.
Nadella's concern about Microsoft's AI strategy has been quite evident. Last October, he announced stepping back from some commercial responsibilities to focus more on AI research, product innovation, and AI datacenter development.
This concern is not unfounded. Microsoft's stock price went through a tough period earlier. After reaching an all-time high last October, the stock price dropped by about 34% over the following five months. Meanwhile, Microsoft's Azure cloud platform and AI-related revenue more than doubled over the past year.
Microsoft also became one of the typical victims of the so-called "SaaSpocalypse" (SaaS doomsday selling). The emergence of AI programming agents triggered a collective sell-off in software stocks. Many investors began to believe that such products signaled that enterprises would no longer buy AI products from SaaS vendors like Microsoft in the future and might not even purchase off-the-shelf software anymore.
Between October 28, 2025, and March 27, 2026, Microsoft's stock price cumulatively declined by 34%. The sales momentum of Microsoft's enterprise-grade Copilot product fell below the company's expectations. Out of the 450 million users of the Microsoft 365 suite, the current percentage of users paying for Copilot features is less than 4.5%. At the same time, consumer-oriented Copilot chatbot usage significantly lags behind ChatGPT, Gemini, and Claude. The once-leading AI programming assistant GitHub Copilot has been surpassed by the AI startups Cursor and Anthropic's Claude Code.
Two years ago, Microsoft seemed to be one of the early winners of the AI era. Thanks to Nadella's forward-looking bet on OpenAI, Microsoft gained exclusive access to the fast-growing AI startup's models and could integrate these models into its product ecosystem. If a business wanted to use OpenAI's technology, Microsoft Azure was the only cloud service provider they could choose. Microsoft even believed at one point that OpenAI gave it the best chance in years to challenge Google Search.
At that time, Nadella had been at the helm of Microsoft for a full decade. He had led Microsoft through the platform shift from desktop software to cloud computing, and now appeared poised to replicate this success in the AI era.
But AI evolves rapidly. In the span of two years, a long cycle could form. The next part of the story is how Microsoft missed out on its early AI lead and how it is now trying to regain the initiative.
Where It Went Wrong
Microsoft's initial surge to the forefront of the AI race was thanks in large part to its collaboration with OpenAI; however, what also partly led to its passive position was this very collaboration.
Microsoft had discovered this young San Francisco startup early on and made its first $1 billion investment in 2019, with subsequent committed investments in OpenAI totaling $13 billion. Leveraging OpenAI's technology, Microsoft rolled out a series of AI products under the Copilot brand across its consumer and enterprise software product lines.
However, by the end of 2022 with the release of ChatGPT, OpenAI's explosive growth and rapid expansion ambitions quickly strained the relationship. The two companies clashed on several fronts: OpenAI always wanted more in terms of computing resources; on intellectual property, Microsoft felt OpenAI was not promptly fulfilling contractual obligations to share technological innovations; on customer relations, OpenAI started directly selling its AI models to the same cohort of enterprise customers using Copilot from Microsoft; and as OpenAI sought a restructuring, disagreements arose over how much stake Microsoft should have in the new for-profit entity.
Nadella knew that betting Microsoft's AI strategy on a still largely unproven startup came with risks. In November 2023, these risks were starkly laid out: the nonprofit board overseeing OpenAI's for-profit business, citing a "failure to consistently uphold candor," ousted CEO Sam Altman and informed Nadella of the decision just minutes before publicly announcing it.
Nadella had to swiftly reassure investors, emphasizing that Microsoft still retained access to OpenAI's technology; simultaneously, he worked with Altman to pressure the board to reverse the decision. Nadella announced that Microsoft was prepared to hire Altman and any OpenAI employees willing to join him at Microsoft. The possibility of a mass employee exodus ultimately forced the board to relent and reinstate Altman.
Within OpenAI, this five-day-long crisis was later referred to as "the blip." However, according to those familiar with Nadella's thinking, the incident shook him deeply. He had to find a hedge for Microsoft's AI bet.
“When Nadella joined the company's AI engineering team for a sprint, it set the tone for how aggressively the entire team would drive the work forward.”
—Jacob Andreou, Microsoft Copilot Executive Vice President
Microsoft's Plan B was Mustafa Suleyman.
Suleyman, a co-founder of Google DeepMind who later left to start his own AI startup, Inflection. In March 2024, Microsoft made a $650 million deal to hire Suleyman and Inflection's tech team, gaining access to their technology. Suleyman was then appointed CEO of Microsoft's new AI division, referred to as MAI, with a dual mandate: to build cutting-edge models internally at Microsoft as a hedge against OpenAI, and to expand the user base of Microsoft's Copilot chatbot.
