At the RBC Capital Markets tech bash, Intel vice president John Pitzer said 14A is already in the definition phase and that the firm is chatting with external customers far earlier than it ever did with 18A. He reckons this early hobnobbing has helped create a more mature process development kit and a more sensible approach to customer needs.
The node will have second-generation gate-all-around transistors and refined backside power delivery, both of which Chipzilla seems oddly keen to remind everyone it tried to introduce all at once with 18A.
Pitzer said: “We are all in on the development of Intel 14A and we're feeling good with the engagements we're having with external customers” and added that the firm is significantly further ahead on 14A than it had been at the same stage with 18A.
Intel has not even shipped its first 18A products yet. Panther Lake will trickle out later this year with volume ramping in early 2026, although the comparison with 14A already appears to be doing 18A no favours.
Pitzer said that during 18A’s early days, Chipzilla only optimised the node for its own internal products because it waited too long to involve customers. He said, “At 14A, it's a second-generation gate-all-around. It's a second-generation backside power.” Pitzer said customer feedback and PDK quality are far stronger this time around.
On the data centre front, the Nvidia partnership revolves around a custom Xeon CPU that Chipzilla will flog directly to Nvidia. The chip will connect to Nvidia NVLink Fusion fabric, giving Chipzilla’s x86 silicon access to the GPU maker’s high-bandwidth interconnect for the first time. Arm’s Neoverse platform will also tap NVLink Fusion.
Pitzer said: “The way the data centre side of the relationship is going to work is we will provide them with a custom Xeon part that they will then integrate into their system,” and confirmed that NVIDIA, not a hyperscaler, will take it to market.
The client side looks livelier. Chipzilla wants to bolt Nvidia's RTX graphics onto a tile inside its new SoCs, starting with high-end notebooks. Nvidia will supply the graphics tile under bailment, so customers pay Nvidia for the GPU bit while Chipzilla integrates the whole bundle into a finished SoC. OEMs will be free to mix and match RTX tiles, potentially leading to an eclectic range of configurations.
Pitzer said: “We have the opportunity to build really a new class of PC parts that we're pretty excited about”, and added that they hope to bring this combo to broader price points once the relationship matures. AMD has already grumbled that this could heat up competition and squeeze prices, but reckons it can take on Intel and Nvidia.
Chipzilla then circled back to the current supply mess. The firm admitted that older 10nm and 7nm parts remain tight, meaning Alder Lake and Raptor Lake prices are going up, while Arrow Lake and Lunar Lake are being marked down to prevent the PC market from running dry.
Pitzer said wafer output from Arizona will not improve until early next year and that the tight supply situation is “going to be here for a bit.”
Panther Lake, built on 18A, will land in the premium bracket in the first half of 2026, leaving Arrow Lake and Lunar Lake to fill the value slots until then. Given that Arrow Lake has already been discounted by almost half in some regions, Chipzilla looks determined to keep its shelves from emptying while it waits for the shinier stuff to roll off the line.
