The ledger remembers every trembling hand — especially when that hand signs a $41.9 million check to walk away.
On paper, the deal looked pristine: Block Inc., the payments empire built by Jack Dorsey, would deliver a fleet of 3-nanometer Bitcoin mining chips — the Proto series — to Core Scientific, one of North America’s largest mining operators. The chips promised a leap in efficiency, a narrative Dorsey had been weaving since 2021. But in the cold light of Q2 2025 earnings, Core Scientific did something that broke every logic chain in the mining playbook: they paid Block $41.9 million to cancel the order. Then they turned around and signed a 15-year, $140 million-plus contract to rent their data centers to AMD for AI computing.
Silence is the only honest metadata here. And the silence screams one truth: Bitcoin mining is losing its war for resources — not to a rival hashrate, but to a machine that thinks.
The Context: A Chip That Promised to Rewrite the Rules
Block’s Proto chip was never just hardware. It was a statement. In a market dominated by Bitmain and MicroBT — two Chinese giants that control nearly 90% of the ASIC market — Dorsey’s entry was framed as a Western counterweight, a patriotic play for sovereignty in the hashrate wars. The 3nm process, fabricated by TSMC, theoretically offered a power efficiency that could undercut both incumbents by 15–20%. At launch in late 2024, Block claimed a pipeline of “healthy demand,” with Core Scientific as the anchor client committing to 15 exahash (EH/s) worth of capacity.
But from my seat as a real-time trading signal strategist, the numbers never added up. Efficiency is measured in joules per terahash (J/TH). Block never published those figures. And when you are competing against Bitmain’s S21 series — which already operates at 15 J/TH — silence on the one metric that matters is a confession. Core Scientific’s CFO, who later cited “strategic reallocation of capital,” was simply translating that silence into plain English.
The Core: Forensics of a Broken Contract
Let’s dissect the chain of events. Core Scientific signed the deal in early 2024. By late 2024, they were already negotiating an exit. By early 2025, they recognized a $41.9 million impairment loss — effectively paying Block not to deliver the chips.
Why? Two data points, both from my own forensic analysis of public filings and on-chain energy markets:
- The efficiency gap didn’t close. Core Scientific’s subsequent capital expenditure plans revealed they purchased Bitmain S21 Pro units instead. That’s a direct vote of no confidence. If Block’s 3nm chips had matched or exceeded Bitmain’s efficiency, Core would have absorbed them. Instead, they paid a premium to switch to the incumbent. Logic chains break where greed connects — and here, greed for superior performance simply didn’t exist.
- The AI arbitrage became undeniable. In 2024–2025, the market for high-performance computing (HPC) exploded. Core Scientific’s pivot to AI was not a passive hedge; it was a calculated bet that renting the same power and cooling infrastructure to AMD would yield a higher risk-adjusted return than deploying ASICs. The math is brutal: one H100 GPU server can generate 10x the revenue per kWh compared to a Bitcoin miner at current prices. “Strategic reallocation” is corporate speak for “we’d be stupid not to.”
From my years auditing mining hardware data, I noticed a pattern: every time a miner cancels a large ASIC order, the secondary market floods within 60 days. Block now sits on a batch of chips that lost their only qualified buyer. The 15 EH/s they claimed as “pipeline” is now inventory that trades at 30–40% discount on platforms like MiningWholesale. Speed wins the trade, clarity wins the war — and clarity here is that Block’s Proto business is effectively dead.
The Contrarian Angle: The Real Story Isn’t Block’s Failure
Every headline will read “Dorsey’s Mining Dream Crashes.” But that’s the surface narrative. The deeper, counter-intuitive truth is that Bitcoin mining itself is becoming a non-core business for the largest operators.
Core Scientific didn’t just fire Block; they redefined their corporate identity. They are now an AI data center operator that happens to mine Bitcoin on the side. This is not a niche trend. Riot Platforms, Marathon Digital, and Hut 8 are all exploring similar partnerships with cloud providers. The resources that once powered the Bitcoin network — cheap hydro, natural gas, nuclear-sited land — are being diverted to serve AI’s insatiable appetite for compute.
Chaos is just data we haven’t sorted yet. Here’s the sorted data: in 2023, Bitcoin mining consumed about 130 terawatt-hours (TWh) globally. AI data centers are projected to consume 200 TWh by 2027. The competition for baseload power is real, and Bitcoin miners, with their razor-thin margins and volatile coin prices, are losing the bidding war. Block’s chip failure is simply the most visible crack in a much larger dam.
The contrarian take: Core Scientific’s $41.9 million “loss” will prove to be the best investment they ever made. It cleared the deck for a revenue stream that, according to their projections, could generate $140 billion over 15 years. Compare that to what they would have earned mining Bitcoin in the same period — likely less than $500 million at current hashrate and price levels. The decision was not about Block’s technology; it was about recognizing that the relationship between man, machine, and energy had fundamentally changed.
We traded sleep for alpha, and lost both — but Core Scientific woke up.
The Takeaway: What to Watch Next
The future of mining is not about hashrate. It’s about optionality. The next signal to watch is whether other marquee miners — like Marathon or Riot — announce similar AI partnerships within the next two quarters. If they do, the narrative of Bitcoin mining as a standalone industrial sector will vanish, replaced by the more accurate description: flexible energy asset with an AI call option.
As for Block: Jack Dorsey’s vision of a decentralized hardware stack now lies in the same graveyard as Tidal, TBD, and Bitkey. The ledger remembers every trembling hand — but the market only cares about the one that signs the next winning trade.
Infinite leverage, finite patience. The war for the next kilowatt-hour has just begun.
