ASML just dropped its Q4 2024 earnings. The numbers are ugly in the best way. Net bookings surged to €7.1 billion—nearly double expectations. EUV orders alone hit €4.8 billion. The stock ripped 10% in hours. But here’s the kicker: ASML’s CEO said they’re still capacity-constrained until 2026.
And you think Bitcoin mining is just about energy costs?
That wafer fabrication line is the real bottleneck. ASML’s machines print the brains of every AI accelerator and every mining ASIC. When the Dutch monopoly on EUV lithography sneezes, the entire crypto hashrate catches a cold.
Context: ASML is the only company on Earth that builds extreme ultraviolet lithography tools. Without EUV, you cannot produce chips below 7nm. That includes NVIDIA’s H100/B200, AMD’s MI300, and every modern Bitcoin mining ASIC from Bitmain, MicroBT, and Canaan. The last three years of mining profitability have been shaped by a gentle truce between AI giants and miners—both fighting for the same limited wafer capacity at TSMC and Samsung. ASML’s order book is the map of that war.
Now, ASML sees a 30% jump in 2025 revenue guidance. Their EUV shipments are expected to hit 70 units next year, up from 55 in 2024. That means more wafer starts for both AI chips and mining ASICs. But here’s the catch: the new capacity is overwhelmingly tilted toward 3nm and 2nm nodes. Bitcoin mining ASICs still live on 7nm and 5nm—older nodes where capacity is shrinking as fabs migrate to advanced nodes. So while total wafer output grows, the slice for mining might not expand proportionally.
Let’s dig into the data. Based on my audit experience at a DeFi lending protocol during 2022’s bear market, I learned to read supply chains the way others read smart contracts. ASML’s current backlog is over €40 billion. That’s five years of revenue at current rates. The lead time for a new EUV tool is 18-24 months. So the machines being shipped today were ordered when Bitcoin was at $30K. The machines miners need for 2026 are being ordered now, while Bitcoin is at $100K. That means the hashrate growth we’ll see in 2025 is already locked in by decisions made in 2023—low prices, cautious fab spending.
The contrarian angle: Everyone is celebrating ASML’s boom as a sign of AI mania. But I see a different pattern. The same ASML tools that enable NVIDIA to double its chip supply also enable mining hardware manufacturers to refresh their product lines. Bitmain’s Antminer S21, with its 200 TH/s at 15 J/TH, relies on TSMC’s 5nm process. That node’s capacity is being cannibalized by 3nm AI chips. ASML’s higher output of high-NA EUV (for 3nm) doesn’t help miners directly—they need legacy 5nm capacity. This is the paradox: more total wafers, but fewer available for mining because the industry is shifting to silicon that miners can’t afford to use. A Bitcoin miner buying 3nm ASICs would be like buying a Ferrari to deliver pizza.
But there’s a second-order effect. As AI chip production ramps, the cost per transistor on older nodes drops because the fixed cost of EUV depreciation is spread over more wafers. TSMC’s 5nm wafer price might fall 15% by 2026 as 3nm takes the premium. That directly reduces the bill of materials for mining ASICs, improving miner margins. I’ve seen this before: in 2017, when TSMC moved to 10nm for mobile processors, 28nm capacity became cheap—and that’s when the first big Bitcoin mining wave with 16nm ASICs exploded.
Now, the vulnerability. ASML’s export controls on China are tightening. The latest US-Dutch agreement restricts service and spare parts for older DUV tools shipped to China. Chinese mining hardware makers (Canaan, Ebang) rely heavily on domestic fabs like SMIC, which struggles to produce competitive 7nm ASICs without EUV. If SMIC loses access to even DUV service, their mining chip yields will collapse. That pushes the entire mining ASIC market toward Taiwan, giving Bitmain even more pricing power. Decentralization in mining hardware supply? A myth. True ownership begins where the server ends—but the server is a TSMC fab in Taichung.
I’ve been skeptical of the narrative that AI and crypto are natural allies. They are rivals for the same physical resources. ASML’s record orders confirm that AI has won the first round—it gets first dibs on the newest, fastest fabs. Miners get the leftovers. But leftovers from a feast are still a feast. The ASIC makers are already signaling higher production in H2 2025, thanks to ASML’s expanded tool shipments to TSMC. The hashrate will climb, but the cost per terahash will fall more slowly than in previous cycles.
The takeaway? Don’t read ASML’s earnings as a blanket bullish signal for crypto. Read it as a map of silicon warfare. The battle lines are drawn: 3nm for AI, 5nm for mining, and a shrinking middle. The next bull run in mining will be driven not by price, but by the sheer volume of wafers that EUV machines can print. And that volume is about to explode—with a delay. The ASML orders of 2023 are the hashrate of 2025. The orders of 2025 will be the hashrate of 2027. Debate is the compiler for better consensus: and the consensus is that mining hardware scarcity is a feature, not a bug.
So keep your eyes on the lithography. Not the price charts. That’s where the real proof-of-work happens.
Postscript: I’ll be watching TSMC’s next Capital Expenditure update. If they raise their 2025 CapEx above $35 billion, miners should expect a supply glut by early 2027. If they hold, the squeeze continues. Either way, ASML has already cashed the check.
True ownership begins where the server ends. And the server starts at a Dutch cleanroom.
