There’s something strange going on at Intel. The company is looking to get leaner as it simultaneously builds up its chipmaking capabilities. The U.S. chip giant’s nascent venture into graphics cards did not go unappreciated by the PC-buying community, especially the budget-end Battlemage GPUs like the B850. Amid the hubbub of its big Panther Lake announcement, one thing seemed to be missing: a clear idea of its future plans for GPUs, with or without Nvidia’s aid.
(Full disclosure: Intel invited me to its chipset fab in Phoenix, Ariz. Travel and lodging were paid by Intel, but Gizmodo did not guarantee any coverage as a condition of accepting the trip.)
With the introduction of Panther Lake and updates to XeSS upscaling software comes Intel’s new Xe3 graphics microarchitecture, which sits under the umbrella of Arc B-Series (Battlemage). Intel said we should expect better performance at lower wattages than its previous Arrow Lake H lineup and much better frame rates in games with the 12Xe core chip variants. But what about everything else? Intel detailed the “Next Arc family” will be labeled Xe3P. No, not Xe4. Will it be a discrete GPU, aka the rumored “Celestial” or C-line of graphics cards?
Intel’s head of architecture, graphics, and software, Tom Petersen, told reporters in a roundtable Q&A that Panther Lake would only hint at what’s implied by the name. “Xe3P is a significant architectural advancement from where we are now,” he said. Whether that means it’s a whole family of products, that doesn’t matter. However, it may still be called “Celestial,” more for the sake of continuity than anything.
“Our naming is not great,” Petersen said. “If we knew what we knew now, we would name those things differently.”
Even Intel doesn’t know what it will do with Nvidia
Intel has other things on its mind. The company needs you to know that its Fab 52 in Chandler, Ariz., that’s generating the company’s new 18A process is up and running. So much so, they strapped me and a host of other journalists and analysts in a white bunny suit to inspect the place. Just to enter this temple to silicon, your body is wrapped head to toe in Gore-Tex waterproof layers, your eyes and feet are covered, and you start to blend in with everybody roving those floors. What can I tell you? Not much. How big is the fab floor where they make the chips in square feet? “A shit ton,” or at least that’s what Intel spokesperson Thomas Hannaford was allowed to say. I couldn’t take pictures. I couldn’t tell you how big the lithography machines were. That would give some competitors an edge, perhaps give an indication to the world of how many chips they planned to ship, or so that’s what Intel claimed.
As I stared up at the flying shuttle robots roving across the ceiling—looking like the two-pronged “Recognizer” vehicles out of a Tron movie—while they carried wafers to and fro across rails in the ceiling, I could tell I was a resource for Intel’s mission statement. Fab 52 has been in production since 2021. Since then, the person who started Intel on this mission for U.S. manufacturing, Pat Gelsinger, was pushed out as CEO and the company went into a year-long spiral that culminated with President Donald Trump pushing the federal government to take a 10% stake in the company. Then, Nvidia came in with its Scrooge McDuck-sized moneybags ($5 billion, to be exact) to pump even more fuel into the chipmaker’s furnaces. Among all the capital changing hands, Nvidia and Intel’s respective CEOs touted a new combo chip that would combine Team Blue’s CPU with Team Green’s GPU.
The fab is only as important as the chips they make with it. And while I could sit here and wax lyrical about the company’s Clearwater Forest data center chips, the PC-buying public only cares about what’s going to end up in their desktop or laptop. Companies don’t like to talk about their futures, but from what Intel execs said last week, the company itself is still trying to figure out what a partnership means.
“It’s brand new,” Petersen said, referring to the still unknown chips it could make with Nvidia. “We don’t know all the answers to that. You’ll know more about that relatively soon. We’re still in the figure-it-all-out mode.”
