Intel CEO Lip-Bu Tan told Reuters at the Cisco AI Summit in San Francisco.confirmIntel is planning to build its own GPU product line, with a very clear goal: to target the data center market, which is currently dominated by NVIDIA.
To achieve this goal, Chen Liwu revealed that he personally "poached" Eric Demers, a former senior executive at Qualcomm, to become Intel's new chief GPU architect.
Reorganizing the GPU team, targeting data centers
In an interview, Chen Liwu stated, "I recently hired a chief GPU architect, and he is exceptionally talented. I'm delighted that he has joined us." He admitted that it took considerable effort to persuade Eric Demers to join the team.
According to a previous report on the CRN website and confirmed by his personal LinkedIn profile, Eric Demers officially joined Intel last month and will report directly to Kevork Kechichian, head of chip development at Intel's data center. This also means that Intel's next-generation GPU strategy will no longer focus solely on consumer-grade Arc graphics cards, but will instead focus on the commercial applications of AI computing and high-performance computing (HPC).
"This is closely related to data centers," Chen Liwu emphasized. Intel is currently working closely with customers to define product specifications based on their actual needs.
Intel Foundry's 14A process technology is gaining favor with customers.
In addition to IC design, Chen Liwu also discussed the progress of the foundry business (Intel Foundry). He revealed that "several customers" are currently in in-depth contact with Intel, with their main interest focused on the advanced Intel 14A process technology.
He anticipates that volume manufacturing of the Intel 14A will likely begin to ramp up later this year. "To attract customers, they need to let us know the specific production volume and products so that we can plan and take the time to build up capacity."
Warning: Huawei's "poor man's version" design method should not be underestimated.
It is worth noting that Chen Liwu specifically mentioned an observation that "shocked" him during the interview. He pointed out that in a recent chip design talent recruitment event, he discovered that Huawei had recruited about 100 "top" chip designers.
Despite the strict export controls imposed by the United States on China, restricting its access to advanced EDA (Electronic Design Automation) tools, Huawei seems to have found a way out. When Chen Liwu asked these designers why they joined Huawei, they replied, "Even if we can't use tools like Cadence or Synopsys, there are still 'poor man' ways' that we can do, and we can do them."
Chen Liwu warned, "To me, they are only slightly behind us. If you are not careful, they could leapfrog us at any time."
Analysis of viewpoints
As the former CEO of Cadence, Chen Liwu's connections and technological insights in the semiconductor industry are beyond question. His personal recruitment of Eric Demers demonstrates Intel's determination to address the issue of "much ado about nothing" in GPU architecture over the past few years.
While Intel has launched its Gaudi series AI accelerators and the Ponte Vecchio GPU designed for data centers, it has consistently struggled to challenge NVIDIA's GPU dominance in terms of software ecosystem and hardware performance. This recent explicit declaration of expanding GPU deployment and targeting the data center market suggests that Intel may be adjusting its existing AI chip roadmap, attempting to develop a more general architecture that better meets the needs of cloud giants.
Furthermore, the comments regarding Huawei are quite intriguing, reflecting that American semiconductor giants are actually very aware of China's "survival instinct" and "overtaking" capabilities under sanctions at a high-level strategic level. For Intel, it's almost a situation where it faces the formidable obstacle of NVIDIA in front and pursuers like Huawei behind, coupled with the urgent need to fill its own foundry capacity. 2026 will definitely be a crucial year for a do-or-die battle.
