Intel is quietly exploring a surprising direction: a return to memory technology. Current CEO Lip-Bu Tan recently revealed in a public conversation that he is researching "new memory architectures" and has recruited executive talent from SK Hynix, offering a rare glimpse into the chip giant's transformation strategy.
In a dialogue with his longtime friend and investor Michael, Tan stated that he established a "pet project" for this purpose, hiring Shil Lee, who previously led SK Hynix, to join Intel. "People can probably guess what direction I'm thinking about, but we're not ready to share details yet," he said. He emphasized that the memory field has lacked innovation, but new technologies are now emerging, transforming it from a commodity business into a strategic area worth deep exploration. The stacking technology between CPU and memory will offer multiple integration paths.
This statement comes at a critical juncture as Intel pushes forward with a difficult strategic restructuring. After taking over as CEO in 2024, Tan faces the task of revitalizing a chip veteran that has fallen behind in three major waves: mobile internet, cloud computing, and AI. He admitted, "In the past, Intel missed mobile internet, cloud computing, and AI. So from now on, I won't miss any major wave." The memory layout is a concrete manifestation of this strategic determination.
Vertical Integration Logic: The Trinity of Products, Packaging, and Foundry
Tan elaborated on the core logic behind his insistence on vertical integration. He believes that combining product design, advanced packaging, and foundry manufacturing creates greater value for customers than any single component. "You must have the complete technology stack, and it can't just be for the PC client; it must extend to new frontiers like agentic AI, edge computing, and physical AI."
On packaging technology, Intel currently has the EMIB next-generation packaging technology. Tan said they are further extending towards glass substrates and synthetic diamond materials—the latter is a heat dissipation material with excellent insulation properties, an area where Tan has personally invested. Regarding heat dissipation paths, he described the evolution from air cooling to liquid cooling, and then to microfluidic cooling, viewing it as crucial supporting infrastructure for AI computing expansion.
In the server and data center business, Tan acknowledged that Intel "has made many mistakes over the years." However, he emphasized that the company has rehired top CPU, GPU, and system architects and expanded its software talent pool to regain a dominant position in the next computing cycle. He stated that he receives calls from CEOs every day hoping to get more CPU supply, a clear signal of demand.
Investor DNA: Early Bets on Credo and SambaNova
Before becoming Intel CEO, Tan was known as a contrarian investor in the semiconductor field. Even before AI became mainstream consensus, he systematically invested in key infrastructure areas like high-speed interconnects and new AI architectures.
In the conversation, he recalled that nine to ten years ago, based on his judgment of GPU high power consumption, he bet on two technology routes simultaneously. One was Credo Semiconductor, focusing on serial buses and wafer-scale expansion, now valued at around $50 billion. The other was SambaNova, which uses a dataflow architecture to achieve stronger performance with lower power consumption. SambaNova is currently completing a Series F round of about $800 million to $1 billion, with Tan following through from Series A and currently serving as chairman.
He also has investments in the photonic interconnect field, having invested in Celestial AI (later acquired by Marvell) and Dust Photonics (later acquired by Credo). "Interestingly, we invested in both Credo and Dust Photonics, and finally Credo acquired Dust Photonics, so the two investments kind of 'merged'," he said.
Transferring Cadence Experience: Cultural Rebuilding and Customer Priority
Tan considers his 15-year management experience at Cadence Design as a key reference for leading Intel. When he took over Cadence, the stock price was around $2; when he left, the cumulative increase was over 3000%. He attributes this transformation to cultural change rather than a simple strategic adjustment.
"My biggest competitor once told me, 'I finally figured out your secret. My customers treat me as a supplier, but the same customers treat you as a partner,'" Tan said. He described a management style centered on quickly responding to customer complaints and actively engaging at the frontline. One customer reported that after lodging a complaint, someone arrived to solve the problem within 24 hours.
He indicated that Intel's current path is "very similar, just more complex"—with a larger workforce and two business lines, products and foundry, needing simultaneous drive. He emphasized, "Staying humble and listening carefully have always been my hallmark," and considers this management style a core tool for driving change in large organizations.
Long-Term Framework: A 10 to 15-Year Platform Vision
Regarding Intel's future, Tan explicitly rejects short-term perspectives. He said in the conversation that every time he takes over a company, his expectation is to stay for 10 to 12 years or even longer. It was 15 years for Cadence, and he measures Intel the same way. "I'm not the kind of person who looks at the short term; I look at the next 10 to 15 years—how to build a bigger platform that truly brings value to the entire industry."
He revealed that he has a "pet project" researching new memory architectures. He hired Shil Lee, who previously led SK Hynix, to join Intel. "People can probably guess what direction I'm thinking about, but we're not ready to share details yet."
"You have to always pay attention to what's happening," he said. "That's also the most interesting part of being a tech investor—you learn something new every day."
Comments