However, the progress didn't go smoothly. Due to Microsoft's partnership agreement with OpenAI, Microsoft was restricted from training models beyond a certain scale. Suleyman told Fortune, “At that time, we could mostly only train Microsoft's own native models, and could only achieve SLM, which is small language model scale.”
MAI's first publicly tested general-purpose language model, named MAI-1 preview, was launched in August 2025 but ranked quite low on various performance leaderboards and was ultimately not widely released.
MAI also failed to turn Copilot into a consumer hit product. According to reports, a year into Suleyman's tenure, Copilot stagnated at around 20 million weekly active users, while ChatGPT's user base soared to 900 million. In 2025, Microsoft attempted a major upgrade to Copilot, aiming to make it more like a task-performing personal assistant, but this upgrade did not reignite growth. As for the AI-powered version of Bing search, it barely made a dent in Google's market share in search.
Meanwhile, Plan A also started to run into trouble.
In 2023, OpenAI's GPT model was far ahead in the industry. However, by early 2025, Anthropic's Claude had frequently topped AI rankings, and many companies preferred to use it for complex tasks. Google's Gemini also became increasingly competitive in visual tasks. Microsoft's Copilot product was still entirely GPT-driven. The engine that once underpinned Microsoft's AI strategy was now starting to feel like a heavy burden.
Microsoft's Chief Commercial Officer, Judson Althoff, admitted that the company did make a few missteps. Firstly, naming both consumer and enterprise products Copilot was inherently confusing. Althoff, who holds a private pilot's license, joked, "The only thing worse than not having a copilot is having more than one copilot."
Microsoft also incentivized sales representatives to promote both the free value-added version and the premium version of the enterprise-grade M365 Copilot simultaneously, but it was only the premium version that truly delivered value to enterprise customers. "We got this one wrong," he said.
Microsoft has also been striving to keep up with the pace of AI evolution. A significant inflection point occurred in 2025 when Anthropic released the Claude Code. Developers only need to describe what they want, and it autonomously writes the complete program. This was no longer just "copiloting" but "autopiloting." Within a mere six months, it reshaped the way software is developed.
Subsequently, in January of this year, Anthropic introduced Claude Cowork. This is an Agent software that can interact with Microsoft productivity tools such as Excel, PowerPoint, and others to autonomously complete tasks.
Claude Cowork poses a significant challenge to M365 Copilot and the AI Agents Microsoft has been encouraging customers to adopt. In fact, it threatens not only Microsoft but most commercial software. It was this recognition that sparked the so-called "SaaSpocalypse" software stock sell-off. Ultimately, over $20 trillion in tech market value evaporated, including a single-day loss of $357 billion in Microsoft's market capitalization.
Microsoft's Correction Course
By the fall of 2025, Nadella realized that Microsoft needed to reboot its AI strategy. Subsequently, the company's actions reflected a delicate balance: on the one hand, it had to innovate rapidly like an AI startup; on the other hand, it still had to serve investors and enterprise customers reliably, akin to the traditional steady Microsoft.
Nadella delegated many of the commercial and day-to-day operational responsibilities to Microsoft's senior executive, Althoff, so that he could focus on AI product development. Althoff mentioned that he was in charge of "Horizon Zero" and "Horizon One," while Nadella was responsible for "Horizon Two" and "Horizon Three." At the same time, Nadella began breaking down internal barriers to make Microsoft faster, flatter, and more agile.
In March of this year, Nadella merged the Consumer and Enterprise Copilot teams. Suleiman no longer oversees Consumer AI products and now leads a rebranded model research project: the Superintelligence team. Suleiman stated that the name reflects the team's ambition and helps attract top researchers.
Jacob Andreou joined Microsoft in 2025, having previously worked at Snap and the venture capital firm Greylock. He is now responsible for the Consumer and Enterprise ends of the Copilot Experience, reporting directly to Nadella. Joining Suleiman and Andreou in the Copilot leadership team are three Microsoft senior vice presidents: Charles Lamanna, in charge of Copilot, AI Agent, and platform; Ryan Roslansky, responsible for Microsoft Office and Microsoft's LinkedIn; Perry Clarke, serving as Chief Technology Officer for application systems.
Lamanna said, "We want it to be a back end, a brain, driving both consumer and work scenarios." Nadella himself will participate in the Copilot leadership team's weekly stand-ups and engage in an ongoing Teams channel specifically discussing Copilot's development progress.
