Video Title: Re-engineering the Semiconductor Supply Chain with Intel CEO Pat Gelsinger
Video Author: No Priors
Translation: Peggy

Editor's Note: Against the backdrop of escalating AI infrastructure investment, discussions in the semiconductor industry are shifting from "Is GPU supply sufficient?" to "Can the entire computing and manufacturing ecosystem support the next phase of AI expansion?" Over the past two years, the market has been more focused on models, compute clusters, and the NVIDIA ecosystem. However, as the long-term growth in AI demand becomes consensus, a more critical question is emerging: If chips, packaging, power, materials, memory, and manufacturing capacity all become bottlenecks, what kind of new semiconductor supply chain does the AI industry need?

This episode of "No Priors" features Intel CEO Lip-Bu Tan discussing Intel's transformation, onshoring manufacturing in the U.S., foundry business, the AI-driven resurgence in CPU demand, and new manufacturing partnerships such as TerraFab. As a long-time semiconductor investor and industry operator from Cadence to Intel, Lip-Bu Tan's value in this conversation lies not in providing a single-company narrative but in presenting how an industry expert reinterprets the semiconductor structure in the AI era.

In this discussion, Lip-Bu Tan breaks down "Intel's revival" into a set of more fundamental structural issues: how to repair the balance sheet, refocus product lines, bring advanced manufacturing back to the U.S., whether AI workloads will redefine CPU value, and how semiconductor investment should revolve around real bottlenecks.

First, Intel's issue has shifted from "product lagging" to "organizational and capital structure rebuilding." In the past, external discussions about Intel often focused on process delays, GPU shortages, and foundry competitiveness. However, Lip-Bu Tan's emphasis is not on a particular product generation but on the balance sheet, organizational culture, and customer trust. The path he proposes is to first "crawl," then "walk," and finally "run": strengthen the financial foundation, streamline product lines, bring engineering teams closer to the CEO and customers, and gradually rebuild the roadmap. This means that Intel's revival is not something that can be achieved with a single product launch but rather a systematic repair involving organizational speed, capital patience, and technological roadmap.

Second, the demand for AI in the compute structure is becoming more complex. In the past, the AI narrative was almost dominated by GPUs, and training clusters became the most clear consensus in the capital markets. However, Lip-Bu Tan points out that with the development of Agentic AI, reinforcement learning, multi-agent scheduling, and edge computing, CPUs are becoming important again. The ratio of CPU to GPU may shift from 1:8 in the training era to 1:4, and even approach 1:1 in some scenarios. This means that AI infrastructure will not have a single chip winner. Future competition will revolve more around system-level combinations for different workloads: CPU, GPU, NPU, advanced packaging, software stacks, and foundry capabilities will all be part of the same compute network.

Third, semiconductor manufacturing is transitioning from a business efficiency issue back to a national infrastructure issue. Over the past thirty years, global chip manufacturing has been highly specialized driven by efficiency, with advanced manufacturing capabilities concentrated in few regions and companies. However, supply chain shocks, AI capacity demands, and geopolitical risks are making reliance on "players in one or two geographical regions" increasingly unsustainable. Chen Liwu likened the U.S. government's investment in Intel to the early relationship between TSMC and the Taiwan government, pointing to a new industrial policy consensus: for capital-intensive, long-cycle, and strategically important manufacturing systems, governments, sovereign wealth funds, and long-term capital will once again become key players.

Fourth, the semiconductor investment logic is shifting from "betting on hot races" to "identifying real bottlenecks." The key word repeatedly mentioned by Chen Liwu is not valuation but bottleneck: interconnects, photonics, EDA, advanced packaging, power conversion, heat dissipation, new materials, memory, helium, electricity, all could become constraints in the AI expansion process. In the past, semiconductor investment was avoided by VCs due to high capital expenditure, long chip cycles, and high customer switching costs; now, as AI demand brings these bottlenecks to the forefront, semiconductors have once again become an area of common interest for risk capital, strategic capital, and industrial capital. This implies that truly valuable investments are not simply chasing the "AI concept" but determining which part is becoming the constraint for the next round of expansion.

Fifth, future computing will not only exist in hyperscale data centers. The past SaaS and cloud computing era formed a highly centralized computing paradigm, but robotics, defense, home devices, Physical AI, and Agentic AI are bringing edge and edge computing back to importance. Chen Liwu does not deny the continued expansion of large AI data centers, but he is more concerned about what applications these infrastructures will ultimately serve. In other words, computing infrastructure can only generate long-term value when combined with sustainable large-scale applications. This also means that the next stage of AI competition is not just "who builds more data centers" but "who can connect computing power, chips, and application scenarios into a scalable system."

