Satya Nadella defends Copilot: the billions aren't wasted (or so he says)

Imagine spending tens of billions of dollars building giant digital temples, only to hear passersby whispering that nobody prays inside. That is roughly where Satya Nadella finds himself. While Microsoft pours concrete and installs GPUs in new data centers, unpleasant rumors have spread through the market: Copilot is supposedly just an expensive accessory gathering dust on the shelf at most corporate customers. Nadella decided it was time to pull out the reports and show the world that his AI brainchild is not merely alive but thriving in the daily routine of millions of people.

The issue isn't just analysts' curiosity. Investors are getting visibly nervous as they look at electricity bills and endless chip purchases from Nvidia. When you spend sums like these, the standard "everything is going according to plan" stops working. You have to prove that people click that icon in the corner of the screen for reasons other than accident or idle curiosity. Nadella says the number of paying GitHub Copilot subscribers grew 180% year over year. That is a serious figure, considering that programmers are arguably the most cynical audience around, one you can't buy with simple marketing slogans. If they pay, the tool probably does save them time.

Outside the cozy world of source code, however, the picture looks considerably more complicated. Microsoft is aggressively building AI into the Office suite, trying to turn Word and Excel into something more than tools for typing text and building tables. Nadella claims the number of Copilot for Microsoft 365 users doubled over the last quarter. It sounds impressive, but the devil, as always, is in the details. We still don't know how Microsoft defines an "active user" in its metrics. Does it mean someone who asked the AI to shorten one email a week, or someone who handed it the full job of keeping meeting minutes? So far, the company is keeping those details to itself.

The roots of this desperate fight for user attention reach back decades. Microsoft spent too long as the "catch-up" player in the mobile era and in internet search. Now the company is leading a technology race for the first time in a very long while, and the fear of losing that pace pushes it to act with extreme aggression.

Every new data center is a huge bet that generative AI will become a new operating system, one without which business simply grinds to a halt. If Nadella is right, we are watching the birth of a new ecosystem comparable in scale to the arrival of the internet. If not, it will be the most expensive monument to inflated expectations in tech history.

Critics often compare Copilot to the infamous Clippy from the early 2000s. But there is a fundamental difference: Clippy didn't need billions of dollars to sustain its virtual existence. Modern AI is an incredibly voracious beast that demands a constant flow of energy and computing power. Nadella insists this beast is already delivering real value to businesses by automating tedious routine. He cites large corporations that supposedly save thousands of working hours. But until that impact shows up directly in net profit in quarterly reports, Wall Street's skepticism isn't going anywhere.

It is interesting to watch how quickly senior leadership's rhetoric is changing. A year ago we were promised magic and the technological singularity; now we are shown dry user-retention charts. This may be a sign of the technology maturing, with pragmatism replacing enthusiasm. Microsoft has laid a foundation on which it is now obliged to build something tangible and profitable. Otherwise, not even Satya Nadella's endless calm and charm will save the company from uncomfortable questions about where shareholders' money actually went.

Bottom line: Microsoft is trying to convince the market that AI is no longer an experiment but an everyday work tool, but we will learn how dependent users really are on Copilot only when the company agrees to disclose daily activity data rather than general subscription figures.