- HEADLINE & INTRO
Microsoft’s Copilot Numbers Look Big — Until You Do the Math
Microsoft wants Wall Street to believe that its colossal AI bet is already paying off. On the latest earnings call, Satya Nadella’s main message was simple: people are using Copilot a lot. Usage is up, data centers are full, and the AI flywheel is supposedly spinning. But once you look past the growth percentages and into the actual adoption patterns, a more nuanced story appears. In this piece we’ll unpack what TechCrunch reported, what Nadella didn’t say, and what this AI spending spree really means for customers, competitors and regulators — especially in Europe.
- THE NEWS IN BRIEF
According to TechCrunch’s coverage of Microsoft’s latest quarterly earnings, the company reported revenue of $81.3 billion, up 17% year over year, and net income of $38.3 billion, up 21%. Cloud revenue crossed $50 billion for the first time. At the same time, Microsoft’s capital expenditures are exploding: it spent $88.2 billion on capex last fiscal year and has already spent $72.4 billion in just the first half of the current one, largely to expand AI data center capacity for Azure, OpenAI, Anthropic and enterprise workloads.
Despite those strong headline numbers, the stock fell as some investors worried that Azure and Microsoft 365 growth came in slightly below expectations. On the earnings call, CEO Satya Nadella emphasised that demand for AI is strong and claimed consumer Copilot usage has nearly tripled year over year. He gave more concrete figures for GitHub Copilot (4.7 million paid subscribers) and Microsoft 365 Copilot (15 million paid seats), as well as rapid uptake of Dragon Copilot in healthcare.
- WHY THIS MATTERS
Strip away the buzzwords and this is a story about a company trying to convince investors it isn’t overbuilding the most expensive computing infrastructure in history.
On one side, the bear case is straightforward: Microsoft is pouring tens of billions of dollars into AI data centers while the two engines that actually monetise that spend — Azure and Microsoft 365 — just under-shot some expectations. Meanwhile, Copilot usage metrics for consumers are presented as relative growth (“nearly 3x”) instead of hard numbers. That’s classic PR when you don’t want the absolute figure scrutinised.
On the other side, there are pockets of very real traction. GitHub Copilot with 4.7 million paying developers is no small business; that suggests clear value in day‑to‑day coding. Dragon Copilot’s 21 million documented patient encounters in a single quarter show how compelling AI can be when it directly removes drudgery from specialised workflows.
The uncomfortable fact for Microsoft is that the crown-jewel product, Microsoft 365 Copilot, is still early. Fifteen million paid seats sounds huge until you remember Microsoft has around 450 million paid Microsoft 365 seats. That’s low‑single‑digit penetration so far. For now, Copilot is an add‑on for the motivated, not yet a universal standard.
Investors are effectively asking: is this AI cycle the next cloud (expensive but ultimately wildly profitable), or the next 3G spectrum bubble (massive capex with uneven returns)? Nadella’s answer is, unsurprisingly, that demand will fully utilise the new data centers. The numbers he shared show progress — but not yet proof.
- THE BIGGER PICTURE
This earnings call fits into a much broader pattern: hyperscalers trying to frame the AI boom as inevitable infrastructure, not speculative excess.
Over the last year, Google has pushed Gemini into Workspace, Search and Android, similarly pointing to growing usage while staying vague on monetisation per user. AWS, slightly behind on the visible “assistant” layer, is talking up Amazon Q and Bedrock as the foundation for customer‑built copilots. All three are telling the same story: AI workloads are filling data centers faster than they can be built.
Historically, we’ve seen this movie before. In the early cloud era, Wall Street worried that AWS and Azure would cannibalise existing server and software revenue before becoming profitable in their own right. In telecoms, operators spent heavily on 3G and fibre long before consumer behaviour caught up. Sometimes the bet paid off; sometimes it took a decade.
What’s different now is the pace and opacity. Training and running large models consumes huge amounts of capital, but the customer‑visible product — a chatbot inside Office, a code assistant in VS Code — feels deceptively simple. That mismatch makes it hard for CIOs and CFOs to decide what “fair value” for Copilot actually is.
Compared to its rivals, Microsoft does have an advantage: it owns the productivity layer where white‑collar workers already live. If it can turn Copilot from a curiosity into a default part of Word, Outlook, Teams and Windows, it can amortise AI capex across a massive installed base. But Google is trying a similar play inside Workspace, and open‑source models are rapidly improving, giving enterprises the option to run their own specialised copilots without locking themselves into one hyperscaler.
- THE EUROPEAN / REGIONAL ANGLE
For Europe, this isn’t just a question of usage numbers; it’s a question of control, compliance and bargaining power.
Copilot’s integration into productivity suites sits squarely in the crosshairs of the EU’s regulatory stack: GDPR for data, the Digital Services Act and Digital Markets Act for platform behaviour, and the upcoming AI Act for high‑risk AI systems. Every time Copilot touches personal data, emails, contracts or patient records, European compliance teams have to ask where that data is processed, how it is stored, and whether it is used to further train models.
Microsoft has already responded with EU‑focused measures like the “data boundary” for European cloud customers and promises about training restrictions. But the more aggressively Copilot is woven into Windows and Microsoft 365, the more scrutiny it will attract from data protection authorities in Germany, France and others.
There’s also a strategic question for European enterprises and public administrations: double‑down on US hyperscalers’ copilots, or invest in European alternatives and open‑source tooling? Players like OVHcloud, Deutsche Telekom, Scaleway and regional AI startups are trying to position themselves as sovereign options, often building on open models rather than closed systems like OpenAI’s.
For European SMEs — from a Ljubljana SaaS startup to a mid‑sized manufacturer in Bavaria — Copilot’s value proposition has to clear a higher bar: subscription prices must be justified in terms of concrete productivity gains, in markets where margins and IT budgets are tighter than in Silicon Valley.
- LOOKING AHEAD
Several things are worth watching over the next 12–24 months.
First, renewal cycles. Many Microsoft 365 Copilot deployments are still in pilot phases or tied to broader enterprise licence agreements. The real test will come when CIOs decide whether to expand, maintain or cut back seats after the first contract period. If Microsoft starts talking more about Copilot renewal and expansion rates, not just seats sold, that will be a sign of confidence.
Second, transparency. Nadella’s reliance on relative metrics for consumer Copilot suggests the absolute numbers may not yet impress. Pressure from investors is likely to push Microsoft to disclose more granular AI revenue, perhaps splitting out Copilot‑related ARR from generic Azure consumption.
Third, competition on price and architecture. As open models improve and on‑device AI accelerators appear in PCs and smartphones, some workloads will shift away from the hyperscaler data centers that Microsoft is frantically building. If enterprises can run a capable local model for routine tasks and reserve cloud AI for the hardest problems, the utilisation assumptions behind today’s capex may need revisiting.
For European organisations, the prudent path is experimentation with clear KPIs: pilot Copilot in specific departments, measure changes in output, error rates and employee satisfaction, and compare that against alternative tools, including local AI assistants and process automation. The risk is not only overpaying for hype, but also locking business logic into a proprietary assistant that may be hard to switch away from later.
- THE BOTTOM LINE
Microsoft’s message is that Copilot usage is booming and its AI data centers will be fully used for years. The numbers behind that story show real momentum, especially among developers and in verticals like healthcare — but also reveal how early we still are in mainstream workplace adoption. For European customers, the decision is no longer whether to touch AI at all, but how much of their future workflows to entrust to one vendor’s black‑box assistant. The key question: are you funding Microsoft’s bet, or leveraging it on your own terms?
