TechFlowPost, citing Wall Street News, reported that Intel CEO Lip-Bu Tan used his first extended appearance on the No Priors podcast to lay out a broad plan for remaking Intel. Tan said his goal is to deliver a 10x return for Intel over five to ten years. He also said the company has already created roughly a 6x return for shareholders over the past 14 months, while adding that this is only the beginning.
Tan’s roadmap centers on a systematic rebuilding of Intel around product focus, stronger engineering accountability, advanced packaging, new semiconductor materials and next-generation substrates. He described Intel’s opportunity as extending beyond the traditional PC client base into data center servers, edge computing, physical AI and agentic AI. In his view, Intel can create customized chips for different workloads if it can effectively integrate XPU capabilities, advanced packaging and foundry manufacturing.
Why Tan Took the Intel Job
The No Priors hosts introduced Tan as a legendary investor from Walden, the former CEO of Cadence and the current CEO of Intel. Asked why he accepted what the hosts described as an extremely difficult role at a critical American semiconductor company, Tan said he is 66 and many people told him he should retire. He gave two reasons for taking the job: Intel is an iconic company that matters deeply to the entire semiconductor ecosystem and to the United States, and after Cadence he wanted to do one more big thing.
Tan said the most surprising moment of the past year was one he had not encountered in any previous job or training: President Trump asked him to resign early one morning, saying there was a conflict of interest and no exception. Tan said he first convinced himself that he did not need the job and was doing it purely to save Intel. He then put his personal emotions aside and focused on what he could do for the company.
He said he secured a meeting on Thursday morning and had another meeting on Monday. Tan told Trump that he was born in Malaysia, grew up in Singapore, graduated from MIT and had lived in the United States ever since without leaving. According to Tan, Trump listened to the explanation and gave him a chance to continue. Tan said he was very grateful for that opportunity.
The Crawl-Walk-Run Approach
Tan summarized his operating method with the phrase crawl, walk, run. In the crawl phase, he said, Intel must remain humble, listen to customers and build step by step before moving faster. Over the past 14 months he has focused on changing culture, making accountability clear and accelerating decision-making. He said he is used to a startup pace, while Intel had layers of bureaucracy and meetings that he had to change.
From the first day, Tan decided that all engineering teams would report directly to him. As an engineer, he said he wanted to know personally where the problems were and what needed to be corrected. His early priorities were listening to customers, making customers satisfied, ensuring the right products, simplifying the product line and setting a clear vision and roadmap for the next five to ten years.
Tan said Intel’s balance sheet was in very poor condition when he took over. He expressed appreciation that the U.S. government became a major shareholder. He said he explained to President Trump that countries such as Japan and Singapore support infrastructure at the national level, and that government support is appropriate for that kind of foundational capability.
He also thanked Jensen Huang, whom he described as an old friend, for investing $5 billion in Intel. Tan said Huang’s $5 billion investment has now grown to $25 billion or more. SoftBank’s Masayoshi Son, whose board Tan previously served on, also provided help. Tan said these moves helped stabilize Intel’s balance sheet.
CPU Demand Returns With Agentic AI and Inference
On the product side, Tan said Intel is simplifying its offerings, listening to customers and preparing next-generation leading products. He said the timing has worked in Intel’s favor because agentic AI and inference are creating very strong CPU demand. In past training workloads, the CPU-to-GPU ratio was about one to eight. Tan now sees that ratio moving to one to four, and even lower.
Tan said he has spoken with AI model developers who told him CPUs perform better in reinforcement learning and in coordinating and scheduling the speed of multiple agents. As a result, he said, demand for Intel CPUs is high. After establishing a stronger data center server product line, the other major business is foundry manufacturing.
He described foundry as a capital-intensive business, a service business and a trust business. Customers must trust a foundry before handing over wafers, and if yield is not good enough, customers may leave because of revenue losses and may not return. Tan said he is highly focused on yield, defect density and cycle time to serve customers with high quality and high reliability.
Tan also said Intel must eventually move toward a full-stack model, not only silicon. Some customers have asked him for an entire rack, which means Intel needs to provide system-level solutions as well as software. He said he is advancing these steps quietly and hiring the best people he can find. He added that he personally handles recruitment and does not use headhunters for those hires.
Foundry, TSMC and the Terafab Collaboration With Elon Musk
Tan said there were many arguments for exiting the foundry business, including claims that it was too expensive or would not work. He decided to continue because advanced manufacturing in the United States is extremely important for the country and for the industry. After supply-chain disruptions, he said, every major semiconductor company must think seriously about resilience and cannot rely completely on one or two geographically concentrated suppliers.
He said Intel has already produced 18A, which he described as a 1.4-nanometer-class process, and is planning 1 nanometer and 0.7 nanometer. As process nodes shrink, manufacturing precision becomes increasingly demanding. Any error at one step can undermine the whole process. Tan said Intel deeply respects TSMC and sees it as a strong partner, while also saying that the industry needs more capacity to serve customers.
