A computer with the capacity to run artificial intelligence locally is a very important goal for Intel, and Lunar Lake, a platform that aims to be particularly efficient, will play a key role in this near future. As you may recall, we recently learned about the technical requirements set by Intel and Microsoft for a local launch of Copilot, and as we go into depth here, those are pretty reasonable goals, but they keep us waiting for the next generation of embedded devices. because the current does not reach them.

A particularly important point, as and for the reasons explained by my colleague Isidro, is the computing capacity of the integrated NPU, a requirement that, in summary, corresponds to two reasons, to guarantee energy efficiency and to ensure the stability of the system when using the Local Second Pilot. So the minimum set by both technologies is that this integrated part is able to reach 40 TOP, a capacity that we can already get with Meteor Lake, but only with the sum of the work of several integrated components including the CPU.

Lunar Lake, Intel’s next generation of integrated laptop chips, aims to meet this goal, surpassing Ryzen 8000’s TOP 10 NPUs and competing with the promised TOP 45 NPUs (it is expected that it may climb to TOP 75 overall) before expected Snapdragon Elitethe expected big leap from Qualcomm into the PC ecosystem, which, after the successful precedent of Apple Silicon, will also represent an important advance in terms of extending the ARM architecture into it.

Now, while it’s important to be able to rely solely on the NPU for tasks like local execution of Copilot, there are undoubtedly other uses where we’ll want to be able to use the full arsenal of integrated, and that seems like a big tech company to have in this sense also promising plans. Specifically, as we read in MSPoweruser, Pat Gelsinger, CEO of Intel, stated that Lunar Lake will be able to reach the TOP 100a milestone that would push it ahead of Qualcomm’s proposal, and we’ll have to wait and see how it stacks up against AMD’s future generation.

There are many reasons why this makes sense in the world. transfer some of the tasks based on artificial intelligence to the local level, such as permanent availability (since we will not be dependent on the Internet connection and the availability of services) and of course privacy, since the data we use will no longer be uploaded to servers and will not be subject to licenses that may allow them to be used for multiple purposes. So something we’ve long predicted is becoming increasingly apparent, namely that if 2023 was the year of AI services, 2024 (and even more so 2025) will be the year of AI done locally.