When Goldman Sachs slapped a Buy rating on Lasertec, Tokyo Electron, and Disco last week, the market nodded in collective agreement. The logic seemed bulletproof: Intel is ramping capex for its 18A and 14A nodes, and these Japanese equipment giants are the only ones who can supply the critical tools. But after spending 15 years auditing semiconductor supply chains and watching narrative cycles in crypto, I've learned that the most consensus trades often conceal the deepest fault lines. Let me walk you through the structural risks that the Goldman report glosses over.
Context: The CPEX Mirage
Intel's IDM 2.0 strategy is the centerpiece of this thesis. To compete with TSMC, Intel is pouring billions into new fabs in Ohio and Arizona, and its capital expenditure for 2026 is expected to rise by roughly $3 billion—mostly for High-NA EUV lithography and advanced packaging (EMIB-T). The three Japanese suppliers sit at different nodes of this value chain: Lasertec owns ~85% of the EUV mask inspection market, Tokyo Electron commands ~50% of coater/developer and a strong #2/#3 position in etch/deposition, and Disco dominates precision dicing and grinding for chiplet packaging. On paper, Intel's push is a direct tailwind.
But paper is not silicon. Here's what the bullish narrative misses.
Core: The 30 Billion Dollar Illusion
A $3 billion incremental capex increase sounds massive, but it's not a blank check for Japanese suppliers. Intel's fab tools are sourced from a global menu: Applied Materials, Lam Research, KLA, and ASML all compete for the same dollars. Splitting $3 billion across dozens of equipment categories means each Japanese company might see only a few hundred million in new orders—hardly enough to justify the double-digit upside that Goldman's price targets assume.
More importantly, Intel's capex is not purely additive. The company is already under severe cash flow pressure. Its 2024 capex was roughly 50% of revenue—twice TSMC's ratio. Intel is heavily dependent on government subsidies (CHIPS Act $8 billion) and external financing. If its foundry business fails to win major external clients (like NVIDIA or AMD), the capex plans could be slashed. We saw this in 2022-2023 when Intel cut spending by 30% after its server chip delays. The same risk applies now.
The Execution Risk Elephant
Intel's historical record on process node transitions is rocky. The 10nm (now Intel 7) was delayed by years. The 18A node, which uses RibbonFET (GAA) and PowerVia (backside power), is extremely complex. Even TSMC struggled with early yields on N3. If Intel's 18A yields are poor, it will need more inspection equipment (good for Lasertec in the short term), but longer-term, it will delay mass production, reduce total wafer starts, and ultimately lower equipment demand. Goldman's report barely touches this—it treats Intel's roadmap as a certainty. Based on my experience auditing foundry roadmaps, I'd assign a 30-40% probability of a significant 18A delay.
Lasertec vs. Disco: The Real Differentiation
Not all Japanese equipment plays are equal. Disco's exposure to advanced packaging (especially EMIB-T for AI chip chiplets) is structural. The chiplet trend is driven by NVIDIA, AMD, and cloud giants, not just Intel. Even if Intel stumbles, AI demand will keep Disco's dicing and grinding tools in high demand. Lasertec's monopoly on EUV mask inspection is also strong—every EUV fab needs it—but its reliance on Intel's High-NA ramp is binary. If Intel delays High-NA adoption, Lasertec's growth rate suffers. Tokyo Electron faces the most competition; Applied Materials and Lam Research are fierce rivals. Intel could easily tilt its procurement toward American suppliers for political reasons, especially under the CHIPS Act's 'Buy American' pressures. That risk is real.
Contrarian Angle: The Geopolitical Trade-Off
The bullish case for Japanese equipment rests on 'friendshoring'—Intel builds in the US, Japanese suppliers benefit. But this ignores a subtle dynamic: the US government may pressure Intel to use American vendors to secure supply chain 'trust'. In 2023, the CHIPS Act required recipients to prioritize US-made tools where available. Applied Materials and Lam are already lobbying for preferential treatment. If that pressure increases, Japanese share at Intel could shrink despite rising overall capex. This is a classic narrative trap: the market sees Intel's spend, but misses the allocation shift.
Moreover, Japan's own export controls on China (e.g., the 23 advanced equipment categories) have reduced revenue from Chinese customers—historically a growth driver for TEL and Disco. While Intel partially compensates, the net effect may be neutral or even negative for Japanese suppliers if US protectionism deepens.
Takeaway: Follow the Chiplet, Not the Fab
Investors should resist the urge to buy all three stocks equally. If you're betting on the Intel turnaround, you're making a high-conviction bet on Intel's execution—a risky proposition. A cleaner play is the chiplet/packaging theme via Disco, which benefits from multi-client AI demand. For Lasertec, monitor High-NA adoption closely. For Tokyo Electron, watch for geopolitical signals. The Goldman thesis has merit, but it's not a 'set and forget' trade. Trust is the only currency that matters—and Intel hasn't earned it yet.
Postscript: After publishing this analysis, several readers asked: 'What about Chinese semiconductor equipment companies as a hedge?' That's a topic for another day. But the core principle remains: Noise filtered. Signal preserved.