Jack Dorsey’s Block Inc. spent years telling the market it would disrupt Bitcoin mining. The 3-nanometer Proto chip was supposed to be the weapon. Core Scientific, the largest publicly traded miner at the time, was the launch customer. Then they paid $41.9 million to walk away.
Ignore the hype. Follow the gas. In this case, the gas is the power that flows through a data center — and where that power is now being directed tells you everything about the structural shift underway.
Context: The Failed Promise
Block’s mining chip story started in 2021 with a splashy announcement. Dorsey framed it as a geopolitical move — decentralizing chip supply away from Bitmain’s dominance. The product was a 3nm ASIC, built on TSMC’s advanced node. Core Scientific signed a massive purchase agreement in early 2024, committing to buy enough chips to power 15 exahash. The deal was worth hundreds of millions.
But by early 2025, Core Scientific terminated the contract. The $41.9 million penalty was the largest single exit fee in mining history relative to the deal size. The stated reason: a shift in business priorities. But reading the quarterly filings, the real reason was simple economics. The Proto chips underperformed. Not in hash rate — but in efficiency. The J/TH ratio was never published by Block. Competitors like Bitmain’s S21 and MicroBT’s M60 were already shipping at 19 J/TH while Block’s alleged 3nm design was reportedly struggling to break 25 J/TH. In an industry where 1 J/TH of inefficiency can destroy margins over a two-year cycle, the chip was dead on arrival.
Core Insight: The Liquidity Drain from Mining to AI
Here is the number most analysts miss: Core Scientific did not just cancel the Block order. Within the same quarter, they signed a 15-year, 200-megawatt hosting deal with AMD for AI compute. The projected revenue from that single contract is $14 billion. Compare that to the few hundred million they could have earned mining Bitcoin with 15 EH/s.
This is not a one-off. Marathon Digital and Riot Platforms are quietly negotiating similar deals. The reason is structural. Bitcoin mining is a commodity business with thin margins, driven by hardware efficiency and electricity cost. AI compute hosting is a premium service with long-term contracts and fat margins. Every megawatt that flows to AMD, Nvidia, or Microsoft is a megawatt that is not mining Bitcoin. The data centers are the same. The hardware is different. The capital allocation has shifted.
Bets are cheap; exits are expensive. Core Scientific paid $41.9 million to get out of an uncompetitive hardware deal — but in return, they secured a revenue stream that makes mining look like a side hustle.
Contrarian Angle: The Decoupling Thesis Is Dead
The market narrative for years was that Bitcoin mining would decouple from traditional infrastructure cycles. The argument was that Bitcoin’s fixed supply and global demand would create an asymmetric return profile, independent of AI or cloud trends. That thesis is now false. Miners are not decoupling — they are re-coupling with the broader compute market, because that is where the liquidity is flowing.
Based on my own experience auditing mining hardware during the 2017 boom, I can tell you that efficiency is the only moat. I turned down a $500,000 advisory role from a token project that claimed it would revolutionize ASIC design — because the team had no fabrication experience. Block had the fabrication partner (TSMC) but lacked the deep design expertise that comes from years of iterative improvement. Bitmain and MicroBT have been iterating for a decade. Block tried to jump straight to 3nm. That’s like a new car company trying to build a Formula 1 engine from scratch. It rarely works.
The contrarian insight is that the mining industry is actually healthier because of this pivot. The capital that would have been wasted on an uncompetitive chip is now funding AI infrastructure, which generates real economic value beyond speculative hash. The market is self-correcting.
Takeaway: Cycle Positioning
Where does this leave us? In a bear market for mining hardware, but a bull market for mining infrastructure that can adapt. Retail miners holding older generation S19s are facing negative margins. Institutional players like Core Scientific have already switched lanes. The next phase will see further consolidation — Bitmain and MicroBT will dominate the chip supply, while data centers will increasingly serve dual purposes.
Follow the gas, not the hype. The gas is now flowing to AI. The mining companies that survive will be the ones that pivot. The ones that don’t will be the ones paying exit fees.
