Re-engineering the Semiconductor Supply Chain with Intel CEO Lip Bu Tan
In a Nutshell
Lip Bu Tan is rebuilding Intel by strengthening its balance sheet, simplifying its product line, improving foundry yield and trust, and pursuing next-generation nodes like 18A and 14A while adding new materials and packaging. AI is driving strong CPU demand—especially for agentic workloads—shifting the training CPU-to-GPU ratio closer to 1:4 and requiring full-stack solutions that combine silicon, software, and advanced packaging. His strategy hinges on U.S. government support, strategic investors like Jensen Huang, and a startup-style culture of speed and accountability to regain competitiveness against TSMC and others by 2030–2032.
These notes were generated by AI and may contain inaccuracies.
Lip Bu Tan noted that nine out of the ten companies he invests in change their business plans halfway through because the market has changed. He prefers investing in teams of entrepreneurs rather than single individuals. He described his approach at Cadence and Intel as first crawling, then being humble and listening to customers, strengthening the balance sheet, focusing on products, simplifying the product line, and driving next-generation leadership products.
CPU demand is currently very high. Lip Bu Tan expressed satisfaction with this trend. He noted that Jensen Huang, his longtime friend, invested $5 billion which has grown to $25 billion over time.
Lip Bu Tan, legendary investor from Walden, former CEO of Cadence for 13 years plus two years as executive chairman, is now CEO of Intel. The discussion covers his plan to transform Intel, the US government as a major shareholder, how to be an effective semiconductor investor, and whether chips can be manufactured in the United States.
At age 66, Lip Bu Tan took the role despite suggestions to retire. He viewed Intel as an iconic company critical to the semiconductor ecosystem and the United States. He decided to do one more job after Cadence.
The most surprising event was President Trump asking him to resign early one morning due to conflict of interest with no exceptions. Lip Bu Tan first convinced himself he did not need the job and was doing it purely to save Intel. He then met with Trump on Thursday morning and again on Monday. He explained his background: born in Malaysia, raised in Singapore, attended MIT, lives in the US, and has never lived outside the country. Trump listened and gave him the opportunity.
After 14 months in the role, Lip Bu Tan focused on changing the culture to drive more accountability and faster decision-making, moving away from bureaucratic layers of meetings toward startup-like speed. He made all engineering teams report directly to him as an engineer by training. Key priorities include listening to customers, addressing their problems, delighting them, simplifying the product line, creating a roadmap and vision for the next five to ten years, and strengthening the balance sheet.
His culture follows a crawl-walk-run-sprint progression. The first step was strengthening Intel's balance sheet, which was in poor condition. The US government became a major shareholder. He referenced how TSMC started with the Taiwan government as a shareholder, and noted similar infrastructure support from governments in Japan and Singapore.
Jensen Huang invested $5 billion, now grown to $25 billion or more. SoftBank's Masayoshi Son, with whom Lip Bu Tan previously served on the board, also provided support. The strategy involves strengthening the balance sheet, simplifying products, listening to customers, and developing next-generation leadership products.
Agentic AI has increased demand for CPUs. The ratio of training CPU to GPU demand has shifted from 1:8 to potentially 1:4. AI model developers indicated that CPUs perform better for reinforced learning and orchestrating agents due to speed advantages. CPU demand is currently very high on the data center and server side.
The foundry business is capital intensive. Success requires the right IP portfolio—for example, low-power IP for mobile applications. It is a service and trust business where poor yield leads to customer revenue misses. Focus areas include yield improvement, defect density reduction, cycle time optimization, and delivering high-quality, reliable service. The goal is to move toward a full-stack offering including software and complete rack-level systems.
Lip Bu Tan personally handles all recruitment without using search firms. He maintains a rolodex of contacts to reach out to directly.
Elon Musk and Lip Bu Tan share the view that semiconductor infrastructure has not kept pace with AI growth in terms of capacity, productivity, and dry efficiency. Musk is described as unconventional, questioning traditional processes. They collaborate weekly on Terafab, where Musk decided to build his own fab. Intel provides technology and processes to help Musk reach production faster. Musk's team is considered very good to work with.