If we were to condense this conversation into one assessment, it would be: AI is shifting semiconductors from a single-chip competition to a comprehensive restructuring of the supply chain, capital structure, manufacturing capabilities, and system architecture. In this sense, the subject of this article is no longer just about whether Intel can make a comeback, but whether the computational infrastructure of the AI era needs to be redesigned from scratch.

The following is the original content (slightly reorganized for readability):

TL;DR

· The bottleneck of AI is no longer just the GPU but an industrial system constraint composed of power, memory, packaging, materials, and manufacturing capacity.

· The key to Intel's revival lies not in a one-off product counterattack, but in the systemic repair of its balance sheet, engineering culture, customer trust, and product roadmap.

· The reimportance of CPUs is not due to a cooling of the GPU narrative, but because Agentic AI, reinforcement learning, and multi-agent scheduling are creating a new computing workload structure.

· Semiconductor foundry is not just a manufacturing business, but a trust business; before customers receive wafers, they must first trust that yield, cycle time, and reliability will not jeopardize their revenue.

· The signal from TerraFab is that AI demand growth has accelerated to the point where top customers are starting to intervene upstream in the manufacturing infrastructure, rather than passively waiting for chip supply.

· The US's rebuilding of advanced chip manufacturing relies not only on fab subsidies, but on a recombination of government capital, long-term funding, industry customers, and manufacturing capabilities.

· The core of semiconductor investment is not chasing hot concepts, but identifying the true bottlenecks restricting industry expansion, such as interconnects, power consumption, heat dissipation, packaging, and new materials.

· Future AI competition will not only occur in mega data centers; edge devices, robots, defense, and Physical AI will push compute back to the application site.

Original Transcript

Host:
Hello, and welcome back to "No Priors." Today, Elad and I are joined by Pat Gelsinger. He has worked at Walden, later served as Cadence CEO, and is now Intel's CEO. We discussed his plan to transform Intel, the US government becoming a significant Intel shareholder, what it takes to be a great semiconductor investor, and whether we can actually manufacture chips in the US. Welcome, Pat.

Why Did Pat Gelsinger Take on Intel?

Host:
Pat Gelsinger, great to see you. Let's start with the most direct question: Intel is a crucial American semiconductor company, but the CEO role is exceedingly challenging. Why did you take on this job?

Pat Gelsinger:
That's a great question. I'm 66 this year. Many would say you're supposed to retire, not take on the toughest job in the industry. I did this for a few reasons. First, Intel is such an iconic company. It's core to the semiconductor ecosystem and core to America. So I decided, after Cadence, one more round.

Host:
Over the past year, a lot has happened. What has surprised you the most?

Pat Gelsinger:
What surprised me the most was something that neither my previous work experience nor training had prepared me for: one early morning, President Trump asked me to resign, citing a conflict of interest with no exceptions.

So I had to first convince myself: first, I didn't need this job. I took it solely to save Intel. Therefore, I set aside personal issues and thought about what I could do to help Intel.

The good news is that I scheduled a meeting on Thursday morning and met with him on Monday. He was willing to hear me out. I told him that I was born in Malaysia, raised in Singapore, then went to MIT, and have lived in the U.S. for a long time. I have never lived in any country other than the U.S.

I told him all this, and he listened attentively and gave me a chance. So I am pleased.

Host:
Now you have the opportunity to truly get to work. You just said that the goal of this job is to "save Intel." In your view, what does Intel's resurgence and prosperity specifically mean?

Pat Gelsinger:
I have been in office for 14 months. A lot has happened in these 14 months.

Firstly, it's about changing the culture. It's clear that we need a stronger sense of responsibility. Secondly, decisions must be made faster. I am very accustomed to the culture of a startup: moving at the speed of light, no bureaucracy, no endless layers of meetings.

So the changes I am driving include: strengthening a sense of responsibility, listening to customers, satisfying customers. Some customers say, "Pat Gelsinger is humble, willing to listen, and willing to solve their problems, striving to satisfy customers."

In addition, from day one, I decided to have all engineers report directly to me. I have an engineering background myself, and I want to know where the problems are, what needs to be corrected. I want to listen to customers, satisfy customers, and ensure we have the right products, streamline the product line, and develop a clear roadmap and vision for the next five to ten years.

Intel's Vision for the Next Decade: Save the Balance Sheet First, Then Rebuild the Product

Host:
What is your vision for Intel in the next decade?

Pat Gelsinger:
I think there are a few things. First, whether at Cadence or Intel, I have always believed: learn to crawl first, then stay humble, listen to customers; the second step is to start walking; and only then, start running, sprinting. That's my culture: one step at a time.

For me, the first step is to strengthen the balance sheet. Intel's balance sheet is in bad shape to some extent. So I am pleased to see the U.S. government become a major shareholder.

