Intel CEO Lip-Bu Tan’s First Podcast Interview: Our Goal Is “10x in 5–10 Years”—Betting on Advanced Packaging, Glass Substrates, and Synthetic Diamonds
TechFlow Selected TechFlow Selected
Intel CEO Lip-Bu Tan’s First Podcast Interview: Our Goal Is “10x in 5–10 Years”—Betting on Advanced Packaging, Glass Substrates, and Synthetic Diamonds
Systematically reconstructing the technology roadmap to break through physical limits.
Navigating Web3 tides with focused insightsSource: Wall Street Insights
Intel CEO Lip-Bu Tan stated his return target for Intel is “10x within 5 to 10 years” and is systematically redefining Intel’s technology roadmap around advanced packaging, novel semiconductor materials, and next-generation substrate technologies.
In a recent podcast episode, Tan elaborated on his transformation plan for Intel: after stabilizing the balance sheet and streamlining the product portfolio, he is now shifting focus toward advanced packaging technologies—specifically EMIB—as well as glass substrates and new materials including gallium nitride (GaN), silicon carbide (SiC), indium phosphide (InP), and synthetic diamond—to address the challenge of traditional process node scaling approaching physical limits. He also revealed that the explosive growth of agent AI and inference workloads is driving a strong resurgence in CPU demand, with the CPU-to-GPU ratio in data center servers shifting from a historical 1:8 toward 1:4—or even lower.
Tan noted that Intel has delivered approximately 6x returns to shareholders over the past 14 months—but “this is just the beginning.” He expects the market to begin recognizing Intel’s full potential between 2030 and 2032—not only in its traditional PC client base, but also across emerging markets such as edge computing, physical AI, and agent AI.
In his view, integrating Intel’s XPU architecture, advanced packaging capabilities, and foundry services into a unified offering will enable customized chip solutions tailored to diverse workloads—a long-term strategic pillar he has anchored for the company.
New Materials as the Breakthrough Lever: Advanced Packaging and Glass Substrates Take Center Stage
Against the backdrop of traditional process node scaling increasingly hitting physical limits, Tan is pivoting toward materials science and advanced packaging. He confirmed that Intel has already begun volume production of its 18A process, is advancing toward 14A volume production, and can envision viable pathways down to 10nm—and even 7nm—yet cautioned that “this path will become progressively more expensive and more difficult.”
To this end, Tan has launched multiple initiatives in packaging materials. He invested in 3DGS, a glass substrate company, drawn by glass’s unique thermal and electrical insulation properties. For chip-to-chip interconnects, Intel is championing its next-generation advanced packaging technology EMIB and has announced new advanced packaging manufacturing collaboration projects in India and New Mexico. With roughly 1,000 patents in the module domain, effectively integrating substrates and modules is a core engineering priority Tan emphasized.
On semiconductor新材料 frontiers, Tan said he has invested in GaN, SiC, and InP—and some of those portfolio companies have since been acquired by major semiconductor firms like Analog Devices (ADI). He has also backed a synthetic diamond wafer company, betting on diamond’s potential as a high-performance thermal management material in chip packaging. “That’s the engineer’s mindset—you keep hitting bottlenecks, then figure out how to leap over them or go around them,” he said.
Foundry Business: Trust First, Yield and Cycle Time Are Core Metrics
Intel’s foundry business was once widely perceived as unsustainable—but Tan chose to double down. His rationale centers on strategic supply chain security: advanced domestic manufacturing is critical for the U.S., and no major semiconductor company should concentrate its supply chain heavily in just one or two geographic regions.
Operationally, Tan has prioritized yield, defect density, and cycle time as the key performance indicators for the foundry business. He stressed that foundry is fundamentally a trust-based business—“customers must trust you before handing over their wafers.” Once yield falls short, customers risk revenue loss and are unlikely to return.
He also clarified that Intel and TSMC are partners—not merely competitors—and that the industry overall needs more capacity to meet surging demand. He forecasts that Intel’s foundry business will begin demonstrating its true potential in the market between 2030 and 2032.
