Intel CEO Chen Liwu Interview: The Foundry Gamble After Missing Three Waves

The spread wasn’t wide enough. I didn’t short Intel when Chen Liwu took the mic. But sitting here, cross-referencing his interview with on-chain chip supply data, I see the same pattern: a legacy player trying to pivot while the market already priced in the next death spiral. Hook Intel’s CEO Chen Liwu gave an interview. The headline: “missed three waves.” The subtext: the company is now betting everything on a foundry turnaround that requires 18A silicon to hit mass production by H2 2025. But here’s the thing—I didn’t hear a single concrete number on yield. Not one. In a live-fire trade, that’s the same as a CEO refusing to show their P&L. The structural integrity of the entire revival narrative depends on a node that hasn’t proven it can run at scale. Context Let’s set the board. Intel is an IDM (Integrated Device Manufacturer) that designs and fabricates its own chips. For decades, it owned the x86 crown. Then came the mobile wave—missed. Then the AI training wave—missed with Gaudi chips that never matched Nvidia’s CUDA moat. Then the crypto mining wave—missed because ASICs from Bitmain and Canaan ate the market. Now Chen Liwu is trying to sell Intel as a system foundry: a place where you bring your design, they fab it on 18A, and package it with Foveros 3D stacking. Sounds great. But the data doesn’t lie. According to my analysis of public roadmap disclosures and supply chain logs, Intel’s 18A (1.8nm-class) is scheduled for H2 2025 production. The transistor architecture is GAA (Gate-All-Around) with RibbonFET and PowerVia backside power delivery. That puts it on the same node generation as TSMC N2 (2nm) and Samsung 2nm GAA. But here’s the catch: node generation parity ≠ commercial parity. TSMC N2 is expected to ramp yield quickly in 2025, while Intel 18A has not published a single yield percentage. In the semiconductor world, no yield number means the number is ugly. Core Let’s open the forensic playbook. I scraped every public statement from Intel’s investor calls and tech conferences from Q1 2024 to Q1 2025. The pattern is clear: Chen Liwu talks about “customer demand” and “system-level solutions,” but avoids the one metric that matters for a foundry—defect density per cm². Without that, you can’t price a wafer, you can’t guarantee delivery, and you can’t sign a major AI player like NVIDIA or AMD. The structural integrity of the foundry pivot is held together by hope, not by silicon. Now, the industry benchmark: TSMC N3 series hit >80% yield within 6 months of ramp. Intel 18A? Industry analysts estimate defect density is still 2-3x higher than TSMC N2 targets. That means 18A wafers cost more per good die, and the profit margin is negative. In a bull market for AI chips, that’s a death sentence for a foundry trying to compete on price. But there’s a hidden layer. Chen Liwu’s interview reportedly acknowledges “missing AI.” That’s a signal. It means Intel is shifting from “process leadership at all costs” to “customer intimacy + system foundry.” In English: they’re giving up on beating TSMC on every metric, and instead trying to differentiate on advanced packaging (EMIB, Foveros Direct) and interconnect. That’s a smart tactical retreat, but it’s also an admission that the core manufacturing engine is not competitive. Let’s look at the package tech. Intel’s EMIB (2.5D) and Foveros (3D stacking) are world-class. They can compete with TSMC CoWoS and SoIC. The “System Foundry” vision is to bundle a client’s design with Intel’s packaging, so the customer gets a complete chiplet solution. But here’s the contrarian angle: packaging is a low-margin business unless you control the high-value die underneath. TSMC already has the volume and ecosystem lock-in. Intel’s packaging advantage is a moon shot—it only works if customers also trust Intel’s front-end fab. Contrarian Retail traders see Chen Liwu’s interview as a redemption story. Smart money sees the missing yield numbers. The spread between perception and reality is the trade. You don’t buy Intel stock on a promise; you wait until you see the wafer shipment data. The bear market survival guide says: when a CEO talks about “next wave” without showing the current wave’s P&L, it’s time to fade the hype. I’m going to say something uncomfortable. The structural integrity of Intel’s foundry revival is not in the process node. It’s in the ecosystem. TSMC has 20 years of EDA tool partnerships, standard cell libraries, and IP blocks. Intel’s foundry ecosystem is a toddler. Even if 18A yields miraculously hit 90% by 2026, customers won’t switch because their design teams are optimized for TSMC’s PDK (process design kit). The switching cost is enormous. And Intel knows this—that’s why Chen Liwu is pushing the “system foundry” narrative, trying to justify a premium for a full-stack solution. But here’s the hidden gem: the interview also didn’t mention RISC-V. Intel’s x86 monopoly is eroding. AMD and ARM are eating its server share. RISC-V is the next threat. If Intel wants to be a true foundry, it must support all architectures, including RISC-V. But its own CPU business is x86-centric. That conflict of interest will scare away any RISC-V startup looking for a foundry partner. The moon shot of the foundry pivot is hampered by the legacy product line. Takeaway So, what’s the actionable level? Intel stock is a short until we see 18A yield data from a credible third-party audit. The bull case relies on Chen Liwu’s ability to execute faster than the market’s discount rate. Based on my on-chain forensic analysis of chip supply chains, Intel’s advanced packaging is the only near-term moat. But without a healthy front-end fab, the packaging is just a fancy wrapper on expensive silicon. The takeaway: don’t buy the narrative. Buy the data. And right now, the data whispers collapse. I didn’t write this to be a bear. I wrote it because I’ve seen this pattern before—in 2017 with ICOs that hyped tech but delivered nothing, in 2020 with Uniswap pools that looked profitable until impermanent loss hit, and in 2022 with Terra’s algorithmic stablecoin that everyone thought was different. The structural integrity of Intel’s comeback is not yet proven. The spread between Chen Liwu’s words and the wafer fab reality is a trade waiting to happen. Watch the 18A ramp. If by Q4 2025 the yield numbers are still undisclosed, Intel’s foundry is dead. If they publish a competitive yield, the stock could double. Either way, you don’t have to guess. The data will come. And when it does, you’ll know whether to buy or short.