WHAT WILL AI PCs DO IN THE FUTURE?
There are benefits to owning an AI PC now, but those benefits range somewhere between niche and slight. However, there’s almost universal industry agreement that those benefits will rapidly increase well within the lifetime of the average PC.
“What you’re doing on an AI PC in 2024 is going to be radically different to 2026,” said Intel’s executive vice president for client computing, Michelle Johnston Holthaus, at CES in January, a sentiment with which the company’s chief rival agreed.
“It will probably be the end of 2025 before you really start to see what an
AI PC is capable of,” said AMD’s senior processor technical marketing manager, Donny Woligroski.
“The GPU is currently the fastest AI processing unit,” Woligroski added. “You’re going to see that performance cross over, with the NPU doing the same levels of [AI] performance as the GPU.”
Lenovo agrees the best is yet to come from AI PCs, pointing out that the software developers have only just started working with production hardware. “It’s almost a chicken and the egg,” said Tom Butler, the company’s executive director of commercial portfolio and product management. “I’ve got to have hardware out there that can catch the software benefits.”
He added: “The front end of this wave is hardware driven, because there was nothing for the software to write to, because a lot of this was in the labs of the silicon providers and OEMs like Lenovo. We couldn’t go out broadly and have these discussions with software companies.”
Now, according to Butler, “we’re actively in conversations with a broad breadth of software companies. We are, the silicon providers are, basically everybody’s looking at this new space to see ‘what can I do now?’”
The pace of product development in the AI industry has been nothing short of breakneck. Subscribe to the
Ben’s Bites newsletter ( bensbites.
It will probably be the end of 2025 before you really start to see what an AI PC is capable of
beehiiv.com), for example, and you’ll find multiple new AI product launches announced every day.
Butler predicts you won’t have to wait too long to see exciting new product launches that tap the capabilities of AI PCs, either. “I think if you just project forward near term – like in months, quarters, not years – you’re going to see a lot more capability coming,” he said.
Dell’s Kevin Terwilliger agrees that we’re only just getting started with AI PC apps. “We’re just at the tip of the spear when it comes to all these AI capabilities,” he said. “Intel has talked about the 100+ ISVs [software developers] that are utilising the Core Ultra processor. And then what we also see as a great example, companies like Dell, large organisations, are developing new AI capabilities to roll out to their end users internally. That’s going to take advantage of this NPU to be able to run efficiently and not have to move a bunch of data into the cloud.”
Lenovo isn’t planning to leave it to third-party developers to come up with AI software, either. At CES it showed off prototypes of its own AI apps that took advantage of the
NPU, instead of relying on the cloud. The company’s AI Now platform includes a Windows Copilot rival that is able to alter operating system functions, as well as perform tasks such as summarising long documents
and writing emails, all without data being sucked into the cloud.
Currently, it only plans to release AI Now in China, but it’s indicative of a coming shift from cloud to local AI processing. “If you think about AI in general right now, most of it is cloud-based,” said Butler. “And that opens up latency concerns, security and privacy concerns. And so the ability to bring that down to the edge device not only overcomes some of those hurdles, but it also makes it more personal, like you’re working on your work for you, not a broad, open public cloud-based platform.”
A big advantage of local AI software is that it could access the different apps and data piles stored on your PC, rather than being restricted to files that you upload to the cloud. Butler said you can think of this as an “orchestration layer”, capable of bringing together the capabilities of different apps, working with system-wide data. “It’s almost at that point a prompt-led conversation with your system, not ‘let me open this app to accomplish a task’.”
He envisages a future where you’ll tell the AI that you want to accomplish a specific task and it will tell you the best, most efficient path to achieve that using the resources and apps you have available on your particular PC. “That’s not in the market or present today, but that’s effectively what you want to drive to from an AI perspective,” he said.