Hook:
Core Scientific just paid $41.9 million to walk away from a deal. Not because the product was faulty—but because the math stopped working. The contract? A massive order for Block's (Jack Dorsey's company) brand-new 3nm Bitcoin mining chip. The result? A $41.9 million hole in Block's hardware ambitions and a strategic pivot from one of the world's largest public miners that screams volumes about where this industry is headed.
I’ve been in the room where these decisions are made. In 2021, I watched a team sink $2 million into a rushed DEX launch because the code looked good on paper. The integer overflow wasn't found until I audited it myself. Speed is a feature, not a bug, until it breaks. This is that breaking moment on an industrial scale.
Context:
Core Scientific, fresh off its 2024 bankruptcy restructuring, is now a beacon of survival in the Bitcoin mining wilderness. The company operates massive data centers across North America, consuming gigawatts of power to solve SHA-256 hashes. Block, on the other hand, is Jack Dorsey's ambitious fintech-behemoth-turned-crypto-evangelist, betting the farm on everything from self-custody wallets (Bitkey) to decentralised identity (TBD/Web5) to—most critically—its own 3nm mining silicon, codenamed Proto.
The deal was simple: Core would buy a fleet of Proto chips, targeting a delivery of 15 Exahash of computing power. It was a classic 'vertical integration' narrative for Block—control the hardware, control the stack. But the crypto winter of 2022-2024 froze liquidity and shattered sentiment. By 2025, with Bitcoin stuck in a grinding bear market and the ETF hype fading into regulatory sludge, the economics of mining flipped. Yields are transient; infrastructure is permanent. Core decided the infrastructure of the future wasn't a mining rig—it was a GPU rack.
Core Insight: The Matrix of Transition
Let’s get technical. The numbers don’t lie, and the numbers here are ugly for Block.
Core Scientific’s decision to terminate its $41.9 million purchase agreement isn't a 'working capital adjustment'. It's a strategic indictment of three realities:
1. The Energy Arbitrage is Dying. Bitcoin mining’s competitive advantage has always been its ability to buy cheap, stranded energy. But AI/HPC (High Performance Computing) is now bidding up that same energy. Core Scientific, in the very same quarter it walked away from Block, signed a 15-year hosting agreement with AMD. Revenue potential: $14 billion. The same sheds, the same power lines, the same cooling systems—but now serving inference requests for large language models instead of producing block rewards. Infrastructure is permanent, but the yield attached to it shifts fast.
2. 3nm Silicon is Overkill for a Dying Use Case. Jack Dorsey’s team bet that a pure-play ASIC at the cutting edge of fabrication would dominate efficiency. They were right—on paper. A 3nm chip for SHA-256 is theoretically 30-40% more efficient than the 7nm chips powering Bitmain’s S19 series. But in practice, the total cost of ownership (TCO) on a 3nm wafer is astronomical. Core ran the numbers: with Bitcoin at $30k-$40k and network difficulty at 100 trillion, a 3nm miner’s marginal efficiency gain doesn’t justify the premium. It’s like buying a Formula 1 car to drive to the grocery store—overengineered for a world that runs on low-speed, high-reliability vehicles. Speed is a feature, not a bug, until it breaks. And the break here is financial viability.
3. The Human Factor: Dorsey’s Gambles are Cooling. This isn't an isolated incident. Look at Block’s portfolio. Tidal (music service): written down. TBD/Web5 (decentralized identity): effectively killed. Bitchat (direct messaging): a ghost town. Bitkey (hardware wallet): tepid adoption. And now Proto (mining chip): abandoned by its only public client. The pattern is clear: a CEO with a vision, but a company without the execution chops to compete in hyper-commoditized hardware markets. Art is the metadata of human emotion. Jack Dorsey’s emotion was conviction. The market’s emotion is indifference.
From my own audits—I spent 2022 forensically reviewing 100,000 transactions across Arbitrum and Optimism—I’ve learned that smart people building smart tech can still fail if the economic incentives don’t align. This failure isn’t about code; it’s about capital allocation.
Contrarian Angle: The Case for Block (And Why It Fails)
Here’s the counter-intuitive angle that most outlets will miss.
One could argue that Core Scientific’s exit is actually a bullish signal for Block. Hear me out: Block is now free to sell its 15 Exahash of 3nm chips to the open market—or better yet, use them to launch its own mining pool. Imagine Block running a pool where they not only mine blocks but also integrate their Bitkey wallet and TBD identity layer. A vertically integrated, Jack Dorsey-branded mining powerhouse could theoretically challenge Foundry and Antpool.
This is the narrative I see floating in the echo chambers. It’s wrong.
Why? Because liquidity is not latency. The global ASIC market is dominated by two players: Bitmain (70%+ share) and MicroBT (20%+). Both have decade-long relationships with farms, hosting providers, and the entire ecosystem. Block is a new entrant with zero track record in hardware reliability, zero inventory for spare parts, and zero after-market support. Core Scientific, as the 'anchor tenant', was the only thing legitimising the product. Without that validation, Block’s chips are orphans. Curation is the new consensus mechanism, and the market just curated Block’s chips out.
Sure, Block might find some distressed buyers—maybe a few Chinese miners looking for a bargain. But that won't replace the lost $41.9 million in revenue. The real winner here? Bitmain, whose new S21 series at 6nm is already shipping in volume. They just watched their most credible competitor take a direct shot to the gut. The protocol is neutral; the user is the variable. And users (miners) are flocking to the safest bet.
Furthermore, the contrarian narrative that 'Block going it alone' is positive ignores the basic physics of ASIC supply chains. A 3nm mask set costs $40 million+. Block needs to sell at least 5-8 Exahash of chips to break even on R&D. Core’s order represented the bulk of their first production. Without a replacement buyer, Block is staring down a multi-quarter inventory writedown. Accounting for that, Block’s stock (SQ) will feel the pain.
Takeaway:
The $41.9 million goodbye from Core Scientific isn’t just a contract cancellation. It’s a tombstone for the idea that Silicon Valley-style disruption can beat vertical integration in a commodity market. Bitcoin mining is no longer a tech story—it’s a utility story.
I don’t predict trends; I ride the volatility. And the volatility here points to a binary future: either Block doubles down on its own mining infrastructure (burning cash) or it shuts down Proto altogether. Either way, the message is clear: the next wave of mining infrastructure won’t be built by Western entrepreneurs chasing a vision. It will be built by the utility providers who can answer one question: “Can this machine make money at $20,000 Bitcoin?”
If the answer is no, the chip is just an expensive paperweight. Yields are transient; infrastructure is permanent. And right now, the most permanent infrastructure in town is AI data centers, not ASIC farms.