As I explained to President Trump, when TSMC started, the Taiwan regional government was also a shareholder. Look at Japan, Singapore—semiconductors are essentially infrastructure, and the U.S. government needs to provide support.

Second, I am also pleased that my old friend Ren Xun Huang invested $5 billion to support me. I'm glad I did at least something right. The $5 billion he invested has now turned into $250 billion, and even more.

There's also SoftBank's Masayoshi Son. I used to serve on SoftBank's board, and he also lent a helping hand. So first, we strengthen the balance sheet, then focus on products. I have significantly streamlined the product line, listened to customers, and driven the next generation of leadership products.

In a way, we are also very fortunate. Now Agentic AI is on the rise, and CPUs are in high demand. In the past, in training scenarios, the CPU-to-GPU ratio may have been 1:8, but now I see it could be 1:4, or even 1:1. I'm glad CPUs are important again.

I have spoken with some AI model developers. They say that in reinforcement learning and in scheduling a large number of agents, CPUs actually have an advantage. So in a way, I'm glad that there is now a very high demand for my CPUs in the market.

Overall, we need to strengthen ourselves at the product level, especially in the data center server-side. Another part is our wafer fab business.

Initially, this was a capital-intensive business and not easy. You need to meet several conditions. You need the right IP to support customers. For example, if the customer is doing mobile-related products, you must have low-power IP. If you don't have these capabilities, you can't serve them.

Foundry is a service business, but it is also a trust business. If the customer gives you the order, gives you the wafers, but the yield is not good, their revenue will be affected, and they may miss opportunities.

So, for us, it is crucial to focus on yield, defect density, cycle time, and ensure that we can meet customer needs and serve customers with high quality and reliability. These are the things I truly care about.

Ultimately, you also have to go full stack. Not just silicon, but you also need software. Some customers will ask me directly: Can you give me a whole system? This means you have to build a system. So, we are quietly and steadily advancing these things step by step, and recruiting the best talent possible.

By the way, all the recruiting I've done myself, without the help of any headhunters. So sometimes, having a strong network, knowing who to call, can be very helpful.

Host:
You've been in this industry for a long time. You were previously the Cadence CEO, I remember about 12 years?

Lip-Bu Tan:
13 years.

Host:
13 years. And then you were the Executive Chairman for another two years, so a total of 15 years.

Lip-Bu Tan:
Initially, I had only agreed to do it for three months.

Host:
Three months?

Lip-Bu Tan:
Yes. So now I'm very careful. Once you say, "I'll only do it for three months," the result might be 15 years.

What is TerraFab? Why is Musk building his own silicon wafer factory?

Host:
It seems like you still have a long way to go at Intel. Another widely discussed project is TerraFab and your collaboration with Elon Musk. Can you talk more about how this project came about? What was your involvement? How did you collaborate?

Lip-Bu Tan:
Of course. Elon Musk, who I believe we all agree is one of the greatest entrepreneurs of this century, and perhaps ever. We had a common view: that the semiconductor infrastructure had not actually kept up with the growth of AI. You need capacity, you need productivity, you need efficiency. These are all issues we both saw: there was indeed a missing link here.

Secondly, I am delighted to be working with him. He is unconventional. I call it "non-traditional." He questions every step: why do things in the traditional way? To some extent, this is very refreshing. I like it. I enjoy working with people with different viewpoints, and then finding the best path together. Both of us will learn a lot in this process. Obviously, he also has his own vision: his robots, his cars, all need a lot of silicon.

Host:
Can you explain what TerraFab is? Many people may not be familiar.

Lip-Bu Tan:
TerraFab is where he decided to build his own wafer fab. At the same time, we are excited to work with him to ensure that we can work together to get him into production faster, to ramp up faster, and to use some of our technologies and processes. This is a joint effort between us. His team is very excellent, and I communicate with them every week. It's exciting to work with them.

Host:
He has also mentioned some ideas, such as hoping to smoke in a clean room, which are usually considered...

Chen Liwu:
Yeah, yeah, and hamburgers. I don't think I would go that far. Maybe some areas of the clean room can accommodate it. But the key is to keep an open mind. We will also listen and see which things are feasible.

How is AI Reshaping the Global Semiconductor Supply Chain?

Host:
It's exciting to see you reshaping this company in the United States: gradually building the foundry business on one hand and collaborating with projects like TerraFab on the other hand. Looking at it from a global AI and semiconductor supply chain perspective, in other words, if you observe AI reshaping the supply chain at a macro level across countries, you will find that different countries are impacted differently.

For example, regarding the claim that AI is causing job losses, I think most of it is currently exaggerated. Many job cuts are actually just a result of over-hiring during the 2020 pandemic. But what I see is that the first to be cut are often outsourcing companies, as businesses are more willing to reduce external workforce before internal staff. So they will cut external customer service, external IT. This has a greater impact on countries with a large BPO industry, such as the Philippines, India, etc. They may be more immediately impacted by AI.

