Core Scientific just paid $41.9 million to cancel an order for Jack Dorsey's shiny new 3nm mining chips.
That's not a contract termination. That's a technical knockout.
The news broke last quarter. But the real story isn't the loss. It's what the loss tells us about the trajectory of Bitcoin mining, the illusion of hardware innovation, and the quiet migration of capital away from digital gold toward something that actually pays the bills.
I've been on the wrong side of these trades before. In 2017, I shorted ICO tokens that everyone thought would moon. Made 40% in three weeks. The lesson? The market doesn't care about your whitepaper. It cares about cold, hard P&L.
Context: The Promise and the Premise
Block (formerly Square) entered the Bitcoin mining chip game in 2022. Jack Dorsey's vision was romantic: build an open, decentralized mining system. The 3nm ASIC chip was supposed to be the cornerstone. Core Scientific, one of the largest public miners, signed on as the anchor customer.
Fast forward to 2026. Core Scientific walks away. Pays a $41.9 million penalty. Then signs a 15-year deal with AMD to host AI compute. The same land, power, and infrastructure that was supposed to mine Bitcoin now serves generative AI workloads. Smart money doesn't hold onto losing positions. It cuts and rotates.
Yield is the rent you pay for holding someone else's bags. In this case, the rent was $41.9 million.
Core: The Anatomy of a Failure
Let's break down the numbers. Core Scientific's contract with Block had a total value of roughly $230 million. The termination fee represents about 18% of the original order. That's a haircut, not a total loss. But the fact that they chose to pay it rather than receive the chips tells you everything.
A 3nm chip is the cutting edge. But cutting edge doesn't mean profitable. The chip's energy efficiency (J/TH) is the only metric that matters in mining. Block never published third-party benchmarks. Core Scientific's own engineers probably ran their own tests and found the chip didn't compete with Bitmain's current-gen machines.
From my own experience running automated trading strategies in 2020, I learned that when the gas fees eat into your yield, you pull the plug. Core Scientific pulled the plug. They ran the numbers and realized that deploying those chips would generate negative net present value. The $41.9 million was cheaper than the eventual loss.
But there's a deeper layer. Core Scientific didn't just cancel Block's order. They pivoted their entire business model. They took the same real estate—land with cheap power, fiber, and cooling—and rented it to AMD for AI training. The revenue projections? $14 billion over 15 years. That's the difference between mining a volatile asset and renting out compute to the fastest-growing industry in the world.
We don't argue with the P&L. The P&L is the truth.
Contrarian: Retail Thinks This Is a Setback. Smart Money Sees a Structural Shift.
Mainstream crypto Twitter will frame this as "Block stumbles in mining" or "Core Scientific dumps Jack." That's surface-level. The real story is that Bitcoin mining is losing its competitive advantage for resources.
Mining is an energy arbitrage game. You find cheap power, build a facility, plug in ASICs, and sell the Bitcoin. The problem? AI compute demand is exploding. Hyperscalers like Google, Microsoft, and Amazon are desperate for data center capacity. They pay 5x what a miner can pay for the same power. The result: miners are converting into AI hosting companies.
Core Scientific is the canary in the coal mine. Its stock (CORZ) has rallied since the pivot. Block's stock (SQ) is down 68% over five years. Jack Dorsey's crypto bets—TBD, Tidal, Bitchat, Proto chips—have all either failed or been wound down. The only thing that worked was Cash App, which now faces a $200 million CFPB fine.
The narrative that Bitcoin is the only digital asset worth building around is cracking. The infrastructure layer is voting with its feet.
Takeaway: The Playbook for the Next Cycle
The mining industry is splitting into two camps: the pure-play miners who will struggle in the next bear, and the hybrid operators who pivot to AI/Crypto infrastructure. Core Scientific is the template. Block's proto chip is the cautionary tale.
If you're allocating capital to crypto miners, stop looking at hash rate. Look at power purchase agreements and downstream contracts with cloud providers. The smart money is already stacking AI data center REITs.
What are you doing?
