The semiconductor equipment cycle is the quiet engine beneath every AI token pump, every mining rig shortage, and every narrative about decentralized compute. Goldman Sachs just dropped a chart that predicts global wafer fab equipment (WFE) spending will hit $2.81 trillion by 2028 — a 36% CAGR from 2025. The crypto market is too busy chasing the next memecoin to read it. But I’ve been reverse-engineering these numbers since 2017, when I watched the Parity wallet bleed 150,000 ETH and realized that hardware liquidity is just code, digitized and leveraged. This article is my pre-mortem on that forecast.
Context: The WFE Engine and the Crypto Supply Chain
Wafer fab equipment is the multi-billion-dollar machinery that etches, deposits, and measures the silicon wafers that become every chip in your phone, your GPU, and your ASIC miner. The industry is cyclical — boom-bust cycles of 3-4 years driven by demand for smartphones, PCs, and now AI. But the current cycle is different. Goldman’s forecast sees WFE spending accelerating from about $1,000 billion in 2025 to $2,810 billion in 2028, driven almost entirely by HBM (high-bandwidth memory) for AI and advanced logic nodes below 5nm. This is the infrastructure that makes AI training possible. And AI training is the new driver of crypto mining — not just for Bitcoin, but for the entire proof-of-work and proof-of-stake ecosystem that now competes for GPU time.
But here’s the catch: the crypto industry doesn’t own any fabs. We rent capacity from the same TSMC, Samsung, and Intel that serve NVIDIA, Apple, and AMD. When AI demand surges, our mining hardware gets pushed to the back of the queue. The 2020-2021 GPU shortage was a preview. The 2024-2025 HBM shortage is the sequel. Goldman’s forecast tells us that the squeeze will last until at least 2028, because the equipment that builds the fabs is itself in short supply.
Core: The Seven Dimensions of the WFE Cycle and Their Crypto Implications
I’ve broken down the Goldman report into seven dimensions, each with a crypto-specific signal. This is the analysis I run on every protocol I audit — trace the dependencies, map the constraints, and find the hidden leverage points.
1. Technology: The GAA and High-NA EUV Inflection
Goldman’s forecast assumes that High-NA EUV lithography (ASML’s EXE:5200 series) will enter volume production in 2026-2027. This is the only way to achieve 2nm and below transistor densities. But each High-NA EUV machine costs €300-400 million and only 50-60 units can be built per year. That means the number of advanced logic fabs that can be built is physically limited. For crypto, this translates to a cap on the number of AI ASICs (like the ones being developed by new Bitcoin mining players) that can be manufactured. The supply of next-gen mining hardware is not just a function of demand — it’s a function of ASML’s crystalline silicon lens production.
I’ve been tracking this since 2022, when I reverse-engineered the supply chain for the Bitmain Antminer S19 XP. The 5nm chip in that miner depends on TSMC’s N5 capacity, which itself depends on EUV tools. The cycle is self-referential. Goldman’s 2028 peak implies that the hardware supply for AI and mining will remain tight until at least 2027, when the current batch of fabs (like TSMC’s Arizona Fab 21 Phase 2) start mass production. After that, we might see a glut of chips — and a crash in mining profitability.
2. Supply Chain: The 100% Dependency on a Single Dutch Company
Every advanced chip in the world passes through an ASML lithography machine. There is no substitute. The supply chain for crypto mining hardware is therefore entirely dependent on the political stability of the Netherlands and the export control regimes of the US. Goldman’s forecast assumes no major disruption to this supply chain. But the experience of Chinese fabs since 2022 shows that the US can and will cut off access to advanced equipment. If the US restricts High-NA EUV exports to any country that hosts crypto mining operations (like Kazakhstan or the US itself), the entire mining hardware supply chain could be severed.
I learned this lesson during the 2022 Terra collapse, when I saw how a single point of failure (the Anchor protocol) could cascade. The WFE supply chain is the Anchor protocol of the semiconductor industry. One geopolitical shock, and the entire yield curve inverts.
3. Capex: The $500 Billion HBM Bet
Goldman’s forecast gives HBM/DRAM the first growth driver slot. HBM is the memory stack that sits next to AI GPUs. It’s made by SK Hynix, Samsung, and Micron, and it requires both advanced DRAM fabs (EUV needed) and advanced packaging (TSV, hybrid bonding). The amount of WFE spending per HBM gigabyte is 3-4x higher than for standard DRAM. This is the hidden lever: every AI token that claims to be “decentralized compute” is actually competing for the same HBM capacity that NVIDIA needs for its B200 chips. The market cap of those tokens is a bet on HBM supply, not on the protocol itself.
I’ve been running a small copy-trading strategy on this thesis since 2024: short AI tokens that have no hardware backing, long the equipment suppliers that are the “picks and shovels” of the AI boom. The numbers are clear: the 2026-2028 WFE cycle implies that HBM supply will remain tight until 2028, which means the price of HBM (and thus the cost of AI compute) will stay high. That’s a tailwind for protocols that own their own hardware (like Bitcoin mining farms) and a headwind for those that rely on rented cloud GPU.
4. Demand: The AI Capital Expenditure Bubble
Goldman’s forecast assumes that AI capital expenditure (capex) by hyperscalers (Amazon, Microsoft, Google, Meta) will grow at 40%+ per year through 2027. This is the core assumption. If AI capex slows, the WFE forecast collapses. The contrarian view I hold is that AI capex is a lagging indicator of actual revenue. The hyperscalers are spending billions on GPUs today, but the revenue from AI services is still a fraction of their total. If the AI bubble pops (like the dot-com bubble did), the WFE spending will follow suit. The crypto market is already pricing in perpetual AI growth, but the semiconductor equipment cycle is a real-economy constraint that will eventually force a reckoning.
