You bought into the hype and purchased a first-gen AI laptop with an Intel Core Ultra (Meteor Lake) chip inside. Congratulations! But in a world that shifted overnight to Copilot+ PCs with much more stringent requirements, were you a fool for doing so?
It’s a tough question, and one that’s already hit home for us, as well. But Intel executives say that’s not necessarily the case, and there are plenty of reasons to own an older AI PC — which includes the possibility that AI PCs will be able to run Copilot+ functions in Windows over time.
Here’s the conundrum: A Core Ultra Series 1 chip (Meteor Lake) produces about 11 TOPS from the NPU, and laptops bearing those chips rolled out the first half of this year. But mere weeks ago Microsoft launched Copilot+ PCs, and placed a restriction upon them: if the NPU doesn’t deliver 40 TOPS, they’ll be unable to access new Windows 11 apps like Recall, plus generative AI in Paint and Photos. Intel’s latest Core Ultra chips, the “Lunar Lake” generation debuting at Computex, meets this threshold with 45 TOPS. Meteor Lake doesn’t. Gulp.
So that begs the question: What does Intel have to tell those early adopters that bought into the AI PC? I asked Michelle Johnston Holthaus, the executive vice president and general manager of the Client Computing Group at Intel, if Meteor Lake buyers wasted their money on an AI PC.
Intel’s Michelle Johnston Holthaus revealing Lunar Lake to press.
Intel’s Michelle Johnston Holthaus revealing Lunar Lake to press.
Mark Hachman / IDG
Intel’s Michelle Johnston Holthaus revealing Lunar Lake to press.
Mark Hachman / IDG
Mark Hachman / IDG
“Well, there’s a lot of assumptions in that question about the advantages of Copilot+,” Johnston replied. “There’s a lot of things that people want to do with AI in Meteor Lake that they can absolutely take advantage of. I think we would argue that in a few years based upon how the software will improve that Meteor Lake will be able to run something like the Copilot+ as those language models become a little more efficient.”
Johnston was referring to the ongoing improvements in the various middleware components: APIs, runtimes, and the language models themselves. Like software drivers, they’re tuned over time, making them more efficient, and theoretically allowing them to run on older, inefficient hardware. But she also unexpectedly threw some cold water on the need for an AI-powered PC.
“You know we haven’t seen Copilot+ PCs en masse yet, and I think we need to take time and understand what it is that users want to take advantage of in that operating system, which things are most important,” Johnston added, pointing out that the cloud was an option, too. “Obviously, we’ve been working with Microsoft to ensure that Lunar Lake meets all of the specifications for Copilot+… So it’s maybe a little overstated to say that everybody needs a Copilot+ PC.
“I think we need to look at how each individual user is going to use it, what they are going to do with AI, and then we really need to define those products for users,” Johnston said. “There are a lot of people out there today buying PCs who aren’t going to use AI. Right? That’s why we say you have to first have to have a great modern CPU.”
“There will be something for everyone,” Johnston said.
Further reading: Your ‘AI PC’ is already obsolete: The curse of early adoption strikes again
