Intel's Lip-Bu Tan Hints at Memory Return, Doubles Down on Full-Stack Vertical Integration to Chase AI's "Next Big Wave"
Deep Tech Week conference, San Francisco — Tech Surge podcast interview with Celesta Capital's Michael Marx
In a wide-ranging conversation recorded during Deep Tech Week in San Francisco, Intel Chief Executive Lip-Bu Tan laid out the clearest public articulation yet of his strategic priorities for the company, ranging from a potential re-entry into memory to an aggressive full-stack vertical integration push spanning CPUs, GPUs, advanced packaging and foundry. The interview, conducted by longtime friend and Celesta Capital managing partner Michael Marx, offered investors a rare, unscripted look at how Tan is applying a five-decade investing career and his Cadence Design Systems turnaround playbook to the largest turnaround project in American semiconductors.
A Signal on Memory That Investors Should Not Ignore
The most consequential new disclosure came almost as an aside. Asked by Marx, half-jokingly, whether Intel — which was founded in 1968 as a memory company before pivoting to microprocessors — might ever go back into memory, Tan did not dismiss the idea. "I think CPU and memory, there's a lot of ways we can really do stacking together and also try to find some new architecture for memory," he said, adding that "a lot of innovation are not there" in the memory industry today. Tan confirmed he recently hired Shih Lee, the former head of SK Hynix, calling it "one of my pet projects" and noting "you kind of know something that I'm thinking about, we're not ready to unfold it." For a company that has spent years retrenching around its core CPU and foundry businesses, a memory ambition — even an early-stage one — would mark a meaningful strategic expansion, and one that investors have not previously had visibility into.
Vertical Integration as the Core Thesis
Tan was explicit that Intel's decision to remain both a chip designer and a manufacturer, unlike the fabless model that companies such as Flex built their businesses on, is deliberate rather than inherited. "Product and advanced packaging and foundry, it tie in together, you create more value for the customer," he said. He acknowledged Intel had lost ground on its historic CPU franchise and said he has brought in new CPU, GPU and system architects to rebuild a genuine full stack, one he wants plugged directly into frontier AI labs rather than developed in isolation. The strategic framing extends beyond PCs and servers into what Tan called "agentic AI, the edge, and also the physical AI that will be the frontier going forward." Notably, Tan was candid about past failures on the server side: "we have a very strong presence in the server side, over the year we make a lot of mistakes, now I need to correct that mistake."
The Cadence Playbook, Applied to a Much Bigger Machine
Tan drew a direct line between his 15-year turnaround of Cadence, where the stock rose from roughly $2 to a peak reflecting a gain of more than 3,000%, and his current approach at Intel. The core lesson, he said, was customer obsession delivered through humility rather than hierarchy. He recounted a former competitor telling him, "my customer treat me as a vendor, but the same customer treat you as a partner," which Tan said stemmed from being willing to share roadmaps and act as a listening post for industry trends rather than hiding behind corporate distance. At Intel, he said the same cultural playbook applies "very much so, but a little bit more complex, more employees, and then we have the product, we have a foundry business, and so you have to drive success on both sides."
The Interconnect and Packaging Bets That Predated the AI Boom
Long before optical interconnects and advanced packaging became conference-circuit buzzwords, Tan was placing early venture bets in the space through Walden International, several of which have since become significant public and private companies. He cited Credo Semiconductor, now roughly a $50 billion market-cap company, and Astera Labs, now around $72 billion, as early convictions on high-speed connectivity that he made "multiple years ago." He also noted the sale of his photonics investment Celestial AI to Marvell and of Ayar Labs to Credo, calling the overlap "when investing is really fun." On the Intel side, he pointed to next-generation EMIB packaging, a move into glass substrates, and even artificial diamond as an insulating material, underscoring that Intel's packaging roadmap is being shaped by the same venture-scouting instincts Tan has used for decades. "I spend a lot of time with different professors to understand what the next frontier is going to be," he said, describing it as how Intel avoids "missing the big wave" the way it did with mobile and early AI.
Competing With Nvidia, SoftBank and Washington as Investors
Tan addressed the increasingly unusual cap table dynamics facing Intel directly, noting that Nvidia, the US government and SoftBank are now all investors in a company that competes with Nvidia in AI compute. His answer was less about strategy than temperament: "I don't view them as a competitor. I view them, you know, market is big enough, of course we just find how to create bigger market for all of us." He framed his own tenure in explicitly long-duration terms, telling his board he is "not the short-term guy" and is instead focused on building for outcomes 10 to 15 years out. Given that Tan could have retired rather than take on Intel's turnaround, that framing carries some weight with investors trying to gauge his commitment horizon, though it also raises the question of how much patience Washington and Wall Street will extend if near-term execution falters.
What It Means for the Narrative Around Intel
None of this changes Intel's near-term financial picture, and Tan avoided discussing "how much the stock's going up or going down" by design in the conversation. But the interview adds meaningful texture to the bull case some investors are constructing around Intel's foundry and product roadmap: a CEO with a four-decade record of early, contrarian conviction in semiconductors, now explicitly signaling ambitions in memory, full-stack AI silicon, and advanced packaging materials that go beyond what has been disclosed in recent earnings calls. The broader context, as global semiconductor sales surged past $790 billion last year and US firms recaptured 53% global market share, the highest since 1984, suggests Tan's decades-long bet on hardware is finally aligned with where capital and policy attention are flowing.