Microsoft opens part of Windows 11’s local AI to PCs with GeForce RTX 30 and leaves Copilot+ without one of its clearest barriers

Microsoft has taken an important step in its AI strategy for Windows 11 by confirming that local language model APIs They will also be able to work on PCs other than Copilot+ PCs, as long as they have an NVIDIA GeForce RTX 30 or higher GPU and at least 6 GB of VRAM. The change expands the scope of some of the system’s local AI and reduces one of the clearest differences that Microsoft had so far drawn between Copilot+ computers and the rest of the Windows ecosystem.

When Microsoft launched the Copilot+ PC concept in 2024, it presented it as the gateway to the new Windows with local AI. To enter that category, among other requirements, an NPU capable of reaching 40 TOPS was necessary, in addition to 16 GB of RAM and SSD. The idea was that these machines would be the ones prepared to execute AI functions directly on the device, with more energy efficiency and less dependence on the cloud.

Now the company qualifies part of that border. In updated documentation on GitHub, Microsoft indicates that developers can now use Language Model API on GPU on devices that do not belong to the Copilot+ category, as long as they meet that graphic requirement. In practice, this means that a portion of Windows’ local AI capabilities are no longer tied exclusively to the presence of an NPU.

What changes and what functions will these GPUs be able to take advantage of?

The opening does not affect the entire Windows 11 AI feature block for now, but rather a specific API aimed at local language models. Microsoft explains that these functions are supported by Phi Silicaa small model prepared to run locally and accessible from applications that call the system’s APIs. If a supported app needs it, Windows Update can download that model to your computer for use directly from the supported GPU.

Among the functions that an application could take advantage of through these APIs are text summarization, rewriting, text to table conversionintelligent formatting and general prompt generation. That is, it opens the door for Windows applications to integrate functions similar to a conversational assistant or assisted editing, but executed locally on PCs with compatible NVIDIA graphics.

This does not yet imply that all the most visible functions of Copilot+ marketing are extended to any computer with a GPU. According to the information available, Recall, Click to Do or Paint AI They are still not planned at the moment for PCs without NPU. For now, the expansion affects the language API layer, not the full set of exclusive features that Microsoft tied to the Copilot+ PC.

Copilot+ loses exclusivity, although it does not disappear

Until now, Microsoft had linked much of the story of local AI in Windows to Copilot+ PCs as a distinct category. However, opening these APIs to GeForce RTX 30 or higher GPUs implicitly acknowledges something technically obvious. And it is that one Powerful GPU can also run local AI models with solvency, although it does not follow the efficiency logic of an NPU. In fact, it is the powerful GPUs that power the heavier models, not the NPUs.

For developers, the novelty expands the installed base on which they can build applications with local AI in Windows 11. For users, it does not yet mean an immediate avalanche of new functions, but it does point to a scenario where part of the system’s native AI will no longer be reserved for a very specific category of recent laptops. Copilot+ does not lose all its meaning, but as a brand it begins to dilute.