If we ask again, how companies in different countries can participate positively in AI's future, it almost requires a country-by-country analysis. Places with cheap energy can build data centers; places capable of training models can do so, but perhaps only the United States and one or two other places have that capability.

How do you view the changes in the global semiconductor industry supply chain? Should certain countries invest more? For example, with Israel's presence of Mellanox, NVIDIA, and Intel, should it do more in semiconductors? Should the Philippines return to a manufacturing base? How do you think about these issues from a global perspective?

Chen Liwu:
That's a good question. Obviously, AI is changing the entire landscape. I think its impact will be greater than the internet, and also more profound. AI initially helps you do things more efficiently. Many intelligent entities can help you accomplish tasks that were originally tedious but necessary, and at a faster pace. So it can significantly improve efficiency. Even in semiconductor design, AI can improve efficiency, for example, in timing closure, how fast the design can be completed; the second is cost. So all of this will help companies enhance efficiency.

AI demand and growth also face several bottlenecks. First, of course, is the well-known power constraint. Some countries simply do not have enough power, so they will be affected. Second, many people fail to realize that helium's impact on the semiconductor industry could also be very significant. Third, as we all know, there is a severe shortage of memory now, and everyone is scrambling for memory. Even if you want to build a fab to increase capacity, it will take several years. CPUs, GPUs are the same, the demand will be very high. Prices will also rise because we have to pass on the costs to customers. So all of this will impact the industry's growth.

Overall, I think the companies that have been most affected are those that have not embraced AI. AI can help businesses improve efficiency across various functions. We should embrace AI and find better ways to use it, whether for predictions, design, or different workloads. The potential here is enormous.

Host:
Many have a simple objection to whether TerraFab or Intel's foundry business can be competitive, which mainly focuses on one issue: some factors are internal to the fab, such as the IP you mentioned and operational speed; and there are also external factors. Elad just talked about a lot of these.

One of them is labor costs and actual manufacturing capacity. By investing in the foundry business, you obviously believe in a possibility: we can manufacture domestically. Elon also believes in this. Can you talk about this issue? How real is this constraint?

Pat Gelsinger:
You're talking about the labor constraint?

Host:
Yes.

Pat Gelsinger:
When I decided whether to double down on the foundry business or exit the foundry business, there were many voices in the market. As you have seen, many people said it was too expensive, it wouldn't succeed, it wouldn't work. But I ultimately decided that this is very important for the United States and for the entire industry.

We have all experienced supply chain challenges. For any large semiconductor company, you must seriously consider supply chain issues. You need a robust and resilient supply chain, not just relying on one or two participants in different geographic regions.

So, I believe more and more people will realize that manufacturing in the US is crucial. And the most advanced process, like our 14A, is about 1.4 nanometers, and we have started planning for 1 nanometer and 0.7 nanometers. The sizes are getting smaller and even finer than a human hair. Therefore, the complexity is very high and not easy to do. If any step goes wrong, all previous efforts will be wasted. So manufacturing has to be very precise.

From this perspective, this will increasingly become a bottleneck. We greatly respect TSMC; it is a great partner. More importantly, both of us need more capacity to serve customers. So we decided to grit our teeth and invest for the long term. I think in the long run, this is very critical and where I can create more value for the industry.

Host:
People have been discussing for a long time that one day we will reach a resolution limit where we cannot further miniaturize. The line width will become too small to advance. When do you think we will truly reach this limit?

Jensen Huang:
This is a great question. I believe we currently have 18A, with 14A entering mass production next, and I can still see a path to 10A and 7A. So I think this road can still go on. But it will become more and more expensive and more and more difficult. That's why we need partners. We can't do it alone. We need to collaborate with material suppliers and equipment manufacturers to truly improve yield and performance.

Another part that is also becoming a bottleneck is advanced packaging. Everyone is familiar with TSMC's CoWoS. Now we also have a very good next-generation solution called EMIB. I have to ensure that it can achieve mass production with the yield that meets customer requirements.

Now Moore's Law is also starting to lose momentum as you mentioned. So I am also researching some new materials, going back to the material level and back to the periodic table. I have invested in three types of materials: gallium nitride, silicon carbide, and indium phosphide, and I am observing how these new materials can drive the next development.

In terms of packaging, I have started to invest in glass. Glass is a very good thermal insulating material, so I invested in a startup called 3DGS. Later on, I realized that Intel has about a thousand patterns on the module, so how the substrate and module combine is very important.

We have just announced a large project with the Indian government to manufacture in India and the state of New Mexico in the United States. So advanced packaging is very important. I have also started to look at synthetic diamonds. It is also a very good insulating material. So I have also invested in Diamond Foundry. These are all next-generation directions worth paying attention to. That is to say, new materials, new substrate materials, and new design methodologies will continue to drive the industry forward.

