Tracing the ghost in the gas receipts – but this time, the ghost isn’t a smart contract bug. It’s a $41.9 million termination fee.
Core Scientific walked away from its order for Block’s 3nm mining chips in late 2025. They paid the penalty. They walked. The chart said everything is fine. The quarterly filing said someone is burning cash to hide a body.
I’ve been staring at on-chain data for nine years. I’ve seen hype die before. But watching a company with a $40 billion market cap abandon its own silicon ambition? That’s a new kind of silence.
Context
Block – Jack Dorsey’s financial services company – dove into mining chip design in 2021. The vision: build open-source, decentralized mining hardware that could break Bitmain’s grip on the industry. By 2024, they had a 3nm chip. They called it Proto. They found one anchor customer: Core Scientific, one of the largest publicly traded miners in North America.
Core Scientific signed a contract to buy 15 exahash of Proto chips. Then they terminated it. The termination cost them $41.9 million in cash. That’s not a breakup fee. That’s a ransom paid to avoid a bad marriage.
Core Scientific’s CEO stated the reason: “strategic shift” toward AI and high-performance computing. They are now leasing their data centers to AMD for a 15-year contract that could yield $140 billion in revenue. They are not abandoning mining entirely – they still run some hash – but they are voting with their feet. And their feet are running toward AI.
Core
Let’s look at the numbers that don’t make headlines.
First, the chip performance. Block never published a benchmark for Proto beyond the 3nm label. In the mining world, 3nm is a process node, not a performance metric. The real metric is joules per terahash (J/TH). Bitmain’s flagship S21 series delivers around 12 J/TH. MicroBT’s M60 series is similar. If Proto couldn’t beat that, there was no reason to buy it – especially when Bitmain offers volume discounts and established reliability.
Core Scientific’s termination letter (disclosed in their Q4 2025 10-K) gave no technical reason. But the timing tells a story. They signed the contract in early 2024. They started receiving chips in late 2024. By mid-2025, they stopped. That’s a 12-month window to test the chips in the field. The test failed.
Second, the on-chain evidence of a weakening mining economy. Bitcoin hashrate hit 800 EH/s in early 2025. But miner revenue per hash is declining. The April 2024 halving cut block rewards to 3.125 BTC. Transaction fees – boosted by Ordinals – provided a temporary cushion, but by late 2025, fee revenue had normalized to under 5% of total miner revenue. The cost of electricity, hardware, and cooling is rising. The margin is thin.
I remember the 2020 DeFi Summer. I sank $50,000 of my own ETH into Uniswap and SushiSwap to test yield volatility. I watched the pools. I felt the impermanent loss. That taught me that liquidity speaks louder than tweets. Now, Core Scientific is saying that AI liquidity – the guaranteed revenue from AMD – speaks louder than Bitcoin mining margins.
Let’s graph the shift. In 2023, Core Scientific generated 100% of its revenue from Bitcoin mining. By Q4 2025, mining revenue was down to 40%. The rest came from AI compute. Their stock price (CORZ) has doubled since the AMD deal announcement. Block’s stock (SQ) is down 68% from its 2021 peak.
Third, the ecosystem signal. Block’s broader crypto strategy is unraveling. Tidal (music) was written down. TBD (decentralized identity) was shuttered. Bitkey (self-custody wallet) saw low adoption. The Cash App regulatory fine – $80 million to CFPB and $150 million to state regulators – is a separate but concurrent wound. Jack Dorsey’s vision of a decentralized future is costing the company real money without delivering results.
The mining chip failure is not an isolated incident. It is the culmination of a pattern: Block’s hardware division lacked the deep engineering expertise of Bitmain or MicroBT. They were trying to enter a game where the incumbents have 20 years of supply chain optimization. And they lost their only customer.
Hunting liquidity where the charts lie – the charts show a growing Bitcoin hashrate. They don’t show that the growth is slowing. If Core Scientific is a bellwether, other miners will follow. Riot Platforms has already announced a pilot AI compute initiative. Marathon Digital is exploring ASIC retirement plans. The narrative that “mining is just energy arbitrage” is becoming a self-fulfilling prophecy.
Contrarian
But let’s play devil’s advocate. The AI hype cycle is real. But it’s also fragile. If AMD’s GPU business stumbles – if AI demand softens – Core Scientific’s 15-year contract could become a liability. The termination clause for that deal isn’t public, but if it’s anything like the Block contract, there’s a multi-million dollar exit fee. Risk cuts both ways.
Moreover, the “miners to AI” narrative might be overblown. According to data from HashrateIndex, only about 5% of total mining capacity has pivoted to AI as of early 2026. Core Scientific is an outlier, not the average. Most miners are still running S19s and S21s, waiting for the next bull run. The majority of hash will remain on Bitcoin as long as the price stays above $70,000 – which it has, barely, in early 2026.
The contrarian bet is that Block’s chip failure is a one-off, not a trend. Bitmain and MicroBT are not going anywhere. The real story is that Block overestimated its ability to disrupt a mature industry. That’s a company-specific failure, not a sector collapse.
Still, I believe the data. The $41.9 million exit fee is a hard number. It’s a price paid to avoid a larger loss. That’s not a signal you ignore.
Takeaway
Decoding the pixelated intent behind the PFP – or in this case, behind the ASIC. Block’s mining chip venture is effectively dead. Core Scientific has found a better business model. The question that keeps me up at night: if mining’s profitability ends, who secures the Bitcoin chain?
The answer right now is: the same oligopoly that always has. Bitmain and MicroBT will continue to produce chips. Miners will continue to run them. But the margin squeeze is real. The next bear market could be brutal. And if AI demand stays hot, we might see a permanent migration of hashpower away from Bitcoin.
That’s not a death knell. It’s a reckoning.
Reading the pulse in the pool balance – the pool balances of miners haven’t changed much. But the hearts behind the numbers are beating faster. Core Scientific just showed us the exit door. Others will follow.
Watch for more termination announcements. Watch for falling hashrate growth. And watch Block’s next quarterly earnings. If they write down the Proto inventory, you’ll know the ghost has finally left the machine.