Hook
What if the most telling signal about Bitcoin mining’s future isn’t a hash rate all-time high, but a $41.9 million termination fee? That’s exactly what Core Scientific paid to walk away from Block’s much-hyped Proto mining chip—a 3nm, 15 Exahash monster that was supposed to challenge Bitmain’s throne. Instead, it became the most expensive dead end in Jack Dorsey’s crypto empire.
I’ve seen this movie before, and it never ends well. When a sole customer—a company that should have been the ultimate beta tester—chooses to swallow nearly $42 million in losses rather than take delivery of your chips, the story isn’t about contract disputes. It’s about a fundamental failure of technical promise and market reality.
But here’s the kicker: Core Scientific didn’t just cancel; it pivoted. Hard. The same week it ripped up the Block order, it signed a 15-year, $14 billion revenue contract with AMD to rent out its data centers for AI workloads. This isn’t a coincidence. It’s a structural narrative shift masking itself as a corporate reshuffle.
Context
Let’s rewind. Block—formerly Square—announced its Proto mining chip in early 2025, positioning it as a democratizing force. Jack Dorsey, ever the Bitcoin maximalist, claimed it would break Bitmain and MicroBT’s duopoly. The chip was 3nm—leading-edge for mining—and promised 15 EH/s of raw compute. Core Scientific, a major public miner (ticker: CORZ at the time), was the launch customer, pre-ordering enough units to make Block’s mining division look viable.
But by early 2026, the fairy tale soured. Core Scientific disclosed a $41.9 million impairment charge related to the contract’s termination. The official reason: “strategic pivot.” The real reason: the chips likely underperformed on efficiency (joules per terahash) relative to Bitmain’s S19 XP or MicroBT’s M60 series. Block had never released independent benchmark data—a red flag I spotted immediately. In my experience analyzing hardware specs during the 2017 ICO frenzy, secrecy equals weakness.
Now, Core Scientific is no longer a miner; it’s an AI infrastructure play. Its Austin, Texas facilities, once humming with ASICs, will host AMD’s GPUs for large language model training. The $14 billion figure is a projection, not a guarantee, but the direction is clear: Bitcoin mining’s best and brightest are swapping picks for shovels in the AI gold rush.
Core Insight
This isn’t just one bad contract—it’s a diagnostic of two collapsing narratives.
First: The myth of mining hardware disruption. Block thought brand loyalty and Jack’s charisma could dethrone incumbents with decades of supply chain mastery. It couldn’t. Bitcoin ASIC design is a physics game, not a marketing game. Block’s chip was a micro-iteration on existing tech, not a breakthrough. Without a cost advantage or superior energy efficiency, it was doomed. The $41.9 million impairment is the market’s way of saying: “You don’t have a product.”
Second: The resource drain from mining to AI. Core Scientific’s pivot is a canary in the coal mine—literally. The same electricity, cooling, and land that powered SHA-256 hashing now serves CUDA cores. This isn’t new; I flagged this in my 2024 piece, “The Algorithmic Herd,” predicting that AI agents would eventually bid up compute costs. But the speed is shocking. Core’s exit from Block’s order accelerated the shift by at least a year.
Let’s quantify this. A standard Bitcoin miner like Core generates roughly $0.10–$0.15 per kWh in profit after all costs. AI HPC leasing can fetch $0.50–$1.00 per kWh. That’s a 3x to 7x premium. When you’re a public company answering to shareholders, the choice is math, not ideology. The market is a story, and the best storytellers win—but right now, AI writes a better story.
I embedded this data in my recent analysis: over the past 6 months, five of the top ten public miners have announced AI partnerships. Only two (Core and one other) signed contracts significant enough to move their revenue mix above 30%. The rest are still talking. Core acted.
The Sentiment Trap
Here’s the contrarian take you won’t read on CoinDesk: The pivot to AI might be a dead cat bounce for miners. A dead cat bounce is still a bounce, but it doesn’t change the cat’s terminal condition.
Yes, Core Scientific locked in a 15-year deal. But that deal is contingent on AMD and the broader AI market continuing to expand at 30%+ CAGR. If AI investment slows—due to regulatory clampdowns, energy shortages, or a new model efficiency breakthrough—those data centers become stranded assets. Miners who sold their ASICs at a loss to build GPU farms could then face a double whammy: no mining revenue and empty racks.
Moreover, the mining hardware market itself isn’t dying; it’s consolidating. Block’s exit leaves only Bitmain and MicroBT as viable ASIC suppliers. That duopoly will squeeze margins for miners who stick with Bitcoin. The next halving in 2028 will force many to close shop. The survivors will be those with access to cheap energy and fat AI contracts—effectively becoming hybrid data centers, not miners.
I’m reminded of the 2017 ICO boom: everyone pivoted to “utility tokens” when the market crashed. Most died. This pivot is different in scale but identical in logic—chasing the hot narrative. If you’re long Core’s stock, you’re betting on AI’s permanence, not Bitcoin’s.
Takeaway
Block’s mining chip failure is a tombstone for the “decentralized hardware” dream. Core’s pivot is a blueprint for survival, but it comes with its own risks. The next chapter in Bitcoin mining won’t be about better ASICs; it will be about capital allocation—choosing which digital economy to serve.
Is the industry ready for a future where Bitcoin’s hash rate grows slower while AI compute demand explodes? Or will the two narratives collide, leaving only the most agile left standing? The $41.9 million question is now ours to answer.