x86 strikes back: Intel comes out swinging as AI PC market explodes (2024)

To set the scene, a couple weeks ago Microsoft announced the new category of “Copilot+ PC” that would be able to comfortably run generative AI models locally on the machine, rather than relying on the cloud. And Qualcomm was the star of the show, with its Arm-based Snapdragon X Elite chipsets powering the first wave of announced Copilot+ laptops. These processors promise great power and power efficiency that should make Windows machines competitive with Apple’s recent Arm-based MacBooks, which have owned the market for portable, powerful, and long-lasting laptops in recent years.

So, game over for the x86 architecture that has traditionally powered PCs? Not if Intel has anything to say about it.

Microsoft’s longtime Windows partner came out swinging today, with CEO Pat Gelsinger using his Computex appearance to stake Intel’s claim to at least part of our AI-imbued future. “Intel is one of the only companies in the world innovating across the full spectrum of the AI market opportunity—from semiconductor manufacturing to PC, network, edge, and data center systems,” Gelsinger said, reflecting Intel’s current racing-to-stay-in-the-race reality (the company has been blindsided by Nvidia’s rise in the AI data center.)

Gelsinger promoted Intel’s upcoming Gaudi 3 AI accelerator chips, which he said would be a cheaper alternative to Nvidia’s market-leading H100s. But the really big Intel announcement was that of its Lunar Lake laptop architecture, which will make its way into PCs later this year.

Lunar Lake is Intel’s second crack at an AI PC platform. Last year’s Meteor Lake was Intel’s first stab at incorporating an AI-focused neural processing unit (NPU) à la Apple. But if you were an Intel AI PC early adopter, I regret to inform you that Meteor Lake-based laptops are now officially weak sauce, after Microsoft said last month that all Copilot+ PCs will have to be able to manage at least 40 trillion operations per second (TOPS) of NPU performance.

Meteor Lake can only eke out 11.5 TOPS at its peak, but the Lunar Lake NPU can manage 48 TOPS—and if you look at combined NPU and CPU and GPU performance, the top of the line Lunar Lake PC will be able to achieve 120 TOPS. Lunar Lake also promises 60% more battery life than Meteor Lake. “It’s x86 power like you’ve never seen it before,” Intel technical marketer Rob Hallock told The Verge, boasting that Lunar Lake would “definitely” beat Qualcomm’s offerings.

The urgency of Intel’s desire to stay competitive in the AI PC race isn’t just demonstrated by the speed at which it’s released a significantly redesigned successor to Meteor Lake—it’s also clear from the fact that Intel has turned to Taiwan’s TSMC to make the key compute and controller components of its Lunar Lake chipset.

This is a very big shift for a company that has always made its own processors, and it’s not exactly a ringing endorsem*nt of Intel’s Foundry chipmaking operations, which it recently separated from the design operation that produced Lunar Lake. But in Gelsinger’s words: “Lunar Lake picked TSMC as the right technology at that point in time.” The next generation, Panther Lake, will almost all be made in Intel’s own factories, but right now keeping up means swallowing pride—this is how Intel gets to promise the release of 80 Intel-based Copilot+ PCs this year, from 20 manufacturers.

Meanwhile, AMD’s Lisa Su also used Computex to announce x86-based Copilot+ PC chips—the Ryzen AI 300 Series, with an NPU offering 50 TOPS. So expect to see loads of PCs with those inside too, from the likes of Lenovo, Acer, Asus, and HP. Heck, if you’re not too bothered about achieving the sort of performance offered by an official Copilot+ PC, chipmaker Hailo has even just announced a $70, 13 TOPS-capable AI accelerator add-on for the cheap-and-cheerful Raspberry Pi microcomputer. It remains to be seen to what degree everyone actually uses the AI capabilities being build into their computers right now, but this market just exploded.

More news below. And by the way, if you want to hear Qualcomm CEO Cristiano Amon’s thoughts on the AI PC sector, he’ll be one of the speakers at Fortune Brainstorm Tech next month in Park City, Utah—details here. Oh, and check out the latest Fortune 500 list, which just dropped.

David Meyer

Want to send thoughts or suggestions toData Sheet?Drop a linehere.

NEWSWORTHY

X > Tesla. Elon Musk recently told Tesla investors that the company would add 50,000 top-end Nvidia AI chips to its infrastructure this year, but surprise! CNBC reports that Musk (who is currently trying to convince Tesla shareholders to give him $45 billion) has diverted 12,000 of those Nvidia chips to his X social network, which has strong ties to Musk’s xAI operation. An equivalent X order of Nvidia H100s will later be redirected to Tesla, but the switcheroo will likely delay the car company’s efforts to develop truly self-driving cars and humanoid robots.

Cruise gathers speed. GM’s Cruise has cautiously put its robo-taxis back on the road in a second city—Dallas this time, after Phoenix in April. As TechCrunch reports, this is again not about ferrying passengers, but rather about sending the vehicles around the city with human drivers, to gather road data. Supervised autonomous driving will follow at some point. Cruise had to pause its entire operations in October, after a San Francisco pedestrian was dragged under one of its vehicles.

Microsoft cloud cuts. Business Insider reports that Microsoft is cutting hundreds—perhaps as many as 1,500—jobs in its Azure cloud business. These cuts affect the Azure for Operators team, which is part of Microsoft’s moonshot organization; the company is also cutting back its mixed reality team. And in other Microsoft news, the European privacy NGO Noyb (founded by the activist lawyer and Facebook antagonist Max Schrems) has filed data-protection complaints in Austria about the Microsoft 365 Education suite of productivity programs for students. As Reuters reports, Noyb claims Microsoft is illegally shifting the legal responsibility for handling students’ data to their schools, which don’t actually hold the information.

ON OUR FEED

“When you look at the frontier technology, at the most cutting edge, that needs to be in co-ordination with the U.S. players and there needs to be reassurances that are given to the U.S.”

—Omar Sultan Al Olama, the United Arab Emirates’ AI minister, says there’s a “marriage” between the UAE and the U.S. in the domain of AI. The Gulf state is trying to pitch itself as a global AI hub, and a recent deal between Microsoft and local champion G42 saw G42 promise to avoid Chinese hardware to retain access to American chips.

IN CASE YOU MISSED IT

Spotify increases subscriber prices for second time in a year—the latest service to jump on the price-hike bandwagon, by Chris Morris

Spotify’s CEO got roasted by artists after he said the cost of creating content is ‘close to zero.’ Now he’s trying to walk back his ‘clumsy’ remark, by Marco Quiroz-Gutierrez

Airbnb CEO reflects on fumbled messaging during layoffs: ‘You don’t fire members of your family’, by Chloe Berger

After being dragged down for months by accounting probe, Autodesk reassigns CFO and sees its stock surge, by Bloomberg

25-year-old Anthropic employee says she may only have 3 years left to work because AI will replace her, by Orianna Rosa Royle

By sunsetting Section 230, Congress could be about to break the internet as we know it, by Ed Brzytwa (Commentary)

BEFORE YOU GO

Amazon’s Project P.I. Amazon is combining computer vision with an AI model called “Project P.I.” to scan products for defects before they go out to the customer. The tech also checks that the color and size of the product conform with what the customer ordered. As The Verge notes, Project P.I. is already doing its thing at “several” North American fulfillment centers, and Amazon is touting the positive environmental impact of cutting down on returns.