Terafab Collaboration: Building Semiconductor Infrastructure with Elon Musk
Tan revealed that Intel’s Terafab project with Elon Musk emerged from a shared assessment: semiconductor infrastructure—including capacity, production efficiency, and power efficiency—is lagging behind the explosive growth of AI demand. Under this partnership framework, Musk plans to build his own wafer fab, while Intel will provide technical and process support to accelerate production. Tan said he meets weekly with Musk’s team, and collaboration is progressing smoothly.
He also noted Musk’s unconventional operational thinking—for instance, discussions about permitting smoking in certain cleanroom zones. “I may not go that far, but perhaps certain areas could allow it—the key is maintaining an open mindset.”
Investors’ Biggest Misconception: Intel Is Still in “Crawl” Mode; Real Potential Emerges Post-2030
Responding to skepticism about Intel’s transformation pace, Tan invoked his longstanding “crawl-walk-run” framework. He explained that the past several months remain firmly in the “crawl” phase: Intel is quietly building out teams in CPU architecture, GPU architecture, and software architecture, aiming to drive breakthrough innovation at startup speed; on the foundry side, the gap with TSMC remains wide, demanding humility and disciplined investment in foundational capabilities like IP, yield, and reliability.
“My VC intuition tells me—to find 10x opportunities,” Tan said. Citing his tenure at Cadence as a benchmark: from interim CEO to retirement, he delivered approximately 76x–85x returns to shareholders. He acknowledged Intel’s larger scale makes replication harder—but “10x within 5 to 10 years” is his explicitly defined goal.
Below is the transcript of the interview:
Host: Welcome back to No Priors. Today, Allad and I are joined by Lip-Bu Tan—legendary investor at Walden, former CEO of Cadence, and current CEO of Intel. We’ll discuss his plan to transform Intel, what it means for the U.S. government to be a major shareholder, how to be a great semiconductor investor, and whether we can manufacture chips domestically. Welcome, Lip-Bu.
Why Take on the Intel Challenge?
Host: Let’s start with an obvious question. Becoming CEO of this critically important U.S. semiconductor company is truly hard work. Why did you take it?
Lip-Bu Tan: Great question. I’m 66 years old—many people say, “You should retire. Why take on the hottest job in the industry?” There are two main reasons: first, Intel is an iconic company, vital to the entire semiconductor ecosystem and to the United States; second, after Cadence, I decided to do something big again.
Host: A lot has happened over the past year. What surprised you most?
Lip-Bu Tan: The biggest surprise was something I’d never experienced in any prior job or training: early one morning, President Trump asked me to resign, citing a conflict of interest—with no exceptions. My first reaction was to convince myself: “I don’t need this job—I’m doing it purely to save Intel.” Setting aside personal emotion, I began asking: “What can I do for Intel?” Fortunately, I secured a meeting on Thursday morning and another on Monday. He heard me out—I shared that I was born in Malaysia, grew up in Singapore, graduated from MIT, and have lived in the U.S. ever since, never leaving. He listened—and gave me the green light to continue. I’m deeply grateful.
Host: You described this role as “saving Intel.” What does Intel winning—and thriving—look like in your mind?
Lip-Bu Tan: It’s been 14 months, and much has changed. First, transforming the culture—establishing clear accountability and accelerating decision-making. I’m used to startup speed—everything moves at light speed—but Intel had layer upon layer of bureaucratic meetings, which I had to change. Second, listening to customers—true customer satisfaction requires humility, active listening, confronting their problems head-on, and solving them. Third, from day one, I mandated that all engineering teams report directly to me. As an engineer by training, I need to know firsthand where things break and what needs correction. Listening to customers, ensuring customer satisfaction, delivering the right products, simplifying the product line, and defining a clear roadmap and vision for the next five to ten years—those are my priorities.
Intel’s Ten-Year Vision
Host: What’s your vision for Intel a decade from now?
Lip-Bu Tan: My approach—whether at Cadence or Intel—is always crawl, walk, run: stay humble, listen to customers, then move forward step by step.
Step one: strengthen the balance sheet. Honestly, it was in terrible shape when I arrived. I’m genuinely pleased the U.S. government became a major shareholder. I explained to President Trump: look at Japan, look at Singapore—this is infrastructure-level support, and governments rightly play a role.
