Intel CEO Lip Bu Tan Bets on Agentic AI and Edge Computing to Revive Chipmaking Giant
No Priors Podcast, June 18, 2026
At 66 years old, Lip Bu Tan took on what many consider the hardest job in tech: turning Intel around. In a wide-ranging conversation on the No Priors podcast, the legendary semiconductor investor and former Cadence CEO revealed his strategy centers on a fundamental shift in AI compute architecture. Tan believes the industry is reaching an inflection point where CPU demand will surge for agentic AI and inference workloads, potentially shifting the ratio of CPUs to GPUs from today's one-to-eight in training to one-to-four or better in the emerging agentic era.
The Agentic AI Thesis Driving Intel's Revival
Tan's most significant insight involves the changing economics of AI compute. While the industry has obsessed over GPU buildouts for training, Tan sees a different future emerging. "Right now the agentic AI and inference, CPU become highly in demand," he explained. "In some way I'm happy right now the demand is very high for my CPU." This represents a dramatic reversal for Intel, which had been largely sidelined in the AI infrastructure boom dominated by Nvidia GPUs.
The CEO elaborated on the technical rationale, noting that AI model developers told him that "in term of reinforced learning, in term of the speed of orchestrating all the agents and turn up, the CPU is actually better." This architectural shift could fundamentally alter the economics of AI infrastructure if Tan's thesis proves correct. Rather than massive centralized GPU clusters, the future may involve distributed computing with CPUs handling agent orchestration and inference at the edge.
Balance Sheet Transformation and Strategic Investments
Tan has moved aggressively to strengthen Intel's financial position, securing investments from unlikely partners. The U.S. government became a major shareholder through CHIPS Act funding, which Tan defended by drawing parallels to TSMC's early backing from the Taiwanese government. More surprisingly, Nvidia CEO Jensen Huang invested five billion dollars in Intel, which has already appreciated to 25 billion in value. SoftBank's Masayoshi Son also participated, with Tan noting his longstanding relationship from his time on SoftBank's board.
These capital infusions addressed what Tan candidly described as Intel's "horrible" balance sheet when he arrived. The partnerships also signal broader industry recognition that semiconductor supply chain resilience requires multiple capable foundries, even from competitors. Huang's investment in particular suggests Nvidia sees value in a stronger Intel, whether as foundry customer, supplier, or simply as a hedge against overreliance on TSMC.
The Elon Musk Collaboration and Terafab
Intel's collaboration with Elon Musk on Terafab represents one of the more unconventional partnerships in semiconductor manufacturing. Tan praised Musk as "one of the best if not the best entrepreneur in this century" and said the two share a view that "semiconductor infrastructure actually is not catch up with the AI growth." The Terafab project aims to build Musk's own fabrication facility, with Intel providing technology and process support to accelerate production.
Tan described the collaboration as "very refreshing," noting that Musk "basically question every step" of traditional manufacturing approaches. When asked about Musk's reported desire to allow smoking in cleanrooms, Tan diplomatically suggested "maybe some part of the clean room you can do that," while maintaining an open mind about challenging conventional wisdom. The partnership gives Intel a foothold in serving Tesla's growing semiconductor needs for vehicles and robots while potentially pioneering new manufacturing approaches.
Cultural Transformation and Organizational Restructuring
The most dramatic internal change at Intel has been cultural. Coming from startup environments, Tan found Intel's bureaucracy stifling. "I'm so used to startup culture and you move fast in the speed of light and going to have that bureaucracy layer of layer of meeting," he said. His response was swift: all engineering now reports directly to him, allowing the engineer-CEO to "know what went wrong and what are the thing that I need to correct."
Tan also simplified Intel's product portfolio and instituted what he calls a "crawl, walk, run" approach focused first on customer satisfaction. He personally handles all executive recruitment without search firms, leveraging his deep industry rolodex. The average age of his team runs in the late 40s and 50s, but he's now bringing in younger talent who understand AI workloads and open-source development. In a revealing aside, Tan noted "my son become my teacher now" when it comes to the latest AI and machine learning developments.
Perhaps most tellingly, Tan described Intel as formerly "a very old legacy spreadsheet company" that he's transforming to become "AI enabled" across design, sales, marketing and operations. This represents a fundamental rethinking of how a manufacturing-intensive company can leverage software and AI tools.
Foundry Strategy and Advanced Packaging
Despite significant skepticism about Intel's foundry ambitions, Tan committed to the capital-intensive business as critical for U.S. semiconductor independence. "I finally decided this is very important for United State and also very important for the industry," he explained, noting that supply chain resilience requires multiple geographic options beyond TSMC's Taiwan concentration.
Intel's roadmap includes 14A (1.4 nanometer) entering production, with planning underway for one nanometer and 0.7 nanometer processes. But Tan also acknowledged that traditional Moore's Law scaling faces increasing challenges. His response involves new materials including gallium nitride, silicon carbide, and indium phosphide. He's invested personally in companies across these areas and is bringing materials science expertise into Intel.
Advanced packaging represents another critical bottleneck. Intel's EMIB (Embedded Multi-die Interconnect Bridge) technology competes with TSMC's CoWoS, but Tan is looking beyond current solutions to glass substrates and artificial diamond for thermal management. He invested in 3DGS for glass packaging technology and diamond foundry for heat dissipation applications. Intel also announced a major partnership with the Indian government for advanced packaging manufacturing in India and New Mexico, recognizing that module assembly expertise is as critical as leading-edge lithography.
