Core Scientific just paid $41.9 million to walk away from Block's 3nm mining chips. The check cleared. The story didn't end there.
It's a quiet bomb in a noisy market. No tweets from Jack. No press conference from Core. Just an SEC filing buried in Q2 earnings—a $41.9M penalty for terminating a contract to buy Block's Proto mining chips. A contract that was supposed to deliver 15 exahash of computational power. A contract that was supposed to prove Block could compete with Bitmain and MicroBT.
Instead, it became a tombstone for Block's mining hardware ambitions. And a flashing neon sign for where the mining industry is headed: away from Bitcoin, toward AI.
Context: The Players and the Prize
Block, Inc.—the company formerly known as Square, helmed by Jack Dorsey—has been on a crypto crusade since 2018. From building Bitkey (a self-custody wallet) to TBD (a decentralized identity platform) to Tidal (a music streaming acquisition), Dorsey has thrown billions at blockchain-adjacent bets. The most ambitious? Proto: a custom 3nm mining chip designed to rival the Antminer S19 and Whatsminer M50. The idea was simple: if Block could manufacture cheaper, more efficient ASICs, it could disrupt the Bitmain-MicroBT duopoly and secure a slice of the Bitcoin mining hardware market.
Core Scientific, once the largest publicly traded Bitcoin miner in the US, was the ideal first customer. Emerging from bankruptcy in early 2024, Core needed to rebuild its hashing fleet. Block's 3nm chip promised lower power consumption and higher hash rates. In mid-2024, they signed a multi-year deal for 15 EH/s of hardware. The tech world took notice.
But by late 2025, Core walked away. The $41.9 million termination fee is a fraction of what they saved by not deploying chips that—according to internal data—likely underperformed. The ledger remembers what the hype forgets.
Core: The Anatomy of a Failure
Let's dig into the numbers. 15 EH/s is significant. For context, the entire Bitcoin network runs at roughly 700 EH/s. Block's contract represented about 2% of total network hashrate—a meaningful chunk for a new entrant. But performance isn't just about raw hashpower; it's about efficiency, measured in joules per terahash (J/TH). Neither Block nor Core disclosed the J/TH of the Proto chips. That silence is deafening.
Based on my years tracking mining chip battles—from the Bitmain S9 dominance to the MicroBT M20 series—I can tell you that a 3nm node alone doesn't guarantee efficiency. Architectural decisions matter more. Block's team, led by former chip designers from Intel and Qualcomm, may have optimized for speed over power. In a market where electricity is the largest operational cost, a few extra J/TH can kill profitability.
Core's decision to terminate, even at a $41.9M loss, signals that the chips' actual performance fell short of breakeven. Tracing the footprint of digital scarcity, I've seen this pattern before—in 2017, when Ethereum time-lock contracts promised instant liquidity but delivered lockups that lasted years. The difference? This time, the failure is physical, not digital. Hardware doesn't get patched with a software upgrade.
Let's break down the financial impact:
- Block recorded an $85.9 million impairment on Proto-related assets in Q2 2025. That's not just the lost revenue; it's the write-down of inventory, tooling, and R&D.
- Core Scientific, meanwhile, redirected the $41.9M penalty into its new AI data center partnership with AMD—a 15-year, $14 billion estimated revenue contract.
- Block's stock has declined 68% over five years. Core's stock has more than doubled since announcing the AMD deal.
The contrast is stark. One company bet on mining and lost. The other bet on AI and won—for now.
But let's not stop at the numbers. The real story is the human one. I remember the 2020 DeFi Summer, when I organized a Twitter Spaces with Uniswap devs to explain AMMs through party analogies. That was the high point of crypto innovation. Now, in 2025, the innovation is happening in Austin, Texas, where Core is repurposing its mining facilities to house AMD GPUs. The same power lines that once ran Antminers now cool Instinct MI300s. Where liquidity meets the human story, the narrative has shifted from digital gold to digital brain.
Contrarian: The AI Pivot Might Be a Mirage
Before we crown Core Scientific as the new king, let's ask: is the AI pivot actually sustainable?
The $14 billion AMD contract is an estimate, not a guarantee. If AI investment cools—or if AMD's chips lose market share to Nvidia—Core could find itself in a similar position to Block: stuck with specialized infrastructure that has no buyer. The mining-to-AI transition is a one-way door. Once you retrofit your facility for liquid cooling and high-density racks, you can't easily switch back to ASICs. And the competition is fierce: traditional data center providers like Equinix and Digital Realty are expanding into AI-optimized colocation. Miners have lower cost of capital for power, but they lack the operational expertise to manage enterprise GPU workloads.
Moreover, the flood of miners pivoting to AI could oversaturate the market. Every major miner—Marathon, Riot, Hut 8—has announced AI pilot programs. If they all succeed, the supply of AI compute will outstrip demand, driving down rental prices. Core's first-mover advantage may evaporate within two years.
The ledger remembers what the hype forgets. During the 2021 Bored Ape mania, I immersed myself in Yuga Labs events, writing about NFTs as digital identity. The hype collapsed when floor prices crashed. The same cycle could repeat with AI mining. The value of an AI data center depends on the demand for LLM inference and training. If the current AI boom is a bubble, the miners who sold their Bitcoin to build GPU farms will be left holding the bag.
Takeaway: What to Watch Next
This isn't just a story about Block or Core. It's a signal about the changing nature of crypto infrastructure. The next 12 months will determine whether Bitcoin mining remains a viable standalone industry or becomes a legacy business cannibalized by AI.
Watch these signals: - Block’s Q3 2025 earnings: Will they announce the closure of Proto? If yes, the hardware dream is dead. - Core’s AI revenue: If AMD revenue exceeds hosting costs, the pivot is validated. If not, question the hype. - Other miners: If Riot or Marathon announce major AI contracts, the trend is confirmed. If they stay with Bitcoin, it's a contrarian bet.
The question I keep asking myself: Are we witnessing the end of mining as we know it, or just a painful transition to a hybrid model where every data center is a polyglot of compute?
From my desk in Jakarta, watching the hashrate charts and AI stock prices converge, I smell a shift. The ghost of Ethereum's Proof-of-Work still haunts the industry, but now it's fighting for relevance against a new ghost—the ghost of AI dominance. Caught in the current of real-time value, the miners who adapt will survive. The ones who don't will become footnotes in the ledger—just like Block's 3nm chips.
