Copilot Plus PC — How Microsoft Invented a Category Nobody Needed

Microsoft spent two and a half years building a category around Copilot Plus PCs — machines with enough on-device AI processing to run tasks locally without cloud help — only to let the branding quietly evaporate as even its own hardware partners stopped using it. The failure wasn't about consumer skepticism toward AI itself. It was about a mismatch so fundamental it exposes how companies mistake specifications for strategies.

Intel made the identical move with Ultrabooks starting in 2011. The category was simple: thin, fast Intel processors in premium chassis. Intel backed it heavily. OEMs launched dozens of them. By 2015, the brand was dead. Not because thinness was unpopular. It was because thickness had never been the problem users were paying to solve.

Both invented hardware categories around a single differentiating spec before solving a real customer problem. Intel marketed thinness as the feature; buyers wanted portability, which they already had. Microsoft marketed on-device AI as the feature. What buyers actually needed. If they needed anything. Was to solve a specific task cheaper, faster, or better than they could before. Running Copilot offline instead of in the cloud solves for latency. Nobody asked for latency to be solved. The feature was technically interesting. The use case was missing.

When Qualcomm and Microsoft stopped mentioning Copilot Plus explicitly, they weren't abandoning AI integration. They were abandoning the fiction that a hardware category could live on a processor architecture alone. The new Surface Pro and Surface Laptop still run the same AI features. They're just not branded as special anymore because branding them as special required the premise that on-device processing power mattered to how people worked. It doesn't, not yet, and possibly not ever if the tasks don't change.

Both invented hardware categories around a single differentiating spec before solving a real customer problem.

The lesson isn't that AI adoption failed or that consumers resist new features. It's that hardware categories built from the inside out. Starting with what's possible. Collapse against categories built from outside in, from what people are actually trying to accomplish. If you're designing anything with multiple components competing to define its purpose, you're working backward. The spec is never the strategy.