Intel CEO Lip-Bu Tan used his first in-depth podcast appearance on No Priors to describe how he is reshaping the company’s technology roadmap and operating model. According to TechFlowPost, citing Wall Street News, Tan has set a long-term return target of ten times over five to ten years. He said Intel has already created roughly a sixfold return for shareholders over the past 14 months, but described that as only the beginning. His plan extends beyond Intel’s traditional PC client base into data center servers, edge computing, physical AI, agentic AI and foundry manufacturing.
The interview introduced Tan as a legendary Walden investor, former Cadence CEO and current Intel CEO. Asked why he accepted such a difficult role at the age of 66, Tan said Intel is an iconic company with importance for the entire semiconductor ecosystem and for the United States. After Cadence, he said, he wanted to do one more major thing.
From saving Intel to a ten-year operating vision
Tan also described the most unexpected episode of his tenure: President Trump asked him to resign early one morning, citing conflicts of interest and saying there would be no exception. Tan said he first convinced himself that he did not need the job and was doing it to “save Intel.” He later secured meetings on Thursday morning and again on Monday, explaining that he was born in Malaysia, grew up in Singapore, graduated from MIT and had lived in the United States ever since. Trump listened to that account and gave him the chance to continue, Tan said.
Inside Intel, Tan said the first tasks were cultural change, clear accountability and faster decision-making. He contrasted the speed of a startup with the layers of bureaucratic meetings he found at Intel. From day one, he required all engineering teams to report directly to him because he is an engineer by background and wanted to know where problems were, what needed correction and which customer issues had to be solved. His repeated themes were humility, listening to customers, simplifying the product line and building a clear five-to-ten-year roadmap.
Tan described Intel’s path as crawl, walk and run. The crawl phase began with repairing what he called a very poor balance sheet. He welcomed the U.S. government becoming a major shareholder and said he explained to Trump that, as seen in Japan and Singapore, semiconductor manufacturing sits at the level of infrastructure and deserves government support. He thanked his longtime friend Jensen Huang for Nvidia’s $5 billion investment in Intel, which Tan said had grown to $25 billion or more. He also credited SoftBank’s Masayoshi Son, noting that Tan had served on SoftBank’s board. With the balance sheet strengthened, Intel shifted to product focus and next-generation leadership.
Agentic AI revives CPU demand as Intel pushes XPU and full-stack systems
Tan said the rise of agentic AI and inference workloads has sharply increased demand for CPUs. In training, the CPU-to-GPU ratio used to be roughly one to eight, he said; now he sees it moving toward one to four and even lower. Some AI model developers told him CPUs perform better in reinforcement learning and in coordinating and scheduling large numbers of agents. That has made the CPU more important just as Intel is trying to rebuild its data center server product line.
At the strategic level, Tan framed Intel’s future around the integration of XPU, advanced packaging and foundry capabilities. The goal is to provide customized chips for different workloads and, over time, full-stack systems. He said some customers have directly asked him to provide an entire rack, which means Intel must deliver not only silicon but also software and system-level solutions. Tan added that he personally handles key recruiting without using headhunters. He has hired top semiconductor talent, is now thinking about what kind of software talent is needed, and wants younger employees who understand workloads and frontier open-source models.
Tan said Intel was historically an old-school company that relied heavily on spreadsheets. He is trying to turn it into an AI-enabled organization, not only in chip design but across the company. Senior technical talent needs to be paired with AI tools, and the design side is joining sales and marketing in adopting AI. He added that his son has become his teacher on AI and machine learning; when he visits to spend time with his grandchildren, he asks his son questions and then tries to translate those lessons into investment and hiring decisions.
Terafab with Elon Musk and the case for U.S. semiconductor infrastructure
Tan gave details on Intel’s Terafab cooperation with Elon Musk. He said both sides agreed that semiconductor infrastructure has not kept pace with AI growth in capacity, production efficiency or power efficiency. Musk’s robots and cars require large volumes of chips, so he decided to build his own fab. Intel will cooperate by providing technology and process support to help accelerate production. Tan said he meets with Musk’s team every week and described the collaboration as exciting.
He also said Musk brings an unconventional operating style and repeatedly asks why a traditional approach must be used. One example was a discussion about allowing smoking in some parts of a cleanroom. Tan said he would not go that far, but some areas could be evaluated with an open mind. For Intel’s own foundry strategy, Tan argued that advanced manufacturing in the United States has strategic value for supply-chain security. Large semiconductor companies cannot rely entirely on suppliers concentrated in one or two geographies.
For Intel Foundry, Tan identified yield, defect density and cycle time as the essential operating metrics. Foundry is a service business and a business of trust, he said: before customers hand over wafers, they must trust the manufacturer. If yield misses targets and customers lose revenue, they can leave and be difficult to win back. Tan also emphasized that Intel respects TSMC and sees it as a strong partner, while the industry needs more capacity to serve customers. He described Intel’s advanced 18A process as a 1.4-nanometer-class technology and said the company is planning for 1 nanometer and 0.7 nanometer.