As an engineer, you will always hit a wall. But after hitting the wall, you either find a way to jump over it or go around it, and ultimately achieve better results. As someone who has long-term investment in semiconductors and has been involved in building the semiconductor industry, from EDA tools to design and manufacturing, having these experiences is actually very helpful. Now I can contribute to the industry in my own way.

Key to Semiconductor Investment

Host:
You said something interesting just now: there is always something you can go around, but there is indeed a physical limit. When you reach scales like 7nm, you will encounter limitations and must look for new materials or other alternative paths.

An interesting question is that we have been discussing this topic for a long time. I remember 20 years ago, someone said that we would eventually reach a point where there is no more space available on a chip. Will you encounter some kind of asymptote where performance differences between different fabs are leveled out?

Li-Wu Chen:
This is a great question. With Moore's Law, in the past, we pursued doubling performance while also considering power consumption and cost. You can double performance, but cost and area cannot maintain the same advantage. So you have to make trade-offs in these aspects unless you find new materials, new design methods, and truly implement them.

I have started to recruit more talents in the materials science field. This is the innovation focus in our field: how do we continue to advance?

I still remember 18 years ago when I was still investing in semiconductors. At that time, most venture capital firms, including some excellent first-tier VCs, were my good friends. At the beginning of the partner meetings, all partners were in the room listening to me talk about semiconductors. Midway through, half of them made excuses to leave. Finally, the other half remaining would ask: Li-Wu Chen, do you have any software service projects? In the end, only two people were left listening to me out of sympathy.

So history has changed. Now semiconductors have become hot again. Look at Jensen Huang's NVIDIA, a company with a market capitalization of $5.3 trillion. Broadcom and TSMC are also at the $2 trillion market cap level. Lisa, my friend at AMD, the company has a market cap approaching $800 billion. And Intel is also close to $600 billion.

So to some extent, semiconductors have become hot again and have become critically important. 15 years, 18 years, 20 years ago, when I was investing in semiconductors, no VC wanted to invest with me except for large companies like Samsung, Arm, and SoftBank. Now I am starting to see many VCs willing to invest in semiconductors, so I am excited.

Host:
Given the tremendous investor interest in this field now, which was once considered too difficult. You have been both a long-time operator and a long-time VC at Walden. Generally, people have many concerns about investing in semiconductors, and I list a few: it has heavy capital expenditure; the success of taping out is very unpredictable; you must have a deep understanding of workloads; and another factor is the high risk of customers switching suppliers.

We have participated in some companies together, which may have already received design wins, but whether they can expand the order size is still a question. There are also cyclical aspects: you build heavy asset manufacturing capacity, but demand may change in a particular year, or it may not change.

How do you view why this industry is challenging? At the same time, there is long-term demand growth from different fields now, such as the recognition of the importance of supply chain diversification and the explosive demand growth in AI. You are still an investor, and now you have made the largest bet of your life to become a CEO. How do you think about these different risks? How would you advise others to invest in this supply chain?

I know this question is very broad, but given your experience, I feel that many people may now have a "YOLO-style investment" mentality: for example, when there is a memory shortage, they rush to buy memory stocks; but at the same time, they are unwilling to take on those things that require a ten-year timeline, such as materials science.

Li-Wu Chen: Well, your question covers a wide range of topics. Let me try to explain.

Firstly, venture capital and entrepreneurship are already in my blood, and I really enjoy this process. Not to brag, but I do have some nice exit cases. To date, I have had 159 IPOs, 126 acquisitions, including in the semiconductor field. If we only look at semiconductors, I have invested in over 200 companies over the years, with 38% of them in the United States. So I usually look at some micro-trends.

Host:
First, let me say, this is truly amazing.

Li-Wu Chen:
Thank you, thank you. I just really enjoy building these companies. But more importantly, on the investment side, the first thing I look at is: where is the bottleneck? What problem do you really want to solve?

For example, I invested in a company called Credo Semiconductor, which has a lab in Australia. At the time, I saw that interconnect had become a bottleneck, so I decided to support it. I also supported Celestial AI, which does photonic interconnect. Because within clusters, interconnect speed is becoming increasingly important, so I believe optical technology will be very important. Look at Ren Xun Huang; he has invested in almost all photonics-related companies.

Additionally, I also look at what solutions the market needs. For example, just now we talked about design complexity and cost, can AI and machine learning be used to drive better design and better solutions? There are several new startups entering the EDA space, trying to enhance performance. I think this is a gold mine.

Then there are new materials. We talked about indium phosphide, so I invested in Inphi, which was later acquired by Marvell. You can also invest in some new materials, such as gallium nitride and silicon carbide. Some of these companies have already started to be acquired, including a power management company called Empower, which has excelled in IVR.

