There is something almost poetic about watching Microsoft get accused of missing the AI wave just as it begins to pull back on the loudest parts of its AI push. Former Microsoft executive Mat Velloso has been making the rounds with a blunt assessment of the company’s strategy, arguing that Redmond “missed the internet wave, the mobile wave, and now it missed the AI wave”. Coming from someone who spent more than a decade inside the company, including four years as Satya Nadella’s technical advisor, the critique lands with a sting.
Velloso’s argument is straightforward. Microsoft bet big on Bing as its flagship AI showcase and saw no meaningful shift in search share. It pushed Copilot into Windows and Office, yet he says “not even 3 percent of paying users use it even when it is pre-deployed right in their faces”. He also points to the company’s push for NPUs in new PCs, only for OEMs to discover that “nobody cares because not a single valuable use case was built for those in Windows or Office”. In his view, the company overbuilt, oversold, and under-delivered.
It is a harsh diagnosis, and it resonates because the symptoms are visible. Microsoft has spent the past few months quietly scaling back Copilot’s presence in Windows, reducing the number of places where the assistant appears and re-centering the OS around fundamentals. That shift has been framed as a response to user feedback, but it also reflects a deeper truth. Consumer AI enthusiasm has cooled, and the data shows it. Even with massive distribution, Copilot’s paid adoption rate sits at roughly 3.3 percent of Microsoft 365 users who interact with it, according to reporting in your tabs.
Yet here is where Velloso’s critique and Microsoft’s reality start to diverge. He frames the company as missing the wave, but the company may simply be stepping off the surfboard before everyone else realizes the tide is shifting. Consumer AI is loud, visible and easy to market, but it has not proven that users will pay for it. Enterprise AI, on the other hand, is where the actual money is flowing. Microsoft already has more than 20 million enterprise Copilot customers, a number Nadella highlighted earlier this year, and that base continues to grow even as consumer sentiment softens.
This is the part of the story that gets lost in the noise. Microsoft is not abandoning AI. It is abandoning the idea that AI needs to be plastered across every consumer surface to be successful. The company is pulling back on branding, reducing forced integrations and focusing on the places where AI actually drives revenue. That is not missing the wave. That is recognizing that the wave is not where everyone thought it was.
Younger users, meanwhile, are increasingly skeptical of AI. Not because they dislike the technology, but because they dislike the way it is being forced into every corner of their digital lives. The backlash is real, and it is growing. Companies that are still treating AI as a universal branding exercise may soon discover what Microsoft already has. Consumer enthusiasm is not the same as consumer willingness to pay, and the gap between the two is widening.
Meanwhile, other companies are still sprinting in the opposite direction. They are doubling down on AI branding, AI features and AI marketing, convinced that the consumer appetite is bottomless. It feels familiar because Microsoft has been here before. When the company became one of the first major players to start charging for cloud storage, it was dragged for it. The internet treated it like a greedy outlier. Then the bills came due for everyone else. Within a few years, every major cloud storage provider quietly adopted the same model because the economics of “infinite free storage” were never real.
That same pattern is starting to repeat with AI. Microsoft’s move toward tokenization is not beloved by consumers, but it is a far more realistic way to account for usage. It is built for developers, not for vibes, and it reflects the true cost of running these models at scale. The data center investments behind AI are massive, and someone has to pay for them. It will not be long before Google, OpenAI, Anthropic and every other LLM provider follows the same path, not because they want to, but because they will have no choice.
Eventually, these companies will face the same reckoning. They will need to pull back, refocus and justify their AI investments with actual ROI instead of vibes. Microsoft is simply getting there first. Again.
So while Velloso’s criticisms highlight real missteps, they also miss the broader strategic pivot underway. Microsoft is not failing to catch the AI wave. It is choosing the part of the ocean where the water is deep, the customers are sticky and the revenue is recurring. The consumer AI hype cycle may be flattening, but the enterprise AI engine is only getting started. And for a company built on long-term infrastructure bets, that is the wave that actually matters.