I’ve seen this before. In 2020, I deployed $50,000 into Uniswap V2 liquidity pools, chasing yield that turned out to be a deceptive incentive for risk. The AI capex surge is the same — it’s a yield that looks good on paper but hides the risk of a sudden stop. If hyperscalers cut their capex by 20% in 2027, the WFE spending could drop by 30-40%, triggering a collapse in hardware prices and a flood of used GPUs onto the market. That would be a golden opportunity for crypto miners to buy cheap hardware, but only if they have cash reserves.
5. Geopolitics: The Taiwan Scenario
Goldman’s forecast is based on the assumption that the geopolitical status quo holds. But the US CHIPS Act, the EU Chips Act, and the Japanese semiconductor revival plan are all trying to build redundancy. The problem is that 90% of advanced logic is still made in Taiwan. If the Taiwan strait becomes a flashpoint, the entire global semiconductor supply chain stops. For crypto, this means the end of new mining hardware for months or years. The price of Bitcoin would likely spike on the supply shock, but the network hash rate would freeze.
I’ve been stress-testing this scenario since 2022. The US government has classified Bitcoin mining as a strategic industry, but it hasn’t invested in domestic fab capacity. The $52 billion CHIPS Act is a drop in the ocean compared to the $2 trillion needed to replicate TSMC’s capacity. The WFE forecast is a best-case scenario, not a worst-case one.
6. Competition: The “Picks and Shovels” Monopoly
The equipment industry is dominated by five companies: ASML, Applied Materials, LAM Research, KLA, and Tokyo Electron. They have pricing power because they are the only game in town. Their gross margins are 45-60%, and they are raising prices. This means that every new fab costs more than the last one, which drives up the cost of every chip — including ASICs for mining. The result is that the barrier to entry for new mining hardware manufacturers is rising. Only the largest (Bitmain, MicroBT, Canaan) can afford to pre-order equipment years in advance. Small players are locked out.
This is a structural shift. In 2017, anyone could design a Bitcoin ASIC and get it made at a Chinese foundry. Today, the foundry capacity is controlled by firms that prioritize AI chips over crypto mining chips. The WFE cycle is confirming that this trend will continue through 2028. The mining hardware market will become more concentrated, which means centralization of hash rate. That’s a security risk for Bitcoin.
7. Valuation: The Equipment Sector as a Crypto Hedge
Goldman’s forecast implies that the equipment sector will see revenue growth of 25-35% annually for the next three years. The current P/E multiples are 20-30x, which is high but not unreasonable if the growth materializes. For crypto investors, this is a hedge. If you believe in the AI narrative (and the crypto tokens that piggyback on it), you should also own a position in ASML or Applied Materials. They are the real “decentralized compute” — they don’t depend on token prices, they depend on physics.
I’ve been practicing this since 2024, when my ETF arbitrage strategy taught me that institutional capital follows infrastructure, not hype. The $12,000 I made from that strategy was boring, but it was repeatable. The same principle applies here: the WFE forecast is the most boring, most important chart in crypto. It tells you that the hardware scarcity will persist, and that the only way to profit from it is to own the suppliers, not the users.
Contrarian Angle: The Retail Herd Is Sleeping on the Wrong Inflection
Every crypto Twitter influencer is talking about the next AI token that will 100x. But they are ignoring the fact that the underlying hardware is becoming more expensive, not cheaper. The retail herd is buying the narrative that “AI will democratize compute,” but the reality is that the semiconductor equipment cycle is consolidating power in the hands of a few oligopolists. The contrarian trade is to short the tokens that are priced for perpetual hardware abundance and go long the equipment suppliers that benefit from perpetual scarcity.
But there’s a deeper contrarian insight: the WFE cycle itself is a cyclical peak. Goldman’s forecast shows growth slowing from 45% in 2027 to 29% in 2028. That’s a deceleration signal. If the trend continues, 2029 could see a decline. The smart money should be preparing for the bust, not the boom. The crypto market has a history of piling into the last leg of a cycle (like the 2021 NFT mania). The WFE cycle suggests that the last leg of the AI hardware cycle will be in 2027-2028, and then the hangover will come.
I’ve seen this movie before. In 2020, I rode the Uniswap yield wave until it broke my board. The WFE cycle is the same: the yield looks great until the liquidity dries up. The question is not whether the cycle will turn, but whether you have an exit plan.
Takeaway: Actionable Price Levels and a Forward-Looking Thought
I’m not going to give you a target price for Bitcoin or any token. Instead, I’ll give you a framework. Monitor the WFE spending data from Goldman, SEMI, and the equipment suppliers themselves. If the quarterly WFE orders start to miss, that’s a lead indicator for a crypto hardware glut 12-18 months later. If they beat, the scarcity continues. The key level to watch is the 2027 WFE forecast of $2.18 trillion. If that number is revised down, start selling your mining rig holdings. If it’s revised up, buy the equipment suppliers.
For the crypto community, the most important takeaway is that the semiconductor equipment cycle is the ultimate “pre-mortem” for the AI token narrative. The technology is real, but the economics are subject to the same boom-bust dynamics that govern every commodity. The code might be trustless, but the silicon is not.
We mined liquidity while the code slept. We rode the wave until it broke our boards. Liquidity is just trust, digitized and leveraged. The WFE forecast is the ledger of that trust. Read it before the market does.