Microsoft is facing a delicate balance: it needs to innovate fast enough to keep up with AI rivals like Anthropic and Google, while also continuing to be a reliable partner in the eyes of large enterprise customers.
Andreou pointed out that two new products can demonstrate the unified Copilot team is operating as Nadella envisioned: one is the consumer-facing Copilot Tasks, a product that Nadella personally participated in designing the prototype for in January of this year; the other is Copilot Cowork for enterprise customers.
He said, "Both of these products are essentially cutting-edge experiences, one for consumers and one for enterprise users. And both have been quickly built by our team by integrating resources in just a few weeks."
Microsoft has also agreed to the long-pending restructuring plan of OpenAI, with significantly reduced restrictive terms. The software giant now holds a 27% stake in OpenAI. If OpenAI goes public as widely expected, this could provide Microsoft with potential upside. However, the exclusivity arrangements in the old agreement have been abandoned: OpenAI can now partner with other cloud service providers, and Microsoft can also use models from other AI companies.
Suleiman stated that the new agreement finally allows Microsoft to build larger-scale, more powerful cutting-edge AI models and ultimately achieve self-sufficiency. However, he also added that Microsoft will need another two to three years to catch up with top AI labs.
The reshaped partnership has also enabled Microsoft to embrace OpenAI's main competitor, Anthropic. In November last year, Microsoft pledged to invest up to $5 billion in Anthropic and started offering its models on Azure. The ability to use Claude to power Copilot has been well-received among enterprise customers and has helped Microsoft create Copilot Cowork.
“It must be acknowledged that OpenAI and Anthropic are helping us move faster.” — Judson Althoff, Microsoft's Commercial CEO
But Microsoft is not simply replacing its reliance on one loss-making AI startup with another. Behind the investment in Anthropic lies another judgment from Microsoft: AI models will become increasingly commoditized. At least in the enterprise market, the real value will not only be in the AI “brain” but will shift to the tools, data, security, cloud computing, and workflow systems that revolve around the brain.
This is where Microsoft believes it can excel.
Microsoft already has many key assets: software tools, security systems, data warehouses, and cloud computing capabilities. Microsoft has also created a series of products branded with IQ to help enterprises create customized workflows, aggregate their own data, and build, deploy, and monitor Agents that run these workflows based on any vendor's AI model.
Althoff stated: “We don't believe that enterprises will switch their information work platforms, development environments, and security environments every time a new model is released.”
This strategic shift has also brought about a new business model.
In the past, Microsoft typically charged based on user licenses, for example, $30 per user per month for Copilot. Customers liked this model because budgeting was easier. But if the AI Agents in these products use models that Microsoft does not own, Microsoft must pay the corresponding token consumption costs to the AI vendor.
Therefore, Microsoft has begun to transition to a hybrid pricing model: the base part is still charged per user license and includes a limited token quota; any excess is charged per token. This is done to prevent the erosion of profit margins from the “model-agnostic” strategy.
Due to cost-cutting efforts, Microsoft has also begun streamlining its workforce. In April of this year, Microsoft announced the launch of its first-ever voluntary employee departure compensation plan, mainly targeting its most tenured employees. The company stated that about 7% of its U.S. employees, approximately 8,750 employees, are eligible for this plan, with an expected cost of $900 million.
Signs indicate that Microsoft's adjusted corporate strategy is paying off. As of the end of March, Azure revenue grew by 40% year-over-year, and Microsoft's overall AI business achieved an annualized revenue of $37 billion, a 123% year-over-year growth. Currently, 20 million M365 users are paid subscribers for Copilot, with a quarter of them being new additions in the first four months of 2026. Althoff stated that adoption is accelerating.
UBS analyst Karl Keirstead mentioned that more and more Microsoft customers are telling him that they are starting to see the value of Copilot. However, overall user penetration is still not satisfactory. He said, "I don't think they've hit the penetration rate that would satisfy Wall Street."
Microsoft's "model-agnostic" strategy may also have a loophole: What if those highly watched AI startups start building Microsoft-style enterprise tools and connectivity systems?
This is no longer a hypothetical situation. In February of this year, OpenAI launched the enterprise-focused Frontier platform, offering many capabilities that Microsoft is building into its new tools. Anthropic is also moving in this direction, introducing the Claude Managed Agents service.
Microsoft's argument is that decades of enterprise customer relationships, reputation for reliability and security, and deep integration with customers' existing software systems will give it an advantage. Althoff mentioned that he welcomes competition. "One has to recognize that OpenAI and Anthropic are helping us move faster," he said.
