Microsoft has dedicated much of 2023 and the rest of 2024 to this extend Copilot’s reach and functionality, its suite of AI-powered tools, features and services. Last February was the first anniversary of the introduction of the new Bing, and although at the time there were indications that these technologies would gain a significant presence in the Redmond catalog, it was the subsequent announcements and the consolidation of what was shown in them. under the heading of Copilot, which made it clear to us that in addition to surprises, AI is also waiting for us in the soup.

We were not wrong to think so, although it is true that the deployment of some announced proposals, such as the integration of Copilot in Windows 11 and Windows 10 It’s going much slower than expected.. Because? Of course, a key factor that many of us have in mind from the first moment is the complex legal fit of AI-based solutions into the regulatory frameworks that are emerging around the world precisely in response to the emergence of AI that we live.

Now, there are other reasons that must undoubtedly be influencing this “slow” rate of deployment as well, and among them there is no doubt to be found costs that Microsoft must bear for offering Copilot as a serviceand we must not forget that currently absolutely all tasks performed by AI models in Redmond are performed remotely, on a huge server infrastructure.

I mentioned it earlier when I talked about the Procyon AI Image Generation Benchmark, and it bears reminding again now: 2024 will be the year of AI computing on the PC, and now we have one more confirmation of that, because reading in Tom’s Hardware, Intel and Microsoft are working to bring Copilot running on PCs. To do this, it is necessary to establish the technical characteristics that the systems must meet, and this is exactly where the two large technology companies meet.

In this sense, the arrival of the first computers with integrated Intel Core Ultra marked an important point, but there is still considerable development to come in the future generation. Because? Because the NPU of integrated devices of this generation is able to reach 10 TOP, but what was proposed by Intel and Microsoft to bring Copilot to PC means for this integrated part, computing capacity of 40 TOPs.

This requirement is determined by Microsoft, which wants Copilot to run on PC to support only NPU, instead of also using it in the GPU and CPU, as they believe that these components should be freed up for other functions. Currently, the peak of performance in this sense is AMD with the Ryzen 8000 series, which can scale up to 16 TOP, but Qualcomm announced that its expected Snapdragon TOP.