Hook
$41.9 million. That is the price Core Scientific paid to walk away from Block's 3nm Proto mining chips. Not a delay. Not a renegotiation. A clean break. In the language of capital markets, that is a tax on unverified assumptions—a signal that the underlying technology failed its first real-world stress test. Volatility is the tax on unverified assumptions.
Context
Block, led by Jack Dorsey, entered the Bitcoin mining hardware space with the ambition to challenge incumbents like Bitmain and MicroBT. Its Proto chip, built on a 3nm process, promised to deliver 15 Exahash (EH/s) of computing power. The only publicly known large-scale customer was Core Scientific, a major publicly traded Bitcoin miner. In late 2024, Core Scientific signed a contract to acquire a significant portion of Block's initial production run. Less than a year later, in early 2025, Core Scientific terminated the contract, booking a $41.9 million impairment charge. The same quarter, Core Scientific announced a 15-year partnership with AMD to lease its data center capacity for AI workloads, a contract valued at up to $14 billion in potential revenue.
The narrative from Block was that they had a "healthy pipeline" of demand. The reality? Their only whale paid millions to escape.
Core: The Structural Failure Beneath the Hype
Let me be explicit: this is not a story about bad luck or market timing. This is a story about infrastructure failing to meet its theoretical promise. The 3nm chip was a process node improvement, not a breakthrough in architecture. The key metric for any ASIC miner is efficiency: Joules per Terahash (J/TH). Block never published independent benchmarks. Core Scientific's decision to terminate—and pay $41.9 million—implies that the deployed chips underperformed relative to existing Bitmain S19 series or MicroBT M50 series. In a bear market, where hashprice is compressed, every joule matters. Code executes logic; humans execute fear. Core Scientific's management chose fear of uncompetitive power bills over loyalty to Dorsey's vision.
But the deeper signal is even more structural. Core Scientific's pivot from Bitcoin mining to AI data center leasing is a textbook case of capital efficiency arbitrage. Bitcoin mining, as a business, operates on razor-thin margins driven by hashprice—a function of Bitcoin price, network difficulty, and transaction fees. AI infrastructure, on the other hand, commands premium pricing with long-term contracts. The $14 billion AMD deal provides revenue visibility for over a decade. Mining Bitcoin does not. The market has priced this differential: Core Scientific's stock (CORZ) has rallied since the pivot announcement. Block's stock (SQ) continues to slide, down over 68% in five years.
From my experience analyzing liquidity models during the 2020 DeFi Summer, I know one thing: when capital exits a sector for a higher-yielding alternative, the narrative of that sector must be reassessed. Bitcoin mining is no longer the growth story. It is a commodity business being squeezed from two sides: falling hashprice and rising competition for energy from AI workloads. The energy that could power a 3nm Bitcoin miner now powers an H100 GPU serving inference requests. The market has spoken.
Contrarian: The Decoupling Thesis
The dominant narrative in crypto media is that Bitcoin miners are simply "pivoting to AI" and that this validates their infrastructure. I disagree. This is a decoupling event. The hardware that was designed for Bitcoin mining is being abandoned. The capital that was locked in mining operations is being redirected. Core Scientific's decision is not a pivot within the same ecosystem—it is a migration out of Bitcoin's economic sphere entirely. If more miners follow this path, Bitcoin's network hashpower growth will decelerate, potentially lowering the network's security margin during bear markets. The hashpower that remains will be increasingly concentrated among the most efficient operators—Bitmain's customers—creating a centralization risk that contradicts Bitcoin's ethos.
Moreover, Block's failure is part of a pattern. Jack Dorsey's entire crypto strategy—the Tidal acquisition, the Web5/TBD project, the Bitkey self-custody wallet, the Bitchat messaging app—has been a series of capital-intensive experiments with zero returns. The $41.9 million impairment is just the latest entry in a ledger of failed bets. The total loss across these experiments likely exceeds $500 million. This is not innovation; it is capital destruction. The market is slowly recognizing that Dorsey's vision for crypto is unprofitable. The infrastructure-first skepticism I bring to every analysis tells me: when the only major customer walks away, the product has no moat.
Takeaway
The $41.9 million is a small number relative to Block's market cap. But it is a loud signal. It tells us that Bitcoin mining hardware is not a winner-take-all market where brand alone suffices. It tells us that the real opportunity in crypto infrastructure is no longer mining—it is serving the AI revolution. Core Scientific understood this. Block did not. The next twelve months will determine whether other miners follow Core's lead. If they do, the Bitcoin network's hashpower may reach a plateau, and the narrative of "digital gold" will face its most serious challenge yet: the hard math of energy economics. As a macro watcher, I see the curve bending. The question is whether Bitcoin's value proposition can bend with it.