Intel 's newest processors are now into production and making a risky bet: that graphics chips , not specialised neural engines , are the future of AI computing . The Core Ultra series 3, codenamed Panther Lake , is the company's most aggressive performance leap in years—and a fundamental departure from how every chipmaker is approaching artificial intelligence.
Doubling down on a controversial strategy: pumping AI horsepower through graphics chips rather than dedicated neural engines, Intel’s new Panther Lake chips show over 50% performance gains in both processing and graphics compared to current models. That’s early benchmarks though.
2-nanometer chips now rolling off Intel's Arizona assembly lines
Panther Lake runs on Intel's 18A process—the first 2-nanometer technology designed and manufactured entirely on American soil. Panther Lake rolled into production this week at Fab 52 in Chandler , Arizona.
The new architecture squeezes 30% more transistors onto each chip while sipping 15% less power, thanks to RibbonFET transistors and a redesigned power delivery system that routes electricity through the chip's backside.
CEO Lip-Bu Tan framed the milestone as existential for American tech leadership, part of Intel's $100 billion bet on domestic production. Fab 52 in Chandler represents the company's fifth major Arizona facility.
Intel chooses graphics over neural processors for AI muscle
Where every chipmaker is obsessing over neural processing units, Intel went a different direction. The company's new Xe 3 graphics architecture delivers 120 trillion operations per second for AI tasks—nearly double the previous generation. The NPU barely budged, inching from 48 to 50 TOPS.
Panther Lake will power everything from laptops to industrial robots, and the first products will roll out in January 2026.
Intel’s also making a 288-core server chip, dubbed Clearwater Forest , that’s made on the same 2-nanometer manufacturing process and is also launching next year.
Doubling down on a controversial strategy: pumping AI horsepower through graphics chips rather than dedicated neural engines, Intel’s new Panther Lake chips show over 50% performance gains in both processing and graphics compared to current models. That’s early benchmarks though.
2-nanometer chips now rolling off Intel's Arizona assembly lines
Panther Lake runs on Intel's 18A process—the first 2-nanometer technology designed and manufactured entirely on American soil. Panther Lake rolled into production this week at Fab 52 in Chandler , Arizona.
The new architecture squeezes 30% more transistors onto each chip while sipping 15% less power, thanks to RibbonFET transistors and a redesigned power delivery system that routes electricity through the chip's backside.
CEO Lip-Bu Tan framed the milestone as existential for American tech leadership, part of Intel's $100 billion bet on domestic production. Fab 52 in Chandler represents the company's fifth major Arizona facility.
Intel chooses graphics over neural processors for AI muscle
Where every chipmaker is obsessing over neural processing units, Intel went a different direction. The company's new Xe 3 graphics architecture delivers 120 trillion operations per second for AI tasks—nearly double the previous generation. The NPU barely budged, inching from 48 to 50 TOPS.
Panther Lake will power everything from laptops to industrial robots, and the first products will roll out in January 2026.
Intel’s also making a 288-core server chip, dubbed Clearwater Forest , that’s made on the same 2-nanometer manufacturing process and is also launching next year.
You may also like
White House swipes at Kamala Harris with 'cackle' roast, after her 'f***' jab at Trump administration - 'these…are crazy'
Drone plot foiled: Belgian PM Bart De Wever targeted in a jihadist-Iispired attack; 3 arrested
Tejashwi Yadav's big poll pledge: A government job for every household
Ambani wedding stationery: How bespoke motifs brought India's grandest celebration to life
Jimmy White's grandson signs first professional contract as he takes up rival sport