Tan also discussed Terafab, the project involving Elon Musk. He said he and Musk share a view that semiconductor infrastructure has not kept pace with AI growth in capacity, production efficiency or power efficiency. Musk decided to build his own fab, and Intel is working with him to accelerate production by contributing technology and process support.
Tan said he meets with Musk’s team every week and described the cooperation as energizing. He said Musk challenges every convention and asks why things must be done in the traditional way. Tan mentioned that Musk raised ideas such as allowing smoking in certain parts of a clean room. Tan said he probably would not go that far, though certain areas might be considered, and the key is to keep an open mind while listening and evaluating carefully.
AI’s Pressure on the Global Semiconductor Supply Chain
Looking at global supply chains, Tan said AI’s effect on the overall landscape will exceed that of the internet and will be more far-reaching. AI helps people complete tasks more efficiently, and in semiconductor design it can improve timing optimization, accelerate time to market and reduce cost.
Tan listed several bottlenecks in AI demand growth. One is electricity, because some countries do not have enough power. Another is helium, whose impact on the semiconductor industry he said many people do not fully recognize. The most urgent issue now is memory shortage. Even if companies expand capacity today, new capacity will take years to come online. CPUs and GPUs are also in short supply, prices are rising, and costs will eventually be passed through to clients.
Tan said the companies most affected will be those that do not embrace AI. He argued that AI can help almost every business function become more efficient, from forecasting and design to various workloads. Companies, in his view, should actively adopt AI and find better ways to use it.
Advanced Packaging, Glass Substrates and Synthetic Diamond
Tan said traditional process-node scaling is moving toward physical limits. Intel has 18A, is pushing 14A into production, and can see paths to 10 nanometers and 7 nanometers, but he said the road will become more expensive and more difficult. That is why he sees the need for closer cooperation with substrate suppliers and equipment makers to improve yield and performance.
Advanced packaging is another major bottleneck. TSMC has CoWoS, while Intel has a next-generation solution called EMIB. Tan said he must ensure that EMIB achieves the yield customers require during high-volume production. When scaling approaches bottlenecks, he said, he goes back to materials to look for ways around them.
In new semiconductor materials, Tan said he has invested in gallium nitride, silicon carbide and indium phosphide. Some of those companies have already been acquired by large semiconductor companies such as ADI. In packaging materials, he has become interested in glass because it is a strong thermal-insulating material, and he invested in a company called 3DGS.
Intel has about 1,000 patents in modules, and Tan said integrating substrates with modules is an important engineering problem. Intel has also announced advanced packaging manufacturing cooperation projects in India and in New Mexico in the United States. Tan is also watching synthetic diamond, another strong thermal-insulating material, and has invested in a diamond wafer company. He summed up the engineering mindset by saying that engineers keep encountering bottlenecks and then find ways to cross them or go around them.
Tan also addressed the question of whether process-node convergence could narrow performance gaps among foundries. He said Moore’s Law is about doubling transistor density, but power and cost do not necessarily fall at the same pace. Performance can double, but area and cost may not decline equivalently unless new materials or new design methods are found. For that reason, he is recruiting more materials-science talent.
How Semiconductor Investment Changed
Tan recalled that 18 years ago, many top-tier venture capital investors were not interested in semiconductors. At partner meetings, he said, half the room would find an excuse to leave after he discussed semiconductors, while the remaining half would ask whether he had software or services deals. Only one or two people would stay out of sympathy.
Now, Tan said, Nvidia under Jensen Huang has a market value of $5.3 trillion, Broadcom and TSMC are each around $2 trillion, Lisa Su’s AMD is close to $800 billion, and Intel is close to $600 billion. Semiconductor has again become a hot and indispensable foundation. Fifteen to twenty years ago, few venture investors wanted to join him in semiconductor deals apart from major institutions such as Samsung, ARM and SoftBank. Today, venture capital has rushed into the field.
Tan said venture entrepreneurship is in his blood. He cited 159 IPOs and 126 M&A exits, with more than 200 semiconductor investments, 38% of them in the United States. His investment method begins with one question: where is the bottleneck, and what problem is being solved?
He said he invested in Cradle Semiconductor because interconnect had become a bottleneck. He invested in Celestial AI because optical interconnect is becoming increasingly important inside clusters. Tan noted that Jensen Huang has invested in nearly every photonics-related company, and he said that is not a coincidence.
In design, Tan sees large opportunities for AI and machine learning to reduce complexity and improve design quality, especially in EDA. He described the area as a gold mine. In power management, he pointed to the large losses involved in converting from 40V to 1V as another bottleneck area he likes.
His framework asks whether the problem is real, whether customers truly struggle with it and who the first target customer is. Tan prefers hyperscale customers because they have the capability and willingness to pay millions over the next few years or provide some form of support if they like the product. Winning one large customer can allow a company to scale.
