Hashrate hit a new all-time high last week—over 650 EH/s—while miner revenue per exahash dropped to its lowest since the 2020 halving. The narrative is familiar: network security is robust, difficulty adjusts, the market finds equilibrium. But look closer. The physical layer—the silicon—tells a different story. Every single one of those ASICs, from Antminers to Whatsminers, was printed using a machine that only one company on Earth can build: ASML. And ASML just increased their revenue forecast and announced capacity expansion. That’s not just a semiconductor story. It’s a Bitcoin centralization story that nobody in crypto wants to talk about.
Let’s be clear. ASML doesn‘t make chips. They make the lithography systems that make the chips. Specifically, their extreme ultraviolet (EUV) machines are the only tools capable of patterning the 7nm, 5nm, and now 3nm nodes used in the most efficient Bitcoin mining ASICs. Without an ASML Twinscan NXE:3400 or an EXE:5000 high-NA EUV, you cannot manufacture a chip with the transistor density required to stay competitive in the mining arms race. The gatekeepers of Bitcoin’s physical security are not the core developers in chat rooms—they are a Dutch company with a 100% monopoly on a single piece of capital equipment.
The numbers are stark. The Bitmain Antminer S21 XP, for example, runs on a 5nm node. Achieving 5nm requires multiple EUV passes. Each EUV machine costs roughly €200 million and takes 12–18 months to build. ASML shipped only 60 EUV units in 2023, and plans to ramp to 90+ by 2026. That entire supply is split among TSMC, Samsung, and Intel—the three fabs that produce logic chips for everything from iPhones to AI accelerators to Bitcoin ASICs. Mining chip designers like Bitmain (which uses TSMC) and MicroBT (which uses Samsung) are at the back of the line. They compete for fab capacity with Apple and NVIDIA. In my experience auditing the firmware of the Antminer S21, I saw the efficiency curves directly tied to the number of EUV layers. Every layer adds cost and lead time. The bottleneck isn’t design talent; it’s ASML’s order book.
Code does not lie, but it often forgets to breathe. The Bitcoin protocol assumes a decentralized distribution of hash power. But the manufacturing of that hash power is concentrated in three fabs that depend on a single supplier. That’s not a theoretical risk—it’s a structural dependency. When ASML announced its expansion plans, the market cheered. More EUV capacity means more advanced chips, lower costs, higher efficiency. For the crypto mining industry, it means the race to 3nm ASICs can accelerate. But here’s the contrarian angle: this expansion will not democratize mining. It will entrench the incumbents. The companies that already have long-term wafer allocation agreements with TSMC and Samsung—those with the balance sheets to prepay for EUV capacity—are the same ones that dominate the mining pool hashpower today. Bitmain’s Antpool, for example, controls over 20% of the global hashrate. The geographical concentration is even worse: most ASML machines are installed in Taiwan and South Korea. If geopolitical tensions disrupt those fabs, the entire Bitcoin network’s ability to produce new miners halts.
Gas wars are just ego masquerading as utility. Replace “gas wars” with “hashrate wars” and the principle holds. The real war is for access to EUV light. ASML’s monopoly is not just a business strength; it’s a single point of failure for the most decentralized network in existence. The Chinese mining hardware makers—Canaan, MicroBT, and increasingly Bitmain’s Shenzhen operations—are already feeling the squeeze from US–China export controls. The Dutch government, under US pressure, restricted the sale of ASML’s most advanced EUV machines to China. That means Chinese ASIC manufacturers cannot get the latest 3nm nodes unless they buy from non-Chinese fabs, which themselves are rationed. The result is a bifurcation: Western fabs produce the highest-efficiency miners, while Chinese fabs are stuck on 7nm or older nodes. This directly translates to a geographic shift in hashrate. Over the next five years, I expect the share of Bitcoin hash power located in jurisdictions friendly to the US to increase from roughly 40% to over 70%, not because of policy, but because of ASML delivery schedules.
Based on my reverse-engineering of mining pool payout structures and my audit of ASIC controller firmware, I’ve seen the code assume a uniform distribution of miner efficiency. But the hardware reality is anything but uniform. The difference between a 5nm ASIC and a 7nm ASIC is roughly 30% higher efficiency. That difference compounds over a 36-month mining lifecycle. The miners with access to the newest EUV-printed chips will always have a 30% cost advantage. Over three years, that advantage translates into the ability to survive two halvings while competitors shut down. The network’s hash power becomes increasingly concentrated in the hands of those who can secure EUV allocation. The code doesn’t care about that—it only sees the sum of hash power. But the community should care. Satoshi’s vision of one-CPU-one-vote is long gone. We are now at the stage where one-ASML-one-vote is closer to the truth.
What does this mean for the average Bitcoin holder? The next difficulty adjustment cycle will not be driven by price alone. It will be driven by the physical capacity of ASML’s Veldhoven factory. If ASML fails to hit its 2026 target of 90 EUV units per year—due to supply chain constraints, labor shortages, or export control changes—the replacement rate of old, inefficient miners slows. The network hash rate growth stalls, and the security level of Bitcoin becomes capped by a capital equipment supplier. Conversely, if ASML succeeds, we see a wave of ultra-efficient 3nm ASICs that could push hash rate past 1,000 EH/s by 2028, but with the vast majority of that power owned by three or four entities.
The fourth halving already showed us the beginning of this trend. Miner revenue collapsed not just because of the block subsidy cut, but because the cost of producing a Bitcoin in fiat terms became dominated by the cost of the ASIC itself. That cost is directly proportional to the cost of the EUV machine that printed it. The economics of Bitcoin mining have become a derivative of semiconductor capital expenditure. And like any derivative, it contains hidden convexity. The contrarian bet here is not against Bitcoin—it’s against the assumption that mining hardware can remain a fungible, open market. It will increasingly become a closed, vertically integrated supply chain.
In my work auditing mining contracts for a large mining pool, I encountered a clause that forced miners to use a specific brand of ASIC for the duration of the contract. When I asked why, the operator said: “Because we only trust machines from fabs that have ASML high-NA EUV allocation.” That statement should terrify anyone who believes in decentralized consensus. The code is not the final arbiter of security—the wafer stepper is.
Takeaway: The next Bitcoin halving will not be decided by monetary policy or hard cap. It will be decided by whether ASML can deliver enough EXE:5000 units to keep the hashrate growing. If you are building a mining operation today, your strategic risk is not the market price of BTC—it’s the lead time on an EUV order that a Dutch company may or may not fulfill. Code does not lie, but it often forgets to breathe. The breath of the Bitcoin network now flows through a very narrow pipe in Eindhoven. Are we comfortable with that?
