Chaos detected. Analysis loading.
Core Scientific just paid $41.9 million to terminate a contract for Block's 3nm mining chips. Not a delay. Not a renegotiation. A full exit. The message is clear: Block's Proto chip was a dud, and the entire Bitcoin mining model is being abandoned for AI.
Context: The Chip That Wasn't
Block, Jack Dorsey's payments company, entered the Bitcoin mining hardware race with grand ambitions. In 2023, they announced a custom 3nm ASIC, aiming to challenge Bitmain and MicroBT. Core Scientific, one of the largest public miners, committed to 15 Exahash of these chips. By early 2025, that deal was dead. Core took a $41.9M impairment charge and walked away.
Why? Because Block's chip never delivered. No public benchmark on energy efficiency (J/TH). No independent validation. Just Dorsey's vision. And Core, facing real capital allocation decisions, chose to eat the penalty rather than deploy hardware that would bleed money.
Core: The Autopsy of a Failed Product
The core insight is not just that Block failed—it's the scale of failure. Core Scientific's decision signals a devastating technical verdict. Based on my years analyzing mining hardware efficiency curves, a $41.9M penalty is only rational if the chip's real-world performance was 15-20% worse than alternatives. That means Block's 3nm chip was less efficient than Bitmain's S19 series (7nm). Extreme? Yes. But the data demands it.
Meanwhile, Core Scientific didn't just drop Block. They pivoted hard—signing a 15-year, $14 billion revenue contract to host AMD's AI chips. They are literally converting mining facilities into AI data centers. The same power, land, and cooling infrastructure now serves a higher-margin master. Bitcoin's loss is AI's gain.
This is not an isolated incident. Block's entire crypto portfolio is rotting. Tidal (music), TBD (Web5), Bitkey (self-custody)—all written down or shut down. Their payment subsidiary Cash App just paid $200M in regulatory fines. Block's stock is down 68% over five years. The narrative of a crypto-native conglomerate is collapsing.
Contrarian: It's Not About Block—It's About Bitcoin's Competitive Disadvantage
The market will frame this as a Block failure. It's not. It's a structural signal that Bitcoin mining is losing the war for resources. Power, capital, talent—all flowing to AI. Core's choice proves that even when you already have the infrastructure, Bitcoin mining's risk-adjusted return is now inferior to AI compute leasing.
Look at the numbers: Core's $41.9M penalty is a fraction of their expected AI revenue. They considered the present value of mining profit versus AI profit—and mining lost. This will repeat. Riot, Marathon, and others are already exploring AI pivots. If the top miners exit, Bitcoin's hashrate growth slows. Security budget shrinks. The doomsday narrative for a post-halving Bitcoin becomes real.
And Block? Their chip design may end up on the secondary market at a discount, further depressing used miner prices. The ripple effect: Bitmain and MicroBT will face less competition, but also a shrinking customer base. The mining industry is consolidating around survivors with AI options.
EOS didn't die; it evolved. Do you?
Takeaway: Watch the Energy Flow
The next signal is not from Block or Core. Watch the power grid interconnections. Every new data center announcement—is it for Bitcoin or AI? If the ratio tilts further toward AI, Bitcoin's security model faces a quiet, lethal erosion. The question isn't if mining chips get better. It's whether anyone will still want to use them.
Based on my experience in the 2022 Terra collapse, this is the same pattern: a gradual resource drain that looks like a sector rotation until it becomes an existential crisis. The old model is dead. The new one runs on LLMs, not hashes.