We chart the code, but the soul chooses the path.
Last week, Goldman Sachs issued a resounding buy call on three Japanese semiconductor equipment makers—Lasertec, Tokyo Electron, and Disco—citing Intel’s planned capital expenditure hike of roughly $3 billion for 2026. The narrative is seductive: a virtuous cycle of reshoring, AI demand, and geopolitical alignment. On the surface, it reads like a technical inevitability—a logical extension of the CHIPS Act and the relentless march toward 2nm. But beneath the clean Excel rows lies a moral hazard that the report glosses over: the assumption that Intel’s execution, America’s industrial policy, and Japan’s technological monopolies will harmonize without friction. As someone who has spent years inside decentralized protocol governance and watched centralized systems fracture under their own weight, I see a supply chain that is paradoxically both more resilient and more fragile than Goldman admits.
Goldman’s thesis is straightforward. Intel, under its IDM 2.0 strategy, is pouring billions into fabs in Ohio, Arizona, and New Mexico. The incremental $3 billion in 2026 capex is earmarked for 18A and 14A node equipment and advanced packaging (EMIB-T). The beneficiaries: Lasertec, which holds an 85%+ monopoly in EUV mask inspection; Tokyo Electron (TEL), the leader in coater/developer and a strong #2 in etch/deposition; and Disco, which dominates precision dicing and grinding for chiplet packaging. The logic is that as Intel buys more High-NA EUV scanners from ASML, it will need every ancillary tool from these Japanese firms. The report even sets price targets: ¥70,000 for Lasertec and ¥83,000 for TEL, implying 30%+ upside.
But here is where my own audit experience kicks in. In 2022, during the bear market, I spent six months auditing failing L1 protocols. The pattern was always the same: a beautiful white paper, a charismatic leader, and a hidden centralization point that, when stressed, collapsed. Intel’s plan is a white paper in steel and silicon. The core insight that Goldman underweights is the execution gap between capital allocation and yield. Intel’s 18A is ambitious—RibbonFET (GAA) and PowerVia (backside power)—but its historical track record on node ramps is littered with delays. The 10nm saga alone cost the company years and billions. If Intel stumbles, that $3 billion capex increment becomes a phantom order. Worse, the equipment makers have already baked that growth into their guidance. Based on my analysis of their backlog reports, Lasertec’s order book depends on EUV-related adoption; any slowdown at Intel would hit its revenue multiple disproportionately harder than TEL, which has a diversified customer base. Disco, however, stands out. Its exposure to chiplet packaging (HBM, AI accelerators) is structurally independent of Intel’s success. The AI wave will need more dicing and grinding regardless of who builds the silicon. So the investment case for Disco is more about the sovereignty of demand than the fate of one foundry.
The contrarian angle runs deeper. Goldman’s report is essentially a bet on intra-alliance trust—that the U.S. will not strong-arm Intel into favoring American equipment giants (Applied Materials, Lam Research, KLA) over Japanese rivals. But the CHIPS Act already contains provisions that prioritize "domestic supply chain security." The so-called Chip 4 alliance (U.S., Japan, South Korea, Taiwan) is not a brotherhood; it is a hierarchy disguised as collaboration. If the U.S. Commerce Department quietly mandates that 70% of Intel’s fab tools must come from American vendors, TEL’s etch/deposition business—which faces fierce competition from Lam and AMAT—could be squeezed. Lasertec and Disco, with their near-monopoly positions, might retain their slots, but the pricing power could erode. This is not a hypothetical; I have witnessed similar dynamics in the blockchain world where trusted bridges become choke points under regulatory pressure. The soul of the supply chain is not code—it is power.
Furthermore, the report ignores the balance sheet fragility of Intel itself. Its capex-to-revenue ratio exceeds 50%, far above TSMC’s 30-40%. To sustain this, Intel needs external foundry customers—Nvidia, AMD, Apple—to trust its 18A process. But those same customers have spent decades building relationships with TSMC. Trust is not transferred; it is earned through consistent yield data. Intel’s IFS segment is still losing billions per quarter. If the customers do not come, the capex spigot tightens. The Goldman recommendation implicitly assumes that Intel will succeed. That is a leap of faith, not an analysis.
So where does that leave the investor? The takeaway is not to aband on the Japanese equipment trade but to distinguish between narratives and structures. The structural case—AI-driven chiplet packaging, which benefits Disco—is robust and decentralized across multiple end customers. The narrative case—Intel’s comeback, which benefits Lasertec and TEL—is a binary bet on one company’s execution. In decentralized systems, we learn to verify, not trust. The same principle applies here: verify Intel’s 18A yield data, verify the CHIPS Act procurement details, and do not assume that the soul of this trade is aligned with your portfolio. The code of supply chains will execute—but your soul must choose the path that survives the inevitable fork.