Physical Limits and Material Science Innovation
When pressed on whether the industry will hit fundamental physical limits to miniaturization, Tan acknowledged the challenge but remained optimistic about workarounds. "I think I can see 10 and seven," he said of future process nodes, "but going to be more and more expensive and more difficult to do." His solution involves partnerships with substrate vendors and equipment manufacturers, plus the new materials research.
Tan's engineering mindset shone through when discussing these constraints: "One thing good about being an engineers you always hitting the wall then you find way to either jump over the wall or you work around the wall." This problem-solving approach extends to his investment philosophy, where he targets bottlenecks in the semiconductor ecosystem. Recent investments include Celestial AI for optical interconnects and Cradle Semiconductor (acquired by Astrella Labs) for interconnect solutions, both addressing speed constraints in AI clusters.
Venture Investment Framework for Semiconductors
Despite semiconductors being unfashionable in venture capital for years, Tan maintained his investment discipline with remarkable results. He's backed 238 companies with 159 IPOs and 126 M&A exits, with 38% in the United States. His framework focuses on identifying genuine bottlenecks, securing hyperscaler customers early, and ensuring entrepreneurs have strong teams rather than depending on a single founder.
Tan recalled presenting to venture firms 18 years ago when partners would leave the room during his semiconductor pitches, with remaining partners asking "do you have any software service?" Now with Jensen Huang's Nvidia worth $5.3 trillion, Broadcom and TSMC at $2 trillion each, and AMD nearing $800 billion, semiconductor investing has returned to favor. But Tan cautions that capital intensity, cyclicality, and customer switching costs remain real challenges.
His advice for semiconductor startups emphasizes targeting a first customer willing to pay millions and potentially provide warrants, focusing on hyperscalers with scale potential. He also stressed the importance of finding long-term co-investors who "really work through difficult time and good time," noting that many investors "walk away" when companies face challenges. Geographic talent concentrations matter too, with Tan highlighting Silicon Valley, Austin, and Israel as key hubs. He praised Israeli entrepreneurs' resilience, recounting conference calls where participants would say "there's a warning I have to go to underground and then the internet may not be good maybe we does use voice."
Edge Computing and Application-Focused Strategy
Tan's product strategy reveals a contrarian bet on edge and client computing rather than exclusive focus on massive centralized data centers. While acknowledging the current "massive buildup" in AI infrastructure, he believes the industry is "supply constraint" rather than demand-limited. More importantly, he sees the future shaped by specific applications rather than general-purpose buildout.
"I always look at all this infrastructure build up at the end you have to look at what is the solution what is the application you want to drive," Tan explained. He drew parallels to the internet era where Amazon and Netflix emerged as winners while others "go sideway and disappeared or being acquired." His focus on applications leads directly to edge computing, where robotics, physical AI, and agentic systems require local compute rather than constant cloud connectivity.
This application-centric view also informs his hiring strategy. Beyond traditional semiconductor expertise, Tan is bringing in software talent to build full-stack solutions. "Not right now in the past you basically provide the server provide the PC for human now you're starting to have Another different dimension is millions of agent they need to access to the compute they access into the software stack," he noted. This suggests Intel is positioning for a world where agents and robots create massive new compute demand at the edge.
Ten-Year Vision and Investor Expectations
Tan set ambitious goals for Intel's transformation, targeting a 10x return for shareholders over five to ten years. This echoes his Cadence tenure where he delivered approximately 76x returns from his start as interim CEO to retirement as executive chairman. He acknowledged the challenge given Intel's larger base but maintained his venture capital mindset of seeking 10x outcomes.
Looking to 2030-2032, Tan believes Intel's foundry potential will become apparent as the building blocks of IP, yield, defect density, and cycle time improvements compound. On the product side, he sees Intel moving up from PC clients into edge computing, physical AI, and agentic AI. "The game is not over yet," he insisted. "We can play on the in the injected AI and also the physical AI."
What investors may be missing, according to Tan, is the magnitude of new markets opening. He characterized Intel as "a multiple of startup culture" that can "leaprog using better technology" across different workloads. The stock has already delivered six times returns in his 14 months as CEO, but Tan views this as "just a beginning." His confidence rests on the thesis that applications haven't yet been built for the AI infrastructure being deployed, and when they emerge, distributed computing will prove more important than the current centralized GPU focus suggests.
Government Partnership and Industrial Policy
Tan offered nuanced views on government involvement in semiconductors, drawing from his unique position with the U.S. government as a major Intel shareholder. He positioned this as essential for infrastructure businesses, comparing it to TSMC's government backing and noting that "for capital intensive business and infrastructure play you need to access to the capital." This represents a significant shift in American business culture, which historically frowned on industrial policy.
The CEO also navigated political challenges, including an early morning request from President Trump to resign over perceived conflicts of interest. Tan convinced Trump to reconsider by emphasizing his American credentials: "born in Malaysia grown up in Singapore went to MIT and I live in US and never live outside country." Trump ultimately allowed him to continue, recognizing Intel's strategic importance. This episode illustrates the political complexity of leading a company now partially owned by the federal government while maintaining private sector discipline.