In a world of noise, code is the only quiet truth. But sometimes, even code can't save you from a bad business decision.
On a Tuesday in late 2025, Core Scientific—once the largest publicly traded Bitcoin miner in North America—dropped a bomb. They paid Block, Inc. $41.9 million in cash to walk away from a binding purchase agreement for Jack Dorsey's Proto 3nm mining chips. Not a renegotiation. Not a delay. A full termination. The chips were already in production. The deal was signed. And Core decided that taking a $41.9 million loss was cheaper than taking delivery.
That is not a soft landing. That is a structural fracture. And as someone who has spent the last nine years auditing smart contracts and tokenomics models, I can tell you—when a sophisticated miner like Core Scientific burns millions to avoid using your hardware, it's not a market hiccup. It's a signal that the underlying design is flawed.
Let me back up. I've been in this industry since 2017, when I was a 20-year-old finance student at the University of Lagos, manually auditing ERC-20 implementations in the Zeppelin Solidity library. I discovered integer overflow vulnerabilities that could drain entire contracts. Back then, decentralized trust wasn't philosophical—it was mathematical. A bug in the code meant real, irreversible loss. That lesson has never left me. Code is the only truth. And Block's mining chip story is a textbook case of code—and business logic—failing the test of reality.
Context: The Promise and the Crash
Block, Inc. (formerly Square) entered the Bitcoin mining hardware space in 2021 with a grand vision. Jack Dorsey, a vocal Bitcoin maximalist, wanted to create an open, decentralized mining ecosystem. The Proto team designed a 3-nanometer ASIC chip—state-of-the-art in semiconductor manufacturing—targeting a hash rate of 15 exahash per second. They partnered with Core Scientific, the largest mining host in the US, to be the anchor customer.
By early 2025, Block had completed the chip design and started manufacturing. Core Scientific, meanwhile, was emerging from bankruptcy and restructuring. They had committed to purchasing a substantial number of these chips. The expectation was that Proto would be a serious competitor to Bitmain and MicroBT.
But then the numbers didn't add up. Core Scientific, in their Q2 2025 earnings call, announced the $41.9 million write-off. CEO Adam Sullivan stated that the company was 'strategically repositioning' toward AI and high-performance computing (HPC). They had signed a 15-year contract with AMD to lease data center space, potentially generating $140 billion in revenue. The mining division was being de-emphasized.
Here's the raw reality: Core Scientific chose $41.9 million in immediate cash loss over the long-term obligation to deploy Block's chips. That's not a subtle signal. That's a sledgehammer.
Core Insight: The Mathematics of Unviability
I've analyzed dozens of mining hardware contracts in my career. The decision to kill a deal at a $41.9 million cost requires a cold, calculated assessment of expected future returns. Let me walk you through the logic.
First, the chip's performance. Block never released independent benchmarks for their 3nm ASIC. The only publicly stated metric was '15 EH/s'—a total hash rate target, not an efficiency ratio. In mining, the critical metric is joules per terahash (J/TH). Bitmain's S19 series operates at around 27 J/TH. The newer S21 series achieves roughly 20 J/TH. If Block's chip couldn't beat 20 J/TH, it was immediately obsolete.
Second, the cost per chip. Block was a new entrant with limited manufacturing scale. Their unit cost would have been significantly higher than Bitmain's, who produce millions of units. Without a massive price discount, Block's chips would have negative net present value for any rational miner.
Third, the timing. Core Scientific's pivot to AI wasn't a whim. They looked at the current Bitcoin hash rate—over 600 EH/s and rising—and the halving that slashed block rewards to 3.125 BTC. The mining margin is razor thin. Meanwhile, AI data center leases are long-term, dollar-denominated, and growing at 30% annually. The return on capital for AI infrastructure now exceeds that of Bitcoin mining by a factor of 3 to 5.
Core Scientific's CFO likely ran the scenario: deploy Block's chips and fight for declining mining margins, or pay $41.9 million to exit and lease the same power capacity to AMD at 10x the revenue rate. The math was brutal. The chips were a liability.
