The $41.9 Million Signal: Why Core Scientific Paid to Walk Away from Jack Dorsey’s Mining Dream

Core Scientific just paid $41.9 million to kill a contract with Block. Not a delay. Not a renegotiation. A full termination. For a hardware project led by Jack Dorsey himself, that cash penalty is the loudest signal a customer can send. The story hasn't yet hit mainstream media, but inside the mining ecosystem, it's already reshaping the narrative. The company that once bet big on Block's 3nm chip—the Proto miner, the supposed challenger to Bitmain's throne—decided it was cheaper to walk away than to take delivery. That decision carries weight. It reveals something structural about the Bitcoin mining business in 2026 and the growing gravitational pull of AI infrastructure. Let me rewind the context. Block (formerly Square) announced its mining chip ambitions in 2023, positioning itself as a decentralized alternative to Bitmain and MicroBT. The pitch was simple: Bitcoin's biggest advocate building its own silicon. Jack Dorsey's credibility gave the prototype instant cachet. Core Scientific, one of the largest publicly traded mining operators, signed on as the anchor customer, ordering 15 EH/s worth of hashpower from Block's 3nm design. The deal was supposed to validate Block's entry into the hardware race. But validation never came. Instead, Core Scientific reported a $41.9 million loss from contract termination in its Q4 2025 filings, converting the agreement into an asset impairment. The miner cited a strategic pivot away from Bitcoin mining toward AI data center services. It signed a massive 15-year deal with AMD to lease out its infrastructure, projecting $14 billion in potential revenue. That's not a shift in strategy. That's a wholesale migration of resources. Now let me dive into the core of what this means technically. The 3nm chip Block produced hit 15 EH/s in aggregate—a respectable figure for a first-generation product. But here's the missing data point: no one published the energy efficiency ratio. Not Block. Not Core Scientific. In mining hardware, J/TH is the only metric that matters. A chip that can't beat Bitmain's Antminer S19 series on power efficiency is dead on arrival. Core Scientific's decision to absorb a $41.9 million penalty suggests the chip's real-world performance fell far short of expectations. The s hype around Block's mining hardware was always more about narrative than engineering. Its launch strategy and community management leaned heavily on Dorsey's personal brand, not on proven technical advantage. I've seen this before. Back in 2020, during DeFi Summer, I tracked how protocols with strong founders but weak fundamentals—like those copying Uniswap's code with no liquidity strategy—collapsed as soon as incentives dried up. Block's mining chip is the hardware equivalent. The moment Core Scientific ran the numbers, they realized the cost of deploying those chips in a competitive mining environment would never beat the alternative: selling their power to AMD for AI workloads. The s hype couldn't mask the unit economics. The data reinforces this. MicroBT and Bitmain control nearly 90% of the mining hardware market. Their chips operate at sub-30 J/TH efficiency. New entrants need to hit that threshold to survive. Block's prototype never publicly disclosed its efficiency, and Core Scientific's termination suggests it didn't. The message is clear: building a competitive mining chip isn't about having a famous CEO. It's about relentless engineering against a two-headed monopoly. Now for the contrarian angle. Most coverage will frame this as Block's failure—another nail in the coffin of Jack Dorsey's crypto ambitions alongside Tidal, TBD, Bitchat, and the $200 million in regulatory fines at Cash App. And that's partly true. Block's stock is down 68% over five years. The company's diversification into crypto-related businesses has been a value destruction machine. But the bigger story isn't about Block. It's about the structural shift in Bitcoin mining. Core Scientific's pivot to AI is a microcosm of what's happening across the industry. Power, land, and capital are flowing away from Bitcoin hashpower and toward general computing for AI. The data shows that mining companies with access to cheap electricity and existing infrastructure are increasingly renegotiating their future—not as miners, but as data center operators. Marathon and Riot are exploring similar deals. The s hype around Bitcoin hashrate growth masks a quiet migration. This creates a blind spot for the market. Retail traders still think of miners as pure plays on Bitcoin price. But the smart money is already pricing in a future where mining is a side business, not the core. Core Scientific's $14 billion AI contract dwarfs its mining revenue. If even 20% of the hashrate capacity among public miners shifts to AI, the Bitcoin network's security budget faces a structural headwind. The narrative of 'digital gold' relies on miners being economically incentivized to secure the network. If they can make more money selling compute to AMD, the incentive changes. Takeaway: Watch the next wave of miner earnings reports. The ratio of AI revenue to mining revenue will become the key metric. For Block, the proto-miner is effectively dead—no anchor customer, no path to scale. Jack Dorsey's crypto hardware dream ends with a $41.9 million check. For the mining industry, the real story is just beginning. Is Bitcoin still the most efficient monetization of electricity? Or is it becoming a legacy floor for compute that yields to higher-paying tenants? The data says the latter. And the narrative is only starting to catch up.