Cadence Lessons, Startups and AI-Driven Organizations
Tan spent nearly 15 years at Cadence and said one of the things he is most proud of is finding and personally developing his successor, who is now an outstanding CEO and is actively embracing AI. Cadence is bringing agentic AI into tools to improve efficiency. Synopsys CEO Sassine is doing similar work, supported by a $2 billion investment from Nvidia, and Synopsys has acquired Ansys to expand into full-system design.
Tan said large companies are moving, but startups still have room to do more disruptive work. Those companies can eventually go public or be acquired by one of the two large companies. His philosophy as a venture investor is to support the founder’s dream: if founders want a quick exit, help them do that; if they want to pursue an IPO from day one, help them follow that path.
Discussing investment decisions over a ten-year horizon, Tan said capital intensity, unpredictability and cyclicality must all be considered. He usually likes to enter early, help build the team, find investors who will stay during difficult periods and bring in strategic investors who can add value in manufacturing, memory, interconnect or other areas. He also values growth investors and hedge fund contacts who can provide public-market perspectives and help entrepreneurs identify areas to avoid.
Looking back, Tan said nine out of ten companies he invested in changed their business plans along the way because the market changed. He prefers entrepreneurs with teams rather than a single founder, and teams that keep an open mind, listen and then form their own judgment. He said the best outcome is not simply doing what an investor says, but using feedback to reach a conclusion the investor can understand or support.
For the next decade, Tan said winners will be companies that focus on a specific area, find the right partners and scale. Full-stack solutions matter. Large companies can build platforms, as Jensen Huang did with CUDA. Startups such as Anthropic and OpenAI can also move at light speed and change the game in a more elegant way. For Intel, Tan wants to combine XPU, advanced packaging and foundry capabilities to build specialized chips for different workloads.
On team structure in the AI era, Tan returned to the crawl-walk-run framework. In the crawl phase he hired the best semiconductor talent. He is now considering what software talent is needed to build full-stack capability, while also noting that Intel’s average team age is in the 40s and 50s. He wants younger talent who understand workloads and frontier open-source models.
Tan said his son has become his teacher on AI and machine learning. When he visits his son’s home to play with his grandchildren, he asks him questions and learns from him. Tan then tries to translate those insights into investment judgment and hiring decisions. He described Intel as previously a very old-school, spreadsheet-dependent company and said he is trying to turn it into an AI-enabled company across the organization, not only in design.
Capital Sources, Investor Misunderstandings and Where Compute Lives
Tan said access to capital is critical for capital-intensive businesses and infrastructure projects. Some venture capital firms are now willing to put $1 billion into a single company, which he said was unimaginable before. For AI factories and foundries, companies must seek support from government funding, sovereign wealth funds or large infrastructure funds. As a public company leader, he also wants long-term growth-oriented investors rather than short-term holders who ask every quarter when the company will buy back shares.
Asked about the biggest investor misunderstanding of Intel, Tan said the company is still in the crawl phase, though people are starting to see its potential. In products, Intel still has share in PC clients, but performance must improve substantially. He is quietly building CPU architecture, GPU architecture and software architecture teams to prepare for leapfrog progress at the speed of a large startup.
In foundry, Tan acknowledged that Intel is still far behind TSMC and must remain humble. It must focus on fundamentals such as IP, yield, defect density and cycle time to make the foundry business more efficient and reliable. Because foundry is based on trust, customers must trust Intel before they hand over wafers. Tan said these efforts take longer, but he believes that by 2030 to 2032 people will begin to see Intel’s true potential.
Tan said the PC client market remains Intel’s base, but the company is extending into edge, physical AI and agentic AI. In the past, companies provided servers and PCs for humans. Now there is a new dimension in which millions of agents need access to compute and software stacks. He sees opportunities for Intel in both agentic AI and physical AI.
Tan framed his ambition through his venture-capital instinct: look for 10x return opportunities. At Cadence, he said, from acting CEO to retirement, the stock rose from $2.4 and generated roughly 76x for shareholders. By the end of his executive chairmanship, the return was about 85x. Intel is larger and harder to replicate, but his goal is still a 10x return over five to ten years.
On where compute will ultimately reside, Tan said the current buildout of large-scale AI infrastructure is correct and he sees no reason for it to slow, because workloads continue to grow. The main constraints are on the supply side rather than the demand side. But he is focused on what applications will run once the infrastructure is built.
He compared the coming AI application cycle to the internet era, when Amazon and Netflix emerged as real applications while others disappeared or were acquired. Some applications are better suited to run at the edge or on clients. Robotics and defense, for example, depend heavily on assumptions about connectivity and on-device compute. Tan said his investment method remains the same: find a real problem, find the right partners, assess whether the application market can sustain scale and, if conviction is strong, double or triple down.