This echoes a pattern I've seen since 2020. During DeFi Summer, I executed a $45,000 arbitrage between Curve and Uniswap. I then wrote a deep analysis of how pegged assets create systemic fragility. The lesson was that interconnected protocols hide hidden leverage. Here, the hidden leverage was the assumption that a new mining chip could compete with incumbents who have years of supply chain advantage and millions of deployed units.
The Systemic Fragility of Block's Crypto Strategy
Proto isn't Block's only failed bet. Let's inventory the casualties:
- Tidal (music streaming): Acquired for $297 million in 2021. By 2025, the service was essentially shut down, with Block writing down the entire value.
- TBD (decentralized identity/Web5): Announced with major fanfare in 2022. By late 2024, the team was disbanded and the project shelved.
- Bitchat (decentralized messaging): Launched and quickly abandoned due to lack of traction.
- Bitkey (self-custody hardware wallet): Released to minimal adoption, facing stiff competition from Ledger and Trezor.
- Cash App's crypto trading: While profitable, the regulatory scrutiny has increased dramatically. Block paid a $200+ million fine in 2025 for failing to properly handle fraud claims.
Each of these was sold to investors as part of a cohesive 'open financial system.' In reality, they were disconnected experiments. The aggregate cost? Over $1.5 billion in market cap destruction—Block's stock fell 68% from its 2021 peak.
Jack Dorsey is a visionary. But vision without execution is just a hallucination. And in the world of hardware, execution means supply chain, manufacturing yields, and cost curves. Code is the only quiet truth—but code alone doesn't make a mining chip profitable.
Contrarian Angle: The Pragmatism Test
Here's the counter-intuitive take: Core Scientific's $41.9 million payment might be the smartest capital allocation in the mining sector this year.
While many will see it as a failure of Block's technology, I see it as a victory for rational decision-making. Core Scientific admitted they made a mistake, paid the penalty, and redeployed resources to a higher-return activity. That's rare in an industry often driven by ego and maximalist narratives.
But there's a deeper blind spot for the wider market. The narrative is that Bitcoin mining is being 'saved' by AI. That's only partially true. Many mining companies are now 'AI-ready' data centers hosting NVIDIA or AMD chips. But the transition is not without risk. The AI infra market could oversaturate. If generative AI demand plateaus, the same 15-year contracts could become burdens. Core Scientific is betting that AI is structural, not cyclical. That's a bet, not a certainty.
Furthermore, Block's failure reveals an uncomfortable truth about the mining oligopoly. Bitmain and MicroBT control over 90% of the ASIC market. Their dominance is not due to superior technology alone, but to vertical integration and relationships with chip fabs. New entrants like Block face impossible odds. The ecosystem's 'decentralization' rhetoric breaks down when hardware is a two-player game.
Takeaway: The Forward-Looking Judgment
What does this mean for you, the reader? Three signals:
- If you hold Block stock, understand that the crypto hardware division is a dead weight. The company should return to focusing on Square and Cash App payments. Jack Dorsey's strategic investments outside the core business have been overwhelmingly value-destructive. The board should ask hard questions.
- If you are a miner, the AI pivot is real but crowded. The early movers like Core Scientific and Hive Blockchain will benefit, but the window is closing. By 2027, most Tier 1 data centers will be AI-hybrid. The pure Bitcoin miner will be an endangered species.
- If you are a developer or entrepreneur building in crypto, learn from Block's mistakes. Building hardware is not like building a dApp. The moat is not code; it's capital, relationships, and patience. Don't confuse a Twitter following with a fabrication line.
In a world of noise, code is the only quiet truth. But even quiet truth can't rewrite the laws of economics. As I wrote in my 2022 post-mortem on three collapsed DeFi protocols: 'Trust is not built by press releases. It is verified by balance sheets.'
Core Scientific verified that Block's chips were not worth the paper they were printed on. They paid $41.9 million for that verification. Next time you see a bold claim about a new mining chip, remember: the code may be elegant, but the math is merciless.