The promise of the “AI PC” felt revolutionary just months ago. Microsoft declared every Windows 11 machine would possess AI capabilities, yet the fervor surrounding Copilot+ PCs at this year’s tech showcase feels…off. The industry has finally met Microsoft’s initial specifications, particularly with a surge in Neural Processing Unit (NPU) power from Intel, but a crucial question lingers: why should anyone care?
A wave of Copilot+ PCs is dominating the conversation, but it feels like a pursuit of a strategy Microsoft itself seems to be subtly abandoning. Microsoft is now downplaying the importance of NPUs, and with few applications truly leveraging their power, the massive investment in these specialized processors feels increasingly misplaced. This is especially true now that the Windows AI Foundry prioritizes GPUs and CPUs for AI tasks – a shift from the original Copilot Runtime’s NPU reliance.
NPUs, once considered essential, now appear less vital to the future of AI on Windows, even as they become ubiquitous. Was Microsoft simply distracted as its hardware partners finally reached the finish line, only to change the race?
When Microsoft first unveiled Copilot+ PCs, a minimum NPU performance of 40 trillion operations per second (TOPS) was a non-negotiable requirement. This presented a significant challenge for Intel, whose processors typically offered only 13 TOPS, with Lunar Lake reaching 48 TOPS. 2024 was touted as the year of the AI PC, but even well into 2025, many reviewed laptops couldn’t meet these demanding specifications.
Conversations with PC manufacturers revealed a sense of pride as they showcased updated versions of last year’s models. “And it’s a Copilot+ PC,” they’d announce, finally achieving the required performance. The industry had, at last, caught up.
Intel’s Core Ultra Series 3 (Panther Lake) is the centerpiece of this year’s advancements, a direct response to their previous NPU shortcomings. Boasting a 50 TOPS NPU and improved multithreaded performance, it promises a substantial upgrade – though independent benchmarks will be crucial to confirm the extent of the improvement.
While Lunar Lake was Copilot+ capable, its limited multithreaded performance meant significant compromises were necessary. Achieving Copilot+ functionality often meant sacrificing overall performance in favor of low power consumption and extended battery life.
AMD’s Ryzen AI 400 series introduces an NPU capable of 60 TOPS, available in both laptop and desktop configurations. Though an improvement over the Ryzen AI 300 series’ 50 TOPS, the practical impact remains questionable. With limited application support, the 10 TOPS increase is unlikely to be noticeable for most users, despite appearing impressive on a specification sheet.
Qualcomm continues to lead in NPU performance, with its Snapdragon X2 Elite and X2 Plus platforms delivering 80 TOPS via the Hexagon NPU. As a launch partner for Copilot+ PCs, Qualcomm remains ahead of the curve. However, the fundamental argument for NPUs remains elusive, mirroring the situation during the initial Copilot+ launch.
These new processor platforms are now delivering capable NPUs across a wide range of laptops. Copilot+ PC-capable NPUs are becoming increasingly common as we move into 2026. But the critical question remains: will it actually matter?
In October, Microsoft announced a plan to extend AI capabilities to *all* Windows 11 PCs. According to Yusuf Mehdi, the goal was to learn from the initial Copilot+ experience and bring AI functionality to a broader audience. “The big thing was, let’s bring that AI capability to all Windows 11 PCs and make it really simple for anyone to try it.”
Now, heading into 2026, Microsoft’s enthusiasm for NPUs appears to have waned. This shift is further underscored by the introduction of the Windows AI Foundry, which allows developers to create AI applications that utilize GPUs, CPUs, *or* NPUs – effectively replacing the NPU-dependent Copilot Runtime.
Copilot+ PC features haven’t sparked a revolution. A recent conversation revealed a common sentiment: many users are still on Windows 11, unaware of newer versions or the subtle AI enhancements available on select machines. The benefits are limited to features like Windows Studio Effects, image generation, Windows Recall, and basic text actions.
Based on Microsoft’s current trajectory, NPUs are unlikely to become the defining feature of the Windows AI experience. A future update could potentially deliver Copilot+ features to a wider range of machines, leveraging the power of GPUs for tasks like image generation and text summarization. That’s the outcome many are hoping for.
Requiring NPUs for Copilot+ PC features was a misstep. Even a high-end gaming PC, equipped with a powerful discrete GPU, remains unable to run these features – a surprising oversight considering GPUs are already the preferred hardware for demanding AI tasks.