Advanced packaging, glass substrates and synthetic diamond
Discussing the physical limits of chip scaling, Tan said Intel has 18A, is pushing 14A into volume production, and can see a path to 10 nanometers and 7 nanometers. But that path will become more expensive and more difficult. That is why, in his view, Intel needs close collaboration with substrate suppliers and equipment makers to improve yield and performance. Advanced packaging is becoming another bottleneck. TSMC has CoWoS, while Intel is promoting its next-generation EMIB approach, and Tan said EMIB must meet customer yield requirements at the manufacturing stage.
Materials science is the other major part of Tan’s route around bottlenecks. He has invested in gallium nitride, silicon carbide and indium phosphide, and some of those portfolio companies have been acquired by large semiconductor companies such as ADI. In packaging materials, he has invested in 3DGS, a glass substrate company, because glass can serve as a strong thermal insulation material. Intel holds roughly 1,000 module-related patents, and Tan said integrating substrates and modules is a major engineering problem. Intel has also announced advanced packaging manufacturing cooperation projects in India and New Mexico. Synthetic diamond is also part of his materials roadmap: he has invested in a diamond wafer company and views diamond as another strong insulating material.
Tan said the core of Moore’s Law is doubling transistor density, but power and cost do not fall at the same rate. To keep improving performance, area and cost, the industry needs new materials and new design methods. He recalled that 18 years ago, when he was investing in semiconductors, many top-tier venture capital investors had little interest in the field. Today, Jensen Huang’s Nvidia has a market value of $5.3 trillion, Broadcom and TSMC are each around $2 trillion, Lisa Su’s AMD is close to $800 billion, and Intel is near $600 billion. Venture capital enthusiasm for semiconductors has returned strongly, he said.
Investing framework, Cadence experience and the 2030-to-2032 marker
Tan summarized his investing record as 159 IPOs, 126 M&A exits and more than 200 semiconductor investments, with 38% in the United States. His method starts with a bottleneck: what problem exists, and who is struggling with it. He cited Cradle Semiconductor as an investment tied to interconnect bottlenecks and Celestial AI as a bet on optical interconnects inside clusters. He also noted that Jensen Huang has invested in almost every photonics-related company, which Tan said is not accidental.
In EDA, Tan sees major opportunities for AI and machine learning to reduce complexity and improve design quality. At Cadence, where he spent nearly 15 years, he said one of his proudest achievements was identifying and training his successor, who is now embracing agentic AI in tools. Synopsys CEO Sassine is doing similar work, with support from Nvidia’s $2 billion investment, and Synopsys has acquired Ansys to expand toward full-system design. Tan said large companies are moving, but startups still have room to do more disruptive work and then pursue IPOs or acquisitions.
Tan’s investment questions remain consistent: Is the problem real, are customers actually suffering, and who is the first target customer? He prefers hyperscale customers because they have the ability and willingness to pay millions of dollars over several years, and sometimes provide additional guarantees, if they like a product. Power management is another bottleneck he watches, especially the loss involved in converting from 40V to 1V. On talent, he mentioned the United States, Silicon Valley, Austin and Israel. He praised Israeli founders who continue to hold meetings during wartime, even when air-raid alerts force them to move to basements and switch to voice calls. Beyond agentic AI, he described physical AI as the next major frontier.
On capital, Tan said AI factories, foundries and other capital-intensive projects require support from government funding, sovereign wealth funds and large infrastructure funds. He noted that some venture firms now put $1 billion into a single company, while Series A valuations have exceeded $1 billion. Entering very early, finding mutual funds that can support scaling, and adding strategic investors in manufacturing, memory or interconnect all matter. Looking back, he said nine out of ten companies he invested in changed their business plans along the way, so he values teams, open-minded founders and investors who stay through hard periods.
Asked about the biggest misunderstanding investors have about Intel, Tan returned to the crawl-walk-run framework. He said the company has still been crawling over the past several months, but people are beginning to see its capability. The PC client business remains Intel’s base, but performance must improve substantially. He is quietly building CPU architecture, GPU architecture and software architecture teams so Intel can move at the speed of a large startup. On foundry, the gap with TSMC remains significant, so Intel must stay humble and focus on IP, yield, defect density and cycle time. Tan said these efforts take time, but he believes that in 2030 to 2032 people will begin to see Intel’s true potential.
Tan sees Intel’s opportunity extending into edge computing, physical AI and agentic AI. He said large-scale AI infrastructure buildout has no reason to slow, with constraints coming mainly from the supply side rather than demand. Once the infrastructure is built, the question is which applications will actually run on it. He compared this with the internet era, when Amazon and Netflix emerged as real winners while other applications disappeared or were acquired. Robots and defense workloads, he added, are examples where edge or client-side compute choices are critical. Returning to shareholder returns, Tan compared the target with his Cadence record: from an initial share price of $2.4, he said shareholders received roughly a 76-fold return by the time he moved from acting CEO to retirement, and around 85 times by the end of his term as executive chairman. Intel is larger and harder to replicate, he said, but his goal remains a tenfold return over five to ten years.