Intel (NASDAQ:INTC) board member and executive Lip-Bu Tan used a wide-ranging conversation to outline what he described as a complex turnaround and expansion effort spanning the company’s core product roadmap and its ambitions to operate a general-purpose foundry business serving external customers.
Balancing products and a service-oriented foundry
Tan said he joined Intel’s board for two years before deciding to take on his current role, calling the company “iconic” and strategically important to the industry and the United States. He described his first 10 to 11 months as “marching off the map,” citing the need to respond to unexpected issues in real time.
Process roadmap: 18A progress and focus on 14A
Tan highlighted Intel’s 18A process node as a key near-term priority, saying that when he took over, yields were “quite poor.” He said he brought in help from industry contacts and companies including PDF Solutions and KLA to drive best-practice yield improvements. Tan stated he pushed for 7%–8% yield improvement per month and said he is now seeing that pace.
He referenced Intel’s announcement of Panther Lake and said he can “really count on” the company’s fabs to produce it. He also said that “a couple of customers” have shown interest in 18A after seeing signs of progress, though he emphasized that he would not name customers due to confidentiality.
Looking further out, Tan said Intel is “laser-focused” on 14A, which he described as the most advanced node in the roadmap. He said 14A is planned for risk production in 2028 and volume production in 2029. Tan added that beyond yield improvement, foundries must also control process variation and provide predictability, and must supply comprehensive IP—particularly low-power IP needed to serve mobile customers—areas he suggested were missing in the past.
Tan said Intel expects to have a “0.5 PDK” available “this month” so customers can use test chips to work with Intel. He added that customers are “engaging heavily,” and he hopes to see volume commitments in the second half of the year as customers specify which products and volumes they want Intel to manufacture.
Signals of customer demand and how Intel plans to respond
While Tan declined to identify customers, he described how Intel intends to demonstrate demand through investment decisions. He said that if observers see Intel putting money into glass substrate—calling it an important material—or adding certain capital equipment to scale, it would indicate that “real customers” have committed and Intel is responding with disciplined investment.
On scaling the foundry business, Tan said he encourages customers to start by assigning a portion of their most important products—“5%, 10%, 20%, 50%”—so Intel can earn trust over time. He tied that approach to relationships built during his time at Cadence, saying trust is critical in foundry services.
AI infrastructure constraints: memory, thermals, interconnect, and software
Asked what limits AI growth first, Tan pointed to memory as the biggest constraint. He said that in his discussions with major players, he has been told there is “no relief until 2028,” arguing that AI workloads consume large amounts of memory. He also discussed rising compute demand, saying customers are “crying for more products” and that Intel did not initially prepare enough production to meet requirements. Tan said he is focused on production and supply chain execution to meet demand, adding that CEOs have called asking for more supply.
Tan also emphasized several additional bottlenecks and areas of focus:
- Thermals and power: He said high-performance processors can be forced to reduce frequency due to thermal issues, making liquid cooling, microfluidic cooling, and immersion cooling increasingly important.
- Interconnect: He said interconnect is becoming critical, shifting from copper to optical to address speed and latency needs.
- Software stack: Tan argued that progress requires a full-stack approach beyond silicon, including cluster management software. He said Kubernetes is valuable but may not pinpoint root causes in complex clusters, and noted startups are targeting tools to diagnose and manage cluster issues.
He also said quantum computing is “around the corner” and framed “physical AI” as a next wave after “agentic AI,” followed by quantum.
Open source, research, and competition with China
Tan said he is passionate about education and expressed concern that foundational research is weakening, citing professors being recruited to Asia and Europe. He argued that public companies often face short- and mid-term pressures that make long-horizon research investment difficult.
On open source, Tan said he is a strong believer and warned the U.S. may be behind China in some respects, describing DeepSeek as a “wake-up call.” He said he initially assumed China would lag due to limited access to advanced GPUs and tools, but he was surprised by the talent concentration he encountered while recruiting CPU architects, claiming Huawei has “100 CPU architects” who are “top-notch.” Tan relayed that candidates told him they can work around tool limitations and are “quietly building” capabilities they lack, leading him to warn that China is “shortly behind” and could “leapfrog” if the U.S. is not careful.
He also contrasted infrastructure development, citing U.S. regulatory approval timelines versus China’s ability to move faster when priorities are set. On open-source AI specifically, Tan said some companies label efforts as open source but later close them off, and he called for continued encouragement of open source to avoid duplicated efforts and accelerate improvement. He acknowledged the economics are challenging due to training costs and said some groups are working to rebuild and fund open-source communities and AI research through new institutes outside traditional universities.
GPUs, architecture flexibility, packaging, and materials
Tan said Intel will build GPUs in the future and noted he has hired a chief GPU architect. He framed CPUs and GPUs as important for different workloads and said he is “not hung up with just x86,” adding that he embraces RISC-V and Arm. He also emphasized the role of “Software 2.0,” describing an approach where software requirements drive hardware design.
He discussed advanced packaging as an emerging bottleneck and said Intel is investing to enable “system wafer kind of packaging arrangement.” On materials, Tan said Intel is “double down on glass,” mentioned diamond as another area of interest, and pointed to gallium nitride as useful for RF and switching applications. He suggested CMOS is “a little bit run out of steam,” making new materials increasingly important.
Advice to enterprise leaders: start with outcomes and rebuild foundations
Speaking to enterprise audiences, Tan advised organizations to begin with the specific problem they are trying to solve and the outcomes they want from AI adoption. He said legacy IT foundations can hinder progress if AI is layered onto outdated systems without modernization. Tan said Intel itself has legacy infrastructure and that he recruited a new CIO, directing her to reassess foundational systems and adopt AI tools function-by-function with measurable success metrics.
He also referenced research suggesting global productivity growth remains low despite AI advances, arguing that broader adoption is needed to see measurable productivity gains. Tan urged leaders to define outcomes, rebuild foundations where needed, and track measurable improvements that can be communicated to boards in terms of productivity and revenue growth.
About Intel (NASDAQ:INTC)
Intel Corporation, founded in 1968 by Robert Noyce and Gordon E. Moore and headquartered in Santa Clara, California, is a leading global designer and manufacturer of semiconductor products. The company is historically notable for introducing the first commercial microprocessor and for driving the x86 architecture that underpins many personal computers and servers. Intel’s core business spans the design, fabrication and marketing of processors, chipsets and related components for a wide range of computing applications.
Intel’s product portfolio includes client and mobile processors marketed under brands such as Intel Core and Pentium, as well as high-performance Xeon processors for data centers and cloud infrastructure.
