Remember a couple of years ago when Intel declared that the “age of the AI PC” had arrived?
Back at CES 2024, the chip giant was saying that its Core Ultra processors would usher in a new era of personal computing, enabling all kinds of new on-device AI capabilities. As Michelle Johnston Holthaus, then the company’s CEO of products, said in a keynote presentation, AI is “fundamentally transforming, reshaping, and reimagining the PC experience.”
Two years later, there’s been a vibe shift.
While Intel is still talking about AI, it now believes its PC processors will play more of a supporting role for cloud-based AI tools. At the CES trade show earlier this month, the company put a bigger emphasis on meat-and-potatoes concerns such as performance and battery life.
“With all the excitement around AI, we always remind ourselves, fundamentals still matter,” Jim Johnson, head of Intel’s Client Computing Group, said at a CES launch event.
A ‘disconnect’ with consumers
David Feng, VP/GM for Intel’s PC client segments, says in an interview that the change in emphasis was intentional. For all the talk about AI PCs, consumers haven’t been all that interested.
“There’s this disconnect between people in the industry who are looking a couple generations or a couple years ahead, versus the general public,” Feng says.
He jokes that for a while, Intel had a hard time getting through meetings without explaining its AI strategy, but when it asked retailers if customers were seeking out AI PCs, the answer was typically “no.”
“I’ll sort of confess in a way, and say, when we first coined the term AI PC, in hindsight we probably spent a little bit too much energy trying to justify running AI on the PC locally,” Feng says.
Unsurprisingly, what consumers want instead are basic PC things like better battery life and improved graphics performance. Intel’s partners are realizing the same thing, with one unnamed Dell executive telling PCWorld that it’s shifting its marketing focus away from AI PCs and “getting back to our roots with a renewed focus on consumer and gaming.”
While Microsoft remains all in on the AI PC concept, it too has started downplaying the value of on-device AI in favor of the cloud, declaring that all Windows 11 computers are AI PCs now.
Meanwhile, Intel began its shift toward more fundamental concerns with its Core Ultra 200V processors, which were an attempt to compete on power efficiency with Apple’s M-Series processors and new PC chips from Qualcomm. Now, Intel is promising further improvements with its Core Ultra Series 3 chips, which uses a new manufacturing process and started shipping in laptops this month.
The new chips have double the number of low-power computing cores, which are optimized for basic tasks such as web browsing and document editing, and those cores are more performant than before. Intel now plans to move all of its processors over to this architecture, including those for desktop PCs and gaming laptops.
“It’s a big leap,” Feng says.
Moving to ‘hybrid’ AI
None of this means that Intel has stopped talking about AI PCs entirely. But instead of emphasizing AI tools that run on-device, Intel is now touting “hybrid” applications, in which the AI primarily runs in the cloud but offloads certain tasks to the PC.
“We’re just more mature about thinking about this,” Feng says. “We’re not going to replace ChatGPT or Perplexity, and nobody’s asking us to replace them. The whole premise of a hybrid is, instead of choosing either or, how about you make them work together?”
For example, ByteDance’s CapCut video editor can now use on-device AI for its “AI Clipper” feature, which analyzes videos for potential highlights. This helps reduce the strain on ByteDance’s cloud servers.
Intel also teased a potential partnership with Perplexity, with Perplexity CEO Aravind Srinivas speaking at the chipmaker’s CES event. While Srinivas didn’t announce anything specific, he talked about how on-device large language models could preserve privacy, reduce latency, and cut cloud computing costs. A browser like Perplexity’s Comet, for instance, might use on-device large language models to provide insights on users’ browser histories, but turn to the cloud for web-based queries.
“Performance, security, economics, control—these make local compute such an obvious thing to work on,” Srinivas said.
Still, it’s early days even for these efforts, so why all the early hype about AI PCs a couple of years ago? Feng says Intel was just signaling that it was the start of a new era. Now, it can thankfully turn its attention to more near-term concerns that PC buyers actually care about.
“Right now, we’re just saying, look, the future is AI PC, but we don’t have to keep beating the drum the same way we beat it two years ago,” he says.