Power management has now become a bottleneck. For example, when going from 40 volts to 1 volt, a significant amount of power is lost in the conversion process, so improving power efficiency is crucial. Therefore, power and heat dissipation have become bottlenecks.

Therefore, I always start with "What exactly do we want to solve?" Is this problem real? Are customers really suffering because of it? If they are, then I start investing.

The next step is to onboard the first customer from day one. I usually like the first customer to be a hyperscaler because of their scale. If they like your product, they are willing to pay millions of dollars over the next few years, even providing a purchase commitment. This is crucial because having a large customer allows you to scale.

So I always look at some formulas: How do you do this? Where do you find talent? Sometimes, finding talent is very important. That's why I am interested in the United States, Silicon Valley, and Austin. Additionally, Israel also has a lot of talent. So I have invested in quite a few projects in Israel.

Because Israel has many disruptive, innovative entrepreneurs who work very hard. Even during wartime, they continue to hold conference calls. Sometimes they may say, "Okay, there's an alert now, I have to go to the basement, the network may not be good, maybe we can only use voice." To some extent, this is even a bit amusing. I really appreciate this resilient entrepreneurial spirit.

Overall, I see a lot of opportunities, especially in the AI field. Now, in addition to Agentic AI, physical AI is also becoming the next huge frontier. You have to look at the problem from a full-stack perspective.

That's also why I am still deeply involved in many cutting-edge models and some of the investment projects I support because I am very bullish on open-source frontier technologies for physical AI. I think it's a gold mine.

Host:
You mentioned that there is an opportunity to use AI to make certain aspects of chip design and testing faster, cheaper, and more creative. Given your experience at Cadence, where do you think the most fertile direction lies? Is there anything that has already started to work?

Lip-Bu Tan:
I was at Cadence for about 15 years, and I'm proud of it. One of the things that I am proud of is that I was able to find my successor on the road and mentor him. Later on, he became a very outstanding CEO. He fully embraces AI and uses Agentic AI to enhance efficiency.

That's the positive side. I think Synopsys is also making efforts in these areas. They received a $2 billion investment from NVIDIA, which I think can help them a lot. They also acquired Ansys to enter the entire system design field.

Overall, these companies are all striving to do their best. But startups also have the opportunity to do some more disruptive things, eventually either go public, be acquired by these two companies, or by Siemens.

So I think the opportunity is for everyone, depending on the entrepreneur's vision. My philosophy has always been: if the entrepreneur wants to sell the company because it is a faster exit path, without a lock-up period and without having to worry about quarterly performance, that's also fine. There are also some entrepreneurs who want to IPO from day one.

As a VC, I think the three of us are all VCs. We support entrepreneurs' dreams and help them achieve those dreams.

Host:
Considering the various directions you mentioned, including future product development, or AI's impact on the semiconductor industry, we now have companies working on materials like Periodic, companies like Purepoint working on EDA and design, and other aspects along the manufacturing supply chain.

Do you think that in ten years, Intel or future semiconductor companies will be fundamentally different because of AI compared to today? If so, where do you see the difference?

Li Wu Chen:
I think they will be. First, going back to the issues you mentioned at the beginning: capital-intensive, unpredictable, cyclical. All of these factors must be considered in your investment decisions.

I usually like to get in very early, build a team. It's quite interesting. I believe you do the same. Secondly, you need to find the right investors to collaborate with you. It's not always about looking at brand institutions; I tend to focus more on individuals. Who truly understands this field? Most importantly, you need to find partners who can weather both difficult and good times with you.

Many people are happy to work with you in good times, but as soon as the company faces trouble, they leave. I like those who truly stick with the company through thick and thin. Some successful companies have almost gone bankrupt several times before taking off. So, finding partners willing to do this is crucial.

In addition, look at strategic investors; can they help the company create value in manufacturing, memory, connectivity, and other areas. I also have friends in the growth stage and hedge fund fields whom I like a lot because they have different perspectives. They understand the public market and can guide entrepreneurs on what paths to avoid. All of this is very helpful.

Overall, this is very interesting. You will realize that entrepreneurship is actually like engineering, both are about problem-solving. At every step, you need to find people who can help you solve problems. Once solved, you move on to the next frontier.

To be frank, looking back, out of the ten companies I have invested in, nine of them would change their business plans halfway because the market changed. So, I prefer entrepreneurs who are a team, not just individuals. Secondly, they must have an open mindset, willing to listen, willing to accept our guidance.

Lastly, they develop their own plans rather than just following what I say. The ideal scenario is when you give them enough feedback, and they draw their own conclusions. As long as you agree with their judgment, even if it differs from your idea, you can accept it. That's the beauty of entrepreneurship. They can move forward faster.