Second, I’m deeply grateful to my longtime friend Jensen Huang—he invested $5 billion in Intel, and I’m proud to say his $5 billion has grown to $25 billion or more. Also, Masayoshi Son of SoftBank—where I served on the board—stepped up. Through these efforts, we stabilized the balance sheet.
Next comes focusing on products, simplifying the lineup, listening to customers, and launching next-gen leading products. Coincidentally, demand for agent AI and inference CPUs is surging—so in a sense, I caught a tailwind. The CPU-to-GPU ratio used to be ~1:8 during training; now I see it moving toward 1:4—or even lower. CPUs are regaining importance—and I’m thrilled.
I’ve spoken with AI model developers who say CPUs actually perform better in reinforcement learning and in orchestrating and scheduling all agents. So demand for my CPUs is very high right now. After solidifying our data center server lineup, the other major business is our foundry. It’s capital-intensive and not easy. You need the right IP portfolio—for example, low-power IP for mobile customers—if you can’t serve them, you’re out. It’s a service business—and a trust business. If yield slips, customers lose revenue and won’t come back. So I’m laser-focused on yield, defect density, and cycle time—ensuring high-quality, reliable service. Ultimately, we must go full-stack—not just silicon. You need software. Customers ask me, “Give me a full rack”—you must deliver system-level solutions. I’m quietly executing all these steps—and recruiting the best talent I can find. By the way, every hire is done personally by me—no headhunters involved.
Collaborating with Elon Musk on Terafab
Host: Another widely discussed initiative is Terafab—and your collaboration with Elon Musk. How did this come together, and how do you collaborate?
Lip-Bu Tan: I believe we both agree: Elon Musk is among the greatest entrepreneurs of this century. We share a conviction—that semiconductor infrastructure hasn’t kept pace with AI growth—in capacity, production efficiency, or power efficiency. We both see the gap.
Second, I truly enjoy working with him. He’s wildly unconventional—challenging every step with “Why do it the traditional way?”—which is refreshing. I love hearing different perspectives, then collaboratively finding the optimal path—both sides learn immensely. He has a crystal-clear vision: his robots and cars require massive volumes of chips.
Terafab specifically means Musk will build his own wafer fab—and we’re delighted to partner, helping him accelerate production using our technologies and processes. It’s a joint effort. His team is excellent—I meet with them weekly—and collaborating with him is exhilarating. He’s floated ideas like allowing smoking in cleanrooms—“I may not go that far, but maybe in some zones, it’s possible—the key is staying open-minded.” We’re actively listening and evaluating.
Global Semiconductor Supply Chain Evolution
Host: Zooming out, how is AI reshaping the global semiconductor supply chain—country by country? What’s your macro view?
Lip-Bu Tan: AI’s impact on the landscape will surpass that of the internet—and be more profound. AI first makes you more efficient: with intelligent agents handling tedious tasks, many jobs get done faster. In semiconductor design, for example, timing optimization and time-to-market improve dramatically—and costs drop.
AI demand growth faces several bottlenecks: first, power constraints—some countries simply lack sufficient electricity; second, helium—its impact on semiconductors is significant, though often overlooked; third, memory shortages—the most urgent issue today—even if you ramp up capacity now, new fabs take years to come online. CPUs and GPUs are equally constrained, pushing prices up—and costs ultimately pass through to end users.
The companies hit hardest are those refusing to embrace AI. AI boosts efficiency across virtually every corporate function—businesses must proactively adopt AI and discover better ways to leverage it—whether for prediction, design, or workload management.
Host: The simplest argument against Terafab and Intel’s foundry competitiveness is labor cost and domestic manufacturing feasibility. What logic underpins your decision to double down on foundry?
Lip-Bu Tan: When deciding whether to double down on or exit foundry, there were many voices—saying it’s too expensive, unworkable. But my final judgment was: it’s critically important for the U.S.—and for the entire industry.