Musk suggested allowing smoking inside the clean room and other unconventional ideas. Lip Bu Tan stated he would not go that far but remains open-minded about exploring feasible changes.
AI's impact will be bigger and more profound than the internet. Initial benefits include efficiency gains through agents handling mundane tasks. AI can improve semiconductor design efficiency in timing and cost. Key bottlenecks include power constraints, helium supply, memory shortages, and increasing demand for CPUs and GPUs. Pricing is rising as costs are passed to customers. Companies not embracing AI will be most negatively impacted.
Despite many voices saying the foundry business is too expensive and will not work, Lip Bu Tan decided it is critical for the United States and the industry. Recent supply chain challenges demonstrated the need for a robust, resilient, geographically diverse supply chain rather than depending on one or two players.
Intel's roadmap includes 18A in production, 14A (1.4 nanometer) entering production, with plans for 10 nanometer, 7 nanometer, 1 nanometer, and 0.7 nanometer nodes. Complexity increases as features become smaller than a human hair. Manufacturing requires extreme precision; any mistake results in total loss. Advanced packaging is becoming a bottleneck. Intel has developed E-IP as a next-generation packaging solution to compete with TSMC's CoWoS.
Lip Bu Tan is investing in gallium nitride, silicon carbide, and indium phosphide. He is also investing in glass substrates through 3DGS for better heat insulation and has 1,000 patents related to modules. Intel announced a major manufacturing program with the Indian government in India and New Mexico. He is additionally investing in artificial diamond through Diamond Foundry as another insulator material. The approach involves hiring more material science talent and exploring new substrate materials, chemicals, and design methodologies.
Moore's Law involves doubling performance while power and cost cannot be doubled, requiring trade-offs unless new materials and design approaches are found. Intel is hiring more material scientists. Semiconductor has become essential again. NVIDIA has a $5.3 trillion market cap, Broadcom and TSMC each have $2 trillion market caps, AMD is near $800 billion, and Intel is near $600 billion.
Lip Bu Tan has 159 IPOs and 126 M&As from his investments. Of semiconductor investments, 238% are in the US. Investment criteria include identifying real bottlenecks and customer pain points. Examples include backing Celestial AI for optical interconnects due to speed requirements in AI clusters, and investing in companies using AI/ML to improve EDA design performance and cost. He targets investments where hyper-scale customers are willing to pay millions and provide warrants. Key factors include talent location, particularly Silicon Valley, Austin, and Israel.
Lip Bu Tan has backed a significant amount of investment in Israel due to the disruptive, innovative entrepreneurs and their strong work ethic. Even during wartime, teams maintain conference calls, sometimes pausing for air raid warnings and moving underground, occasionally switching to voice calls when internet connectivity is affected. Tan finds this resilient entrepreneurship inspiring and sees substantial opportunity, particularly in AI.
Beyond agentic AI, he identifies physical AI as the next major frontier. He remains involved with frontier model companies and has backed investments in open-source frontier technology for physical AI.
Tan spent nearly 15 years at Cadence and takes pride in identifying and developing his successor, who has become an effective CEO embracing AI and driving agentic AI for greater efficiency. Both Synopsys and Cadence are advancing in this space, with Synopsys receiving a $2 billion investment from Nvidia and acquiring Ansys to expand into system design. While established players are making progress, Tan sees opportunities for startups to pursue more disruptive approaches, potentially leading to IPOs or acquisitions by larger EDA companies.
As a VC, Tan supports entrepreneurs' visions whether they seek quick exits through acquisition or aim for IPOs from day one. He believes VCs should help founders fulfill their own dreams rather than impose predetermined paths.
Tan believes semiconductor companies will look radically different in ten years. He emphasizes that capital-intensive and cyclical businesses require factoring unpredictability into investment decisions. His approach includes going in very early, assembling teams, and finding the right co-investors—preferably knowledgeable individuals rather than purely institutional firms—who will remain committed through both successful and troubled periods.
