Ming-Chi Kuo: Nvidia's N1X devices are expected to ship around 10 million units in the next two years
Claire Weston
Kuo's latest supply-chain survey pegs Nvidia N1X PC shipments at roughly 10 million units over two years, but he argues the real bottleneck for on-device AI adoption is the operating system, not the chip.
What does 10 million units actually mean?
Analyst Ming-Chi Kuo (郭明錤) of TF International Securities estimates Nvidia's N1X Windows PCs will ship roughly 10 million units over the next two years.
He calls this a niche market, aimed squarely at power users who need local AI compute.
In plain terms = this chip is not chasing the mainstream. It is built for people who must run large models on their own machine.
When do the first products appear?
The first N1X-powered PCs are expected to debut simultaneously at Taipei Computex and Microsoft Build in San Francisco.
Microsoft Surface and Dell are among the brands confirmed to launch devices.
This means → Nvidia is not going alone. It has Microsoft + major OEMs committed, and the ecosystem groundwork is already underway.
Can shipments be revised upward — and what is the bottleneck?
Kuo identifies two conditions: pricing, and whether Windows can actually put on-device AI compute to work.
His observation: across both Windows and Mac today, mainstream AI usage still runs through browser-based LLM access or cloud APIs. Nearly all compute depends on the cloud.
In plain terms = no matter how powerful the chip, if the OS has no compelling use case for local compute, users will not feel the difference.
What do two 2026 PC-market trends tell us?
MacBook shipment forecasts jumped from 5 million to 10 million units, but buyers are drawn to low prices, design, and ecosystem — not on-device AI.
Cheap mini PCs (such as OpenClaw and Mac mini) are gaining attention for running AI Agents around the clock, yet those Agents' inference compute still comes from the cloud.
This reflects a simple reality: the actual drivers of PC upgrades today remain price and user experience, not local AI power.
What is the real barrier for on-device AI?
Kuo argues that on-device AI's defining advantage over cloud AI is the ability to deeply integrate cross-app user data and workflows while preserving privacy.
That integration requires deep OS-level support. Today, PC operating systems are still at an early stage — mostly adding AI features inside first-party apps, with only shallow cross-app integration.
Put simply = the value of on-device AI is not "faster compute." It is connecting the data scattered across all your apps — and only the OS can do that. It has not done it yet.
What does this mean for ordinary users?
The N1X's biggest value: it gives AI power users a credible alternative to Mac, balancing AI compute, memory, design, and portability.
For users who need to run LLMs locally, the N1X is worth considering.
But to drive a mass upgrade cycle, beyond reasonable pricing, Windows OS support for AI remains the critical bottleneck.
Content is for reference only, not financial advice.