However, some question whether a company of Microsoft's scale can truly keep up with the agility of AI-native startups. UBS's Keirstead stated, "Microsoft, as frankly all software companies, is facing a situation they haven't faced in a decade-plus: highly innovative brand-new competitors. The expectation that a large established company like Microsoft could pivot as quickly as OpenAI and Anthropic might be asking too much."
Bank of America analyst Tal Liani takes the "Nadella camp" side. He believes that AI companies are unlikely to build out the full product suite that Microsoft offers. This means that Microsoft doesn't necessarily have to win the AI race; it just needs not to lose it.
He said, "It may not be the best, but as long as it's good enough and provides high value through bundling, this is actually where Microsoft's value lies."
However, even just "not losing" comes at a high cost.
Like other hyperscale cloud providers, Microsoft is investing heavily in data centers and custom chips. In the 2025 fiscal year, Microsoft's capital expenditure reached $88.2 billion, roughly on par with peers such as Google Cloud and Amazon AWS. However, in hindsight, this was still too conservative. Surging demand has left Microsoft struggling with a shortage of computing power and unable to recognize contracted AI revenue as actual revenue at the expected pace.
"I thought we would catch up," CFO Amy Hood admitted during last October's earnings call, "but we didn't."
Now, Microsoft is further ramping up its efforts. The company projects that capital expenditure in 2026 could reach around $190 billion, more than three times the spending in 2024. Wall Street has previously been nervous about such spending levels, but now seems willing to tolerate such massive investments. However, if investor sentiment were to reverse, Microsoft would be more exposed to risk than ever before.
In November 2025, an independent developer named Peter Steinberger released OpenClaw. This is a free, open-source system that can turn any AI model into a long-running, always-on Agent: it can develop software, act as a virtual executive assistant, and even manage inventory for an online store.
OpenClaw has been widely embraced by developers and AI enthusiasts. Reportedly, Nadella is also among them.
But despite its popularity, OpenClaw has one glaring issue: to truly be effective, it needs access to systems, data, payment information, and passwords, making it highly risky. At the same time, it consumes tokens at a staggering rate.
In March of this year, Nadella stated at a tech conference in San Francisco, "I can't launch OpenClaw at Microsoft. I don't have the authority to do so, as it would be seen as Microsoft releasing a virus. But at the same time, it is indeed a remarkable innovation."
Nadella has tasked the consolidated Copilot team to build the Microsoft version of OpenClaw: maintaining the fun and ease of use of a consumer-grade product while possessing the security and governance capabilities required by enterprises. Andreou sees this as a test for the new organization: "This is what we call a victory here."
Lamanna believes this could be the key to igniting Copilot's growth. He said, "The hardest question has always been: How do you help people change the way they work?"
If a perpetually running AI assistant is truly viable, it will make this change easier to happen. It also means that the basic unit of AI will shift from a 'model' to a 'always-on Agent.' This is precisely a paradigm shift that will test whether Microsoft's so-called 'Connected Organization' strategy can continue to stand at its core when transformation happens. Lamanna stated that an enterprise-ready Microsoft version of OpenClaw is not far off.
“Gigawatt-Scale”
During the week of March 30, Suleiman convened the new Superintelligence team in Miami for a three-day offsite meeting. This team, consisting of about 500 people from around the world, aimed to chart a roadmap for achieving 'gigawatt-scale' AI training operations. This scale of training will enable Microsoft to directly compete with OpenAI, Anthropic, Google DeepMind, Meta, and xAI.
Suleiman stated that it is crucial for Microsoft to achieve self-sufficiency by 2030. Microsoft will lose access to OpenAI's technology by 2032.
The entire team gathered in a massive banquet hall to listen to Suleiman and Nadella deliver keynote speeches and participate in an 'Ask Me Anything' session. According to Suleiman's recollection, Nadella described this moment as a 'reset of the company' that Microsoft was undergoing to address the AI platform shift.
It was a profound statement.
After the keynote speeches, the meeting split into different workflows. Each team gathered around one of the 40 whiteboards placed around the banquet hall to brainstorm and plan the upcoming eight-week sprint tasks. Nadella didn't leave but stayed behind.
Over the next three hours, he moved between tables, speaking to researchers, offering suggestions, and sharing ideas.
If this is indeed a 'reset,' Nadella is playing the role of a startup CEO. He doesn't take any advantage for granted. He knows that Microsoft could lose everything and still have everything to fight for.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