We’ve all faced supply chain challenges. Any major semiconductor company must seriously consider supply chain resilience—building a robust, diversified supply chain—not relying entirely on one or two geographically concentrated suppliers. More and more people will realize domestic U.S. manufacturing is essential.
Our most advanced processes—like 18A, equivalent to ~1.4nm—are already in production; we’re planning for 1nm and 0.7nm. Process nodes shrink below hair-widths—complexity skyrockets. One misstep can invalidate everything. Hence, manufacturing precision becomes increasingly critical—and increasingly the bottleneck.
We deeply respect TSMC—we’re strong partners—and the industry needs more capacity to serve customers. So we’re doubling down. Long-term, this is pivotal—and where I can create maximum value for the industry.
Physical Limits and Advanced Packaging
Host: People have long debated when chip scaling hits physical limits—when lines get so narrow they can’t shrink further. When do you think we’ll truly hit the wall?
Lip-Bu Tan: We have 18A in volume production, are advancing 14A, and can see viable paths to 10nm and 7nm—yes, it’s doable, but increasingly costly and difficult. That’s why we need partners—to work closely with substrate suppliers and equipment vendors, jointly improving yield and performance.
Another emerging bottleneck is advanced packaging. TSMC has CoWoS; we have EMIB, our next-gen solution—and I must ensure it achieves customer-required yields at volume production.
When traditional scaling hits walls, I return to materials for breakthroughs—GaN, SiC, InP—I’ve invested across all three. In packaging materials, I’m focused on glass—a superb thermal insulator—and invested in 3DGS. Intel holds ~1,000 module patents; integrating substrates and modules is a key challenge. We recently announced advanced packaging manufacturing collaborations in India and New Mexico. And I’m also tracking synthetic diamond—an exceptional thermal management material—and invested in a diamond wafer company.
That’s the engineer’s mindset—you hit bottlenecks, then figure out how to leap over them or go around them. Having deeply engaged across the full semiconductor lifecycle—from EDA tools to design to manufacturing—I’m thrilled to apply that experience to contribute meaningfully to the industry.
Host: Could process node convergence flatten performance differences among foundries—creating some kind of asymptote?
Lip-Bu Tan: Moore’s Law is about transistor density doubling—but power and cost don’t scale down proportionally. You can double performance, but area and cost may not halve. Unless you find new materials or new design methodologies—that’s why I’m aggressively hiring materials science talent. It’s now central to innovation in this field.
Eighteen years ago, when I invested in semiconductors, many top-tier VCs showed zero interest. I recall presenting semiconductors at a partner meeting—half the room walked out mid-presentation; the remaining half asked, “Do you have any software or services deals?” Only one or two stayed out of sympathy. Today, NVIDIA—led by Jensen Huang—is valued at $5.3 trillion; Broadcom and TSMC each hover near $2 trillion; my good friend Lisa Su’s AMD approaches $800 billion; Intel near $600 billion. Semiconductors are hot again—indispensable infrastructure. Fifteen to twenty years ago, almost no VC would co-invest in semiconductors with me—except giants like Samsung, ARM, and SoftBank. Now VCs flood in—enthusiasm is enormous. I’m deeply gratified.
Challenges in Semiconductor Investing
Host: You’re both a long-term investor and an operator. Semiconductor investing is tough—capital intensive, outcomes unpredictable, requiring deep workload understanding, high switching costs for customers, and strong cyclicality… How do you assess these risks—and advise others where to invest along this supply chain?
Lip-Bu Tan: Venture entrepreneurship is in my blood—I truly love it. Not to boast, but for context: I’ve led 159 IPOs and 126 M&A exits—with over 200 semiconductor investments, 38% in the U.S.
My investment methodology starts with one core question: Where’s the bottleneck—and what problem are you solving? For example, I invested in Cradle Semiconductor because interconnects became the bottleneck; I invested in Celestial AI because optical interconnects are growing critical in clusters—Jensen Huang has invested in nearly every photonics company. That’s no coincidence.
At the design layer, can AI and ML reduce complexity and improve design quality? I see huge opportunity in EDA—several startups are pursuing this. It’s a goldmine. In新材料, GaN, SiC, and InP are all in my portfolio—some acquired by ADI and others. Power management—from converting 40V down to 1V—is highly inefficient—and another bottleneck I’m targeting.
My investment framework is always: Is the problem real? Are customers truly struggling with it? Critically: Who’s the first target customer? I prefer hyperscalers—they have the scale, willingness, and commitment. If they love your product, they’ll spend millions—and offer guarantees—because landing one major customer enables rapid scaling.
Talent is equally vital—U.S., Silicon Valley, Austin, and Israel are my key focus areas. Israeli founders are exceptionally disruptive and incredibly hardworking. Even during wartime, they hold meetings—sometimes saying, “Air raid siren—heading to the basement; connection may drop—let’s switch to voice.” Their resilience inspires me.
Beyond agent AI, physical AI is the next frontier—requiring full-stack attention. That’s why I remain deeply involved in cutting-edge model-related investments—I’m extremely bullish on open-source frontier tech for physical AI. It’s another goldmine.
Lessons from Cadence
Host: You mentioned AI enabling faster, cheaper, and more creative chip design and testing. Drawing on your Cadence experience, which directions are most fertile—and what’s already working?
Lip-Bu Tan: I spent nearly 15 years at Cadence—and one of my proudest achievements was identifying and personally mentoring my successor, now an outstanding CEO actively embracing AI—integrating agent AI into tools to boost efficiency. Synopsys’s Sassine is doing the same—backed by NVIDIA’s $2 billion investment—and acquired Ansys to expand into full-system design.
Big players are moving—but startups still have room for disruptive innovation, eventually going public or being acquired by the two giants. It depends on the founder’s vision. My consistent philosophy: If a founder wants a quick exit, help them achieve it; if they aim for IPO from day one, help them get there. As VCs, we support founders’ dreams—and help them realize them.
Scaling and Investment Decisions
Host: Looking ahead 10 years, will AI render Intel—or future semiconductor companies—unrecognizable, given your points on capital intensity, unpredictability, and cyclicality?
Lip-Bu Tan: Yes, I believe so. These traits—capital intensity, unpredictability, cyclicality—must inform investment decisions. I typically enter early and build the team; find investors who’ll stick with you through tough times—not just fair-weather friends; and seek strategic investors who add value in manufacturing, memory, interconnects, or elsewhere. I also work with growth-stage and hedge fund partners—they bring unique public-market insights, helping founders avoid pitfalls. That’s invaluable.
Honestly, looking back, nine of the ten companies I invested in changed their business models midstream—because markets shifted. So I prefer founders with strong teams—not solo operators. They need open minds—willing to listen and accept advice—but ultimately form their own judgments. The best outcome isn’t “he told me what to do, so I did it”—but rather, you give enough feedback, and they independently derive the conclusion you endorse or understand. That’s the joy of entrepreneurship.
Looking ahead 10 years, winners will be those who focus on a niche, find the right partners, and scale effectively. Full-stack solutions matter. Large companies can emulate Jensen Huang—focusing on CUDA and platform-building—and succeed. Startups can follow Anthropic or OpenAI—changing the game elegantly—and move at light speed to become dominant.
For Intel, I want it to play this role—we have XPU, advanced packaging, and foundry. Integrating them to deliver workload-optimized custom chips is my direction.
Team Restructuring in the AI Era
Host: Software is changing dramatically—who to hire, who can manage multiple agents. Many now prefer hiring people aged 30–50, as their team-management skills transfer directly to managing agents. In hardware or foundry contexts, how do you see team structure and capability evolving?
Lip-Bu Tan: Back to the crawl-walk-run framework. During “crawl,” I recruited the semiconductor industry’s best talent. Now I’m thinking about what software talent we need to build full-stack capability—and noticing our team’s average age is mid-40s to 50s. I need younger talent who understand workloads and cutting-edge open-source models.
Interestingly, my son has become my teacher. Every time I visit him and play with my grandson, I ask him about AI and ML—he knows more than I do. I’ve learned a lot—and try turning those insights into investment decisions and hiring choices.
Intel used to be a very old-school, spreadsheet-dependent company. I’m transforming it into an AI-powered enterprise—not just in design, but across the entire organization, reducing spreadsheet reliance. We’re combining senior technical talent with AI tools—not just in sales and marketing, but now actively in design too.
Industrial Policy and Capital Sources
Host: For capital-intensive enterprises, accessing funding has always been tough. Industrial policy created TSMC—the world’s most important foundry—but this approach has long been unpopular in U.S. business culture. How do you view this?
Lip-Bu Tan: Capital access is critical for capital-intensive businesses and infrastructure projects. Today, some VCs are willing to commit $1 billion to a single company—unimaginable before. So early-stage strategies must either enter very early, at reasonable valuations—or wait until Series A—but now Series A valuations exceed $1 billion, making entry difficult.
Capital that supports scaling—like mutual funds—is welcome. They’re less sensitive to ownership stakes. For capital-intensive projects like AI factories or foundries, government funding, sovereign wealth funds, or large infrastructure funds are essential. Sovereign wealth and government capital will grow increasingly vital.
As a public company, I consciously focus on long-term growth-oriented investors—not short-term capital asking quarterly, “When will you buy back shares?” Of course, shareholder returns are a legitimate concern—but I must also build the business. Striking that balance is crucial.
Investors’ Biggest Misconception About Intel
Host: What’s the biggest misconception investors currently hold about Intel?
Lip-Bu Tan: Several points. First, back to crawl-walk-run: I’m still crawling—but people are already spotting potential. On products: we retain PC client share, but must significantly boost performance—so I’m quietly building CPU, GPU, and software architecture teams to prepare for breakthrough leadership, moving at startup speed and leveraging superior technology.
On foundry: our gap with TSMC remains large—we must stay humble, focus on fundamentals—IP, yield, defect density, cycle time—to make foundry more efficient and reliable. It’s a trust business—customers must trust you before handing over wafers. These take time—but I believe by 2030–2032, people will begin seeing Intel’s true potential.
PC clients remain our foundation—but we’re expanding into edge, physical AI, and agent AI. Previously, you served humans with servers and PCs—but now there’s a new dimension: millions of agents needing compute access and software stack access. I believe Intel has opportunities in both agent AI and physical AI—the game isn’t over.
AI is just beginning—you have Jensen Huang dominating training, you have edge, you have agent AI, and physical AI—a massive opportunity where everyone still has a shot. That’s where I’m all-in. Past 14 months delivered 6x returns to shareholders—but this is just the beginning, with huge upside remaining.
My VC instinct tells me—find 10x opportunities. At Cadence, from interim CEO to retirement, I delivered ~76x returns to shareholders; as executive chairman, closer to ~85x. Intel is larger and harder to replicate—but my goal is 10x: 10x within 5 to 10 years. As someone whose DNA is VC—that’s my goal.
Where Will Compute Reside?
Host: Some argue data centers will grow ever-larger—gigawatt-scale is just the start, and centralization is inevitable. Yet your business vision includes edge and client computing. How do you see compute ultimately distributed across data centers, edge, and client—or is it fully workload-driven?
Lip-Bu Tan: Massive AI infrastructure buildout is correct—and I see no reason for slowdown, as workloads keep growing. Current constraints are predominantly on the supply side—any slowdown stems from supply constraints, not demand.
But I’m more focused on: once all this infrastructure is built, what applications will run on it? You must identify truly scalable applications—just as Amazon and Netflix emerged in the internet era, while others vanished or got acquired. AI will follow the same pattern: massive growth, then consolidation, culminating in one or two real winners.
Focusing on applications is key. Netflix is a real application. Amazon is a real application. Both won. And certain applications are inherently better suited for edge or client—robotics, defense—where on-device compute choices are critical. Your assumptions about connectivity and on-device capability define what’s possible—something SaaS-era companies overlooked.
My investment method is always: find a real problem, identify the right partners, assess whether the application’s market size is sustainable—if you truly believe in it, double down or triple down. That includes betting on applications not yet scaled.
Host: Thank you so much for joining us today—it’s been a genuine pleasure.
Lip-Bu Tan: Thank you for the invitation.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
