The WFE Cycle: Why Crypto Miners Should Care About ASML's High-NA EUV

Goldman Sachs just dropped a number that should make every crypto miner sit up. $281 billion. That's the projected wafer fab equipment (WFE) spend by 2028. A 36% CAGR from 2025. The bank is betting on AI demand for HBM and advanced logic. But here's the thing: the same machines that make NVIDIA's B200s also make Bitcoin ASICs. And the queue is about to get longer. We didn't see this bottleneck coming in 2021. Back then, the narrative was all about hash rate and energy costs. Now, the real constraint is lithography. ASML's High-NA EUV tools—each costing €300-400 million—are the gatekeepers of the next generation of chips. If TSMC and Samsung are busy filling orders for AI accelerators, where does that leave Bitmain's next-gen 3nm miner? Context: WFE is the collective term for the machines that fabricate semiconductor wafers. The industry is dominated by a handful of suppliers: ASML for lithography, Applied Materials for deposition, Lam Research for etching, KLA for inspection. These companies are the 'picks and shovels' of the digital age. The current cycle, as per Goldman, is driven by three pillars: HBM memory for AI, advanced logic (5nm and below), and a recovery in DRAM. The report forecasts WFE spending to hit $218 billion in 2027 and $281 billion in 2028. That's a massive increase from the ~$100 billion seen in 2024. For crypto, the implications are layered. First, Bitcoin mining ASICs are typically built on trailing-edge nodes—7nm, 5nm, and now moving to 3nm. But these nodes are the same ones used for AI accelerators and high-end smartphones. When AI demand surges, foundries like TSMC allocate capacity to the highest-paying customers. NVIDIA pays a premium. Bitmain does not. The result: longer lead times for ASIC orders, higher prices, and a potential cap on hash rate growth. Second, the HBM boom matters for GPU mining. Ethereum's move to proof-of-stake killed that market, but other coins (like Kaspa) still use GPU mining. HBM is a key component for high-performance GPUs, and its tight supply trickles down. Goldman expects DRAM tightness to persist through 2028, meaning HBM prices stay elevated. That makes GPU mining rigs more expensive, reducing the marginal profitability for altcoin miners. Third, the geopolitical layer. The US export controls on advanced chipmaking equipment to China have already disrupted the plans of Chinese miners. Bitmain, Canaan, and others rely on TSMC (Taiwan) and Samsung (South Korea) for their ASICs. But if those foundries face capacity constraints due to AI orders, Chinese miners lose access to the most advanced nodes. The Chinese government's 'Big Fund III' is pouring $34 billion into domestic equipment, but the gap in advanced lithography is years away from closing. The result: a bifurcated mining market—those with access to TSMC's 3nm and those stuck on older, less efficient nodes. Yields don't lie. The Goldman report notes that advanced node yields improve with more equipment investment. TSMC's N3 is now above 80% yield. But the cost of achieving that yield is enormous. Each new fab requires billions in WFE, and the depreciation hits the foundry's margins. For crypto miners, this means that the cost of a new ASIC is not just a function of the chip design, but of the entire semiconductor capex cycle. If foundries raise prices to cover their depreciation, ASIC prices rise. We saw this in 2022 when the Bitcoin mining rig market collapsed—not because of hash rate, but because of oversupply and falling chip prices. The next cycle will be different. Core insight: The WFE cycle is a hidden variable in crypto's macroeconomic model. Most analysts focus on coin price, halving events, or energy costs. They ignore the fact that the mining hardware supply chain is a derivative of the global semiconductor industry. When Goldman says WFE spending will grow at 36% CAGR, it implies that the semiconductor industry is absorbing more capital than ever. That capital has to come from somewhere—either from end customers (like AI companies) or from subsidies (like CHIPS Act). Crypto miners, being a smaller buyer, get squeezed. Based on my experience auditing the 2020 DeFi yield arbitrage, I learned that liquidity depth is the primary constraint. The same applies here. The liquidity of ASIC supply is determined by the foundry's capacity allocation. And right now, AI is the 800-pound gorilla. The 2024 ETF liquidity bridge showed that institutional capital flows into Bitcoin are decoupled from on-chain liquidity. Similarly, the demand for advanced chips is decoupled from crypto's needs. The result is a structural deficit. Contrarian angle: The decoupling thesis is wrong. Many in crypto believe that the industry is immune to traditional semiconductor cycles because mining is a niche. But the data suggests otherwise. The WFE spend is a leading indicator for chip availability. If the equipment cycle peaks in 2028, as Goldman suggests, then the next few years will see a steady increase in supply. That could actually be bullish for miners—more capacity means more chips, eventually. But the timing is off. The 2025-2027 period will be a crunch. The contrarian view is that the WFE cycle is actually a bearish signal for the next Bitcoin halving (2028). By the time the halving hits, the supply of new ASICs will be high, competition will be fierce, and the marginal cost of mining will drop. The narrative of 'scarcity' will be replaced by 'efficiency'. Another contrarian insight: The AI hype is a double-edged sword. Goldman's forecast assumes AI demand sustains until 2028. But if the AI investment bubble bursts (as many anticipate), the WFE spend could drop 30-50%. That would free up capacity for crypto miners, leading to a flood of cheap ASICs. That scenario is bullish for hash rate but bearish for mining margins. The smart money is not betting on the direction of coin prices; it's betting on the timing of the next equipment delivery. Takeaway: Watch the equipment delivery timelines. The next Bitcoin halving might be less about price and more about whether TSMC has capacity for 3nm ASICs. We didn't see that coming. Yields don't tell the whole story—the real story is in the order books of ASML and Applied Materials. If you're a miner, your next purchase decision is not just about hash rate per watt; it's about whether the fab can actually deliver. The macro watcher's job is to map these interconnections. The semiconductor cycle is now the crypto cycle.