Going back to your question, looking ahead over the next decade, what kind of companies will win? This is just my personal view: companies that can clearly articulate their strategy, laser-focus on a specific niche, find the right partners, and have the ability to scale will win.

To some extent, this goes back to my view on being full-stack. You need to have a full-stack solution. It could be a large company that transforms into a major platform. For example, Jensen Huang, whom I admire. He focused on CUDA, focused on software libraries. He said, "I want to become a platform company," and he did achieve that.

It could also be a startup, such as Anthropic, OpenAI, which have elegantly found a path to change the game. Startups move extremely fast, advancing at the speed of light, and can also become leaders.

I hope Intel can also play such a role because we have XPU, NPU, advanced packaging, and foundries. By putting all these together, we can build dedicated chips for different workloads. I am moving in this direction.

Interviewer:
That makes a lot of sense. Part of my earlier question was about where you are headed, and another part was to ask if this will fundamentally change the way you work. Because in the software world, I see a significant change happening now: whom you hire, who you want to join your team, many people are starting to manage multiple intelligences.

So now, many people I know are more willing to hire people in their thirties, forties, or fifties because they are used to managing teams. They believe this can directly transition to managing intelligences, including understanding how to set up complex tasks, how to do QA, and so on.

I am curious, in the physical world, or in a semiconductor factory environment, how do you view team structures, skill requirements, or the changes after AI is overlaid? Is this a natural slow evolution, or will there be radical changes in certain areas? For example, in the materials field, is it now enough to use these three models along with some chemical knowledge? So I am very interested in how you perceive that future world.

Li Wu Chen:
Great question. Going back to what I said earlier about "crawl, walk, run." In the "crawl" stage, you first need to recruit the best talent in the semiconductor industry. Now I am starting to think, in order to build a full stack, what software talent do I need to bring in.

Currently, the average age of my team is around 40s to 50s, and I need to bring in some new talent. They understand workloads, understand cutting-edge models, understand open source, which is crucial.

Now, my son has become my teacher. Every time he invites me to his house, we play with the grandchildren while I ask him about AI and machine learning. He is deeper into it than me, so I have learned a lot and I am also trying to understand investments and bring in some talent.

We are transforming Intel. It used to be a very old-fashioned, traditional, spreadsheet-dependent company. Now I'm transforming it into an AI-enabled company, using AI in design and getting the entire organization to embrace AI. This way, it won't rely so much on spreadsheets and manpower.

You have to combine top talent with the best AI tools, not just for organizational management, not just for sales; now I'm also starting to think about marketing, design, and other aspects, all embracing AI.

Host:
I think for many investors, at least for me, over the past few years after starting my company, thinking about different sources of capital for capital-intensive companies has been a very educational process.

I've done a lot of software investments in the past. If you say, I need $150 million before reaching a certain critical scale, then you need some very smart friends with a completely different balance sheet.

You've been in this for a long time. You also have a unique experience of working with the government, this large stakeholder. How do you view this industrial policy? It has brought huge success stories like TSMC, one of the world's most important companies. However, in American business culture, industrial policy has long been unpopular. How do you think this notion should change now? Where does it apply?

Peter Thiel:
That's a good question. Clearly, for capital-intensive businesses and infrastructure projects, you need to access capital. To some extent, even for early-stage venture capital, many investments are starting to become capital-intensive. In the past, a VC would be willing to invest $1 billion in a company, which was unheard of in the VC industry, but now it's happening.

So, to some extent, you have to adapt. I like to look at it from a bell curve perspective. Either you enter very early because now Series A rounds can be valued at over $1 billion, so you have to enter at the pre-seed stage, before the company reaches a valuation of $20 billion or $30 billion. This is very rare today, so you have to pick the right projects.

Another part is finding capital to help the company expand. That's also why some mutual funds are starting to be willing to enter the pre-IPO market, joining me in investing in early-stage projects. I welcome them because they are not as sensitive to the requirement of "must own 20% of the company." There are not as many 20% stakes available now. So, you have to find the right investors.

In capital-intensive fields, such as AI factories and foundry businesses, you do need to leverage government funding, sovereign funds, and some very large capital. There are now some large funds specifically supporting infrastructure, and we also hope to tap into some of that capital to ensure we can scale our operations.

Overall, governments and sovereign wealth funds have become very important. At the same time, as a publicly traded company, I am also interested in focusing on some longer-term, growth-oriented investors because they can help me grow the business, rather than just focusing on short-term capital allocation, asking whether to buy back stock. These questions are also good, but at the same time, I still have to build the business. So balance is very important.

The Most Misunderstood Aspects of Intel by Investors

Host:
In your opinion, at this point, what are the most misunderstood aspects of Intel by investors?