He values partners who have helped companies through near-bankruptcy situations that eventually succeeded. Strategic investors who can add value through manufacturing, memory, or connectivity expertise are also important. Growth-stage investors and hedge fund managers provide additional perspectives on public markets and guidance on avoiding certain paths.
Tan notes that nine out of ten companies he has invested in changed their business plans midway due to market shifts. He seeks entrepreneurial teams rather than single founders, prioritizing open-mindedness and willingness to listen to coaching while ultimately formulating their own conclusions.
In Tan's view, winning companies ten years from now will be those that articulate and maintain laser focus on one niche area, find the right partners, and scale effectively. Full-stack solutions will be essential. He admires Jensen Huang's approach of building Nvidia into a platform company focused on CUDA and the AI ecosystem.
Startups like Anthropic and OpenAI demonstrate how to execute full-stack approaches more elegantly and change the game at startup speed. Intel's combination of XPUs, advanced packaging, and foundry capabilities positions it to build purpose-built silicon for different workloads.
Tan discusses how AI is changing team structures and hiring practices. Intel is moving from recruiting primarily semiconductor talent to also bringing in software talent to build full-stack capabilities. The company's average team age in the late 40s to 50s requires bringing in new talent who understand workloads and frontier models, including open-source approaches.
Tan mentions his son has become his teacher on AI and machine learning topics. Intel is transforming from a legacy spreadsheet-driven company to one embracing AI across design, engineering, sales, marketing, and all organizations, reducing dependence on manual processes.
For capital-intensive businesses like AI factories and foundries, access to capital is critical. Tan notes that venture capital is evolving, with some firms now willing to invest $1 billion into companies—an amount previously unheard of. Series A rounds exceeding $1 billion valuations and pre-money valuations of $20-30 billion are becoming more common.
Beyond traditional VC, mutual funds are moving into pre-IPO stages. Government funding, sovereign funds, and large infrastructure-focused funds are increasingly important for scaling operations. As a public company, Intel is deliberately focusing on long-term growth-oriented investors rather than those seeking immediate capital returns or share buybacks.
Tan acknowledges investors may misunderstand several aspects of Intel. After four months of foundational work, recognition of the company's potential is growing. Intel needs to deliver better-performing products in the PC client segment where it still holds market share. The company is quietly building CPU, GPU, and software architecture teams to leapfrog competitors with a startup-like culture focused on speed and better technology.
Beyond current products, agentic AI and physical AI represent major market opportunities. On the foundry side, Intel is behind TSMC in performance metrics and must focus on fundamentals: IP development, yield improvement, defect density reduction, and cycle time efficiency. Trust is essential in the foundry business—customers need confidence before entrusting wafers.
Tan believes Intel's potential will become more visible by 2030-2032. The company's opportunity extends beyond traditional PC and server provision to supporting millions of agents requiring compute and software stack access. He sees Intel positioned to compete in both agentic AI and physical AI.
Despite achieving 6x returns for shareholders in 14 months, Tan views this as just the beginning. Drawing from his venture capital background, his goal is 10x returns over five to ten years. At Cadence, returns reached approximately 76-85x from his time as interim CEO through retirement as executive chairman, though he acknowledges the larger base at Intel makes similar multiples more challenging.
Regarding the debate between massive centralized data centers versus edge and client compute, Tan believes the industry is currently supply-constrained rather than demand-constrained. He emphasizes focusing on applications rather than infrastructure alone. Not every company building infrastructure will win—success depends on identifying large, sustainable applications.
Drawing parallels to the internet era, he notes that winners like Amazon and Netflix succeeded by focusing on specific applications, while others consolidated or disappeared. Tan's investment thesis centers on three elements: identifying problems that genuinely need solving, determining which players to partner with, and evaluating application size and sustainability. When applications are genuinely large and credible, he advocates doubling or tripling down on investment.
He sees robotics and defense applications as areas where on-device compute choices significantly impact capabilities, noting that assumptions about home robots and their connectivity determine what can be achieved.
Keep No Priors: AI, Machine Learning, Tech, & Startups in your library
Save the videos and channels worth coming back to, and find them again in one place.
