Microsoft Shipped Copilot to 450 Million Desks. Almost Nobody Is Paying for It

On March 17, 2026, Microsoft announced a significant reorganisation of its Copilot operation, unifying its consumer and commercial teams under a single leader, appointing a new EVP to run the product, and freeing its AI chief Mustafa Suleyman to focus on building the company's own frontier models.

The company framed it as a strategic evolution and a sharpening of focus. However, the numbers tell a completely different story.

The M365 Copilot ships to 450 million commercial seats. It is bundled into Word, Excel, Teams, and Outlook. In other words, Microsoft has handed its AI assistant the most enviable distribution advantage in the history of enterprise software. It sits on the desktops of nearly every corporate worker who uses a Microsoft product.

And yet, M365 Copilot has reached only 15 million paying users at up to $30/user/month. That is 3.3% penetration of its own installed base. 97 % of M365 users have looked at Copilot, shipped to them by default, and chosen not to pay for it.

For context, Satya Nadella himself confirmed the 15 million figure on Microsoft's January 2026 earnings call, presenting it as a milestone. Not everyone saw it that way. For a product with this level of built-in distribution, 3.3% is not a milestone.

“Microsoft’s disclosure of 15 million Microsoft 365 Copilot paid users represents disappointing uptake of the tool, just 3.3% of the 450 million-strong Microsoft 365 user base, despite reorganizing the Microsoft 365 product and go-to-market around Copilot,” said J.P. Gownder, vice president and principal analyst at Forrester.

Microsoft will point to consumer Copilot as evidence of momentum since daily active users nearly tripled year over year and conversations per user doubled. These are real numbers, but consumer Copilot is free.

Tripling the users of a free product does not close a $72 billion investment gap. The metric that matters is paid conversion. And on that measure, 3.3% is where the conversation ends.

A Dozen Copilots, Zero Clarity

The Bloomberg report from March 17 adds an important dimension that the official messaging overlooks. It notes that the reorganisation follows complaints, from inside and outside the company, that fielding an array of Copilot assistants was confusing.

At one point, Microsoft marketed more than a dozen Copilot variants like tools for software developers, security professionals, finance workers, consumers, and enterprise teams. Analysts who advise Microsoft customers and investors struggled to navigate the complexity.

This goes to the heart of why 97% of M365 users have not converted. When enterprise buyers cannot clearly articulate what they are purchasing, or whether they are being asked to pay multiple times for similar products, they do not purchase. They wait. And in the AI market, waiting means defaulting to a competitor.

Microsoft has spent the past year trying to simplify by reducing standalone products, folding features together, and, as of last week, launching a new higher-priced M365 E7 tier at $99 per month that bundles Copilot features.

The reorganisation is the latest move in that cleanup effort. It is a restructuring of a product that shipped too many things at once and confused the very buyers it was trying to serve.

The Capital Question and the Suleyman Signal

The scale of Microsoft's AI investment makes the adoption gap harder to ignore. The company spent $72 billion in capital expenditure over the past two quarters, in AI infrastructure spending. That is the foundation on which Copilot runs, the compute, the infrastructure, and the capacity to serve enterprise AI at scale.

15 million paying users is what that investment has produced on the Copilot side so far. The $72 billion also supports Azure's broader AI services and Microsoft's partnership with OpenAI, which powers everything from M365 Copilot to Bing.

It is not a Copilot-specific spend. But it does illuminate the scale of the bet Microsoft is making, and the distance between that bet and current commercial traction.

Perhaps the most telling element of the reorganization is what happened to Mustafa Suleyman. He joined Microsoft in 2024 when the company acquired the engineers and intellectual property behind his AI startup Inflection AI. He built out Microsoft AI as a distinct unit, drove AI integration into consumer-facing products like Bing and MSN, and positioned himself as the executive leading Microsoft's charge in the consumer AI race.

That race, Reuters notes, is one in which Microsoft watched rivals, including OpenAI and Google, amass a far larger individual chatbot user base. Suleyman and his team watched that happen. Now, responsibility for closing that gap falls to Jacob Andreou, who joined Microsoft last year from venture firm Greylock Partners, having previously spent eight years at Snap.

Suleyman is being repositioned to lead Microsoft's superintelligence and frontier model efforts, building the company's own AI models independent of OpenAI.

"We now have the resources that we need to train frontier-scale models over the next few years," he told Bloomberg. "That's fundamentally what my mission is going to be."

Whether that is a step up or a step aside depends on Microsoft's strategic priorities. What is clear is that the person who led consumer Copilot through its most critical adoption window is no longer leading consumer Copilot.

The OpenAI Dependency

There is a longer-term strategic dimension to all of this that the reorganization hints at without addressing directly. OpenAI, which powers most of Microsoft's AI offerings including M365 Copilot, now accounts for roughly 45% of Microsoft's remaining performance obligation.

That is an extraordinary level of dependence on a single technology partner, particularly one that competes directly with Microsoft's own products in the consumer market.

Suleyman's shift to model development is a hedge. If Microsoft can build its own frontier models, it reduces its reliance on OpenAI's technology and its exposure to that 45% concentration risk.

This way, the reorganization is about Microsoft preparing for a future in which the OpenAI partnership may not be what it is today.

Microsoft did not announce any product changes on March 17. A spokesperson confirmed that the data safeguards offered to corporate AI users would continue in future versions of Copilot.

“More consolidation is coming,” Suleyman said. The goal, in his words, is to give consumers and corporate customers 'the best of both worlds.'

The four-pillar structure that emerges from the reorganization is designed to move from a fragmented collection of products to an integrated system.

Microsoft has the distribution. It has the infrastructure. It has the enterprise relationships that most AI companies would trade anything for. What it has not yet demonstrated is the ability to convert any of that into a product that the people who already use its software actually want to pay for. That is the problem they are trying to solve.