Pat Gelsinger:
There are quite a few. First of all, let's go back to the "crawl, walk, run." In the past four months, I have been in the crawl phase. But people have begun to realize its potential. Another very important point: we must truly bring out the best products. For example, in the PC client, we still have market share. But we do need to build better performance. So I am quietly assembling CPU architecture, GPU architecture, and software architecture teams to enable us to act faster like a culture composed of multiple startups and achieve a better technological leap.

In addition to products, there is some new energy pouring in, such as Agentic AI, Physical AI. These are huge markets we can invest in.

In terms of foundry, we still have a long way to go with TSMC, both in terms of performance and other aspects. So we must remain humble, to build those basic modules, such as the IP, yield, defect density, cycle time that I mentioned earlier, and make it more efficient and reliable. Foundry is a trust business. Customers must trust you first before they hand over the wafers and rely on you. So these things take more time.

But I believe that by 2030, 2031, people will begin to see how big our potential is. In terms of products, the PC client is our foundation. Then we will move into edge computing, enter Physical AI and Agentic AI.

In the past, we mainly provided servers and PCs for humans. Now you will see another dimension: millions of intelligent entities that also need computational power, need to access the software stack. So I think this part is where we have the opportunity to participate.

The game is not over. We can continue to play in Agentic AI and Physical AI. This is the direction I want to go.

AI is just beginning. The training part is led by Jensen; the intelligent entities in edge computing, Agentic AI, and Physical AI, I think are all huge opportunities. Everyone has a chance. So this is the direction I want to strive for.

I hope investors will understand that although in the past 14 months we have already created a 6x return for shareholders, this is just the beginning. We still have a lot of room to grow.

Host:
From here, there is still venture capital-style returns ahead.

Chen Liwu:
Yes. I've always been looking for 10x opportunities. As someone with a VC mindset, you always aim for 10x.

At Cadence, during my tenure as CEO, starting from an interim CEO point of $2.42, by the time I stepped down as Executive Chairman, I had created approximately an 85x return for shareholders. Close to 76x, even 85x.

At Intel, it's challenging to achieve this because the base is larger. So I said, well, let's aim for 10x. If we can achieve 10x in five or ten years, I think that would be a very good return. As someone with a VC mindset, that's my goal.

Will Mining Power Always Stay in Data Centers?

Host:
I wish you success in accomplishing this very ambitious mission on top of this already sizable base. There was an implicit assumption behind your description just now, which is where the workload will run. Some would say we will only build larger and larger data centers, 1 gigawatt is just the beginning. Centralized operation and even centralized inferencing will dominate in terms of efficiency.

But some would consider the edge, the client side. Do you believe there will be some kind of equilibrium in future computing? Or will it be determined solely by the workload itself? How do you see it?

Chen Liwu:
That's a very good question. Right now, AI infrastructure is being massively built out, and I think that's the right thing to do. I don't think it will slow down because workloads are increasing dramatically.

Host:
We are currently facing a supply limitation.

Chen Liwu:
Yes, a supply limitation. So if there's anything that could slow down development, it's the supply constraint.

But on the other hand, I always view all this infrastructure buildout as eventually serving what solutions, what applications. I focus more on the applications. If you can identify a huge application, or several applications together that make sense, and focus around that, not everyone involved in the build will win. Some will win big, some will slowly fade, or move sideways.

Just like the internet era. You can see some companies end up very big, like Amazon, Netflix; some companies go towards marginalization, disappear, or get acquired. So for me, the approach is the same. What we really need to focus on is, what applications do they want to serve? How big is that application? Is it sustainable? Is it too crowded?

If it gets too crowded, maybe in the end, only one or two will be left, and the others will be consolidated. So, this industry will experience rapid growth, then start consolidating, and eventually maybe one or two companies will emerge as the real winners. We have seen this movie before, so I am not surprised.

Focus on applications. Netflix is an application, Amazon is a true application. In my opinion, they are the winners.

Host:
But you are assuming that some of these applications would be better served through client-side or edge computing rather than relying entirely on data centers?

Chen Liwu:
Exactly.

Host:
I have also invested in some robotics and defense companies myself, so I know that edge computing is a very important option. For example, if there are robots in the future at home, what computing power and connectivity you assume exist at home will determine what you can do. I think in the SaaS era, this was somewhat forgotten.

Chen Liwu:
Yes. My investment logic is: Find a problem that really needs to be solved. Second, find players you can collaborate with. Third, look at the application. How big is this application? Is it sustainable? If it is really big, and you believe in it, then double or triple down.

Host:
But you also include betting on applications that have not yet been widely deployed.

Chen Liwu:
Yes.

Host:
Great. Thank you very much for joining the show today. It's been a pleasure talking to you.

Chen Liwu:
Thank you very much.

[Video Link]

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia