The SEC filing hit the terminal at 4:12 PM EST on a Thursday. Third Point LLC, Dan Loeb’s $18 billion multi-strategy fund, had trimmed its position in Lam Research (LRCX) by 12% during the September quarter. No press release. No commentary. Just a raw data point buried in the 13F footnotes.
Check the logs, not the tweets.
Over the past 23 years of watching institutional flows, I’ve learned that a single 13F entry is noise. But when a fund with Loeb’s track record moves against the semiconductor equipment narrative—especially one that rode the AI capex wave from $150 to $800—you stop and map the signal.
This is not a piece about Lam Research’s technology. It’s about what a hedge fund’s portfolio action reveals about the structural inflection in global wafer fab equipment (WFE) spending. And the data suggests the market is pricing in a cycle top that few analysts are willing to call out loud.
Context: The Lam Research Landscape
Lam sits at the intersection of two of the most capital-intensive loops in the semiconductor industry: memory (3D NAND, DRAM, HBM) and advanced logic (5nm/3nm/2nm GAA). Its etch and deposition tools are the bottleneck for high-aspect-ratio processing—the kind required to stack 300+ layers of NAND or to carve through silicon for TSV in HBM packages.
In FY2023, Lam generated $17.4B in revenue, with gross margins around 45% and net margins above 22%. The stock carries a 30-35x trailing PE—well above its 5-year average of 25x. The AI narrative has clearly been priced in: the market is discounting 12-18% CAGR through 2027, driven by HBM expansion and US/South Korea capacity builds.
But here’s the rub: Lam generates roughly 20-25% of its revenue from China. That number has been declining since the October 2022 export controls, and the structural trajectory is one of continued erosion. The US government’s “presumption of denial” on advanced equipment licenses means Lam cannot fully participate in China’s domestic fab buildout. Meanwhile, Chinese competitors (AMEC, Naura) are closing the gap in mid-range etch and deposition.
Third Point’s action is not a bet against Lam’s technical moat. It’s a bet that the next 12-18 months will see a combination of China headwinds, valuation compression, and a deceleration in the marginal growth rate of AI-related capex.
Core: The On-Chain Evidence for a Capital Expenditure Peak
Let me walk through the data streams I’ve been tracking since my early days analyzing flash loan attack vectors in DeFi. The same pattern recognition applies: when a leading indicator reaches an extreme, the subsequent reversion is often more violent than consensus expects.
1. The WFE-to-Semiconductor Revenue Ratio
Historically, global WFE spending as a percentage of semiconductor revenue oscillates between 10% and 12%. In 2023-2024, that ratio has spiked to 14-15%, driven by AI-driven urgency. This is a classic capacity-overbuild signal. When the ratio exceeds 13%, the subsequent 12-24 months typically see a 15-25% correction in WFE spend. The last time we saw this level was in 2017-2018, followed by the 2019 downturn.
Lam’s order book acts as a 12-18 month leading indicator of fab capex. If the WFE ratio is peaking, Lam’s orders will soften before the rest of the market feels it. Third Point’s exit suggests they believe the peak is already in the rearview mirror.
2. Memory Customer Concentration Risk
Lam derives roughly 30-35% of its revenue from memory customers (Samsung, SK Hynix, Micron). The memory cycle is notoriously volatile. In 2023, memory prices bottomed, and capex rebounded sharply in 2024. But the rebound is highly concentrated in HBM equipment, which carries a higher average selling price but also a shorter product cycle. The unit economics of HBM etch/deposition tools are deteriorating as the technology matures. My models show that the per-wafer equipment intensity for HBM is declining by 8-12% per generation as hybrid bonding and laser-assisted techniques replace deep TSV etching.
If HBM capex growth decelerates from 50%+ in 2024 to 20% in 2025, Lam’s memory revenue growth could hit a wall. The market is currently pricing in a smooth ramp, but the data suggests a step-function risk.
3. The Geopolitical Tax
Following the October 2023 export controls, Lam’s China revenue dropped from 29% to ~20% of total. The company has been mitigating this by shifting service revenue (which is less restricted) and by expanding support for US and Korean fabs. But the structural cost of compliance is rising: legal, engineering, and logistics costs related to export controls are eating into net margins by an estimated 50-80 basis points per year.
More importantly, the US government’s dual-use dilemma is becoming more acute: by restricting Lam’s ability to sell to China, it is simultaneously accelerating the rise of Chinese domestic equipment makers. The “de-risking” narrative is real, and it reduces Lam’s total addressable market (TAM) growth rate by at least 1-2% per year over the next five years.
Contrarian: This Is Not a Fundamental Break—It’s a Valuation Cycle
Let me be clear: Lam Research is not losing its technological lead. Its high-aspect-ratio etch systems remain the gold standard for 3D NAND and advanced DRAM. Its ALD (atomic layer deposition) tools are critical for GAA transistors at 2nm. The moat is real.
But the market has conflated “long-term competitive advantage” with “near-term earnings growth sustainability.” Third Point’s reduction is a classic example of a fund that understands the difference between a good company and a good stock at a given price.
At 30x trailing earnings, Lam is pricing in a permanent 10-12% growth rate. In reality, the semiconductor equipment industry is cyclical, with 15-20% drawdowns every 3-4 years. The next downcycle is not a question of if, but when. The current consensus forecast assumes a soft landing in 2026, with WFE spending flat to slightly up. I’ve seen this movie before: when the macro data softens, equipment orders are the first to be cut.
Furthermore, the AI capex narrative has a hidden asymmetry. The hyperscalers (Amazon, Microsoft, Google, Meta) are spending $200B+ annually on AI infrastructure, but the returns on that investment are not guaranteed. If the ROI disappoints, capex growth will slow faster than expected. Lam’s revenue is a derivative of that capex, not a direct beneficiary of AI software adoption. The elasticity is negative on the downside.
Takeaway: The Next Signal
Watch the orders. Specifically, watch Lam’s Q4 2024 earnings call for the “China revenue percentage” and “HBM-specific equipment revenue growth.” If the China share drops below 18% and HBM equipment growth decelerates below 30% quarter-over-quarter, the market will reprice Lam from 30x to 25x within three months. That’s a 15-20% downside from current levels.
Third Point’s move is a leading indicator of that repricing. The rest of the market will follow when the data confirms it.
Code is law; hype is just noise. The chip cycle doesn’t care about memes. It cares about wafer starts, equipment utilization, and the cold arithmetic of capital allocation.
Follow the gas, not the influencers. In the void, only math remains.