How will Intel’s AI PC help you work smarter?

Intel has released a new series of Core Ultra processors with dedicated neural processing units (NPUs), which it claims heralds the dawn of the AI PC.

The bulk of today’s AI tasks are processed by huge servers in the cloud. It’s one of the reasons why Nvidia’s earnings and share price have rocketed in the past 18 months: its GPUs are the hardware behind services such as ChatGPT.

Intel wants to attack that on two fronts – with AI server chips of its own (which we’ll come to shortly) and with the Core Ultra processors that are capable of performing AI tasks locally, instead of in the cloud.

Let’s be clear, however: you’re not about to be running services such as ChatGPT from your local PC. That still requires phenomenal levels of processing power that only the cloud can provide. Instead, Intel will look to tackle more low-key jobs on its NPUs.

Related reading: Lenovo to mark MWC 2024 with more AI PCs  

Work smarter?

So, how will Intel’s AI PCs help you work smarter? One way is to give you a chance to stand out on work video calls.

For example, the company said the NPUs are already being used to blur backgrounds on Zoom calls. Intel’s AI PC Acceleration Program includes partners such as Krisp, a company that specialises in reducing background noise on voice calls on services such as Discord. Other partners perform tasks such as video compression.

In other words, the NPUs will likely be used for very specific tasks that loosely fall under the “AI” umbrella, rather than the chatbots and generative art services that have grabbed much of the public’s attention over the past year or so.

The NPU isn’t doing all of the heavy lifting, either. The AI workloads will be spread across CPU, GPU and the NPU to fully exploit the available processing power. Intel executive vice president, Michelle Johnston Holthaus, claims that by 2028 these so-called AI PCs will comprise 80% of the PC market, and that “together with our vast ecosystem of hardware and software partners, Intel is best positioned to deliver this next generation of computing”.

AMD may have something to say about that. Last week, it unveiled a new generation of Ryzen 8040 processors, which also include AI capabilities.

Intel’s AI server offensive

Intel’s doing its best to loosen Nvidia’s stranglehold on the AI server market, too. The company will next year launch its Gaudi3 chips, designed as a direct competitor to Nvidia’s H100, which the graphics firm can barely make fast enough to keep pace with demand.

Again, Intel is behind here: AMD announced its third-generation Instinct MI300 accelerators at an event in California last week. And a Dell VP stepped on stage to say that it was taking orders right now.

Intel CEO Pat Gelsinger suggested the AI industry was tiring of its reliance on Nvidia’s CUDA infrastructure and was looking for alternatives. “We think of the CUDA moat as shallow and small, because the industry is motivated to bring a broader set of technologies for broad training, innovation, data science, et cetera,” Gelsinger said, according to a report on Tom’s Hardware.

However, Intel doesn’t have a great record of disrupting markets where it’s surrendered an early lead. For example, the company made several attempts to smash ARM’s grip on the smartphone and tablet markets, without success. The company also failed to gain a foothold in the console market. The first generation Xbox did use an Intel processor, but it has since made way for AMD and Nvidia parts.