Code betrays when we do.
On paper, the deal between Block and Core Scientific looked like a textbook alignment of vision and execution. Block, led by Jack Dorsey, poured resources into designing a 3-nanometer Bitcoin mining chip, code-named Proto. Core Scientific, one of the largest public mining operators, committed to purchasing the chips in bulk—enough to power 15 exahash of computing capacity. The narrative was seductive: decentralization championed by a tech icon, operationalized by a mining giant. Then came the termination. Core Scientific paid $41.9 million to walk away from the contract, choosing to recognize a loss rather than take delivery of the hardware.
Context
To understand the significance, you need to see the full picture. Block’s Proto chip was its first foray into mining silicon, a market dominated by Bitmain (roughly 70-80% market share) and MicroBT (15-25%). The 3-nanometer process node, while not revolutionary by semiconductor standards, represented a leap for a company whose primary expertise lies in payments and social media. Core Scientific, meanwhile, was emerging from bankruptcy restructuring in early 2024, having survived the crypto winter by shedding debt and refocusing on operational efficiency. The termination announcement came with a broader strategic shift: Core Scientific signed a 15-year agreement with AMD to lease its existing data center capacity for AI and high-performance computing (HPC), a contract tentatively valued at $14 billion in potential revenue. The mining chips were no longer part of the plan.
Core
The $41.9 million write-off is not an anomaly—it is a signal. In my years analyzing protocol infrastructure, I’ve learned to treat contract terminations as diagnostic events. They often reveal what balance sheets and press releases try to hide.
First, consider the signal-to-noise ratio. Block claimed a "healthy pipeline of demand" for its Proto chips. But healthy pipelines do not produce $41.9 million penalty payments. The only logical conclusion is that the chip’s real-world performance—its energy efficiency (joules per terahash), stability under load, or overall cost per hash—failed to meet Core Scientific’s projections. Mining is a commodity business where fractions of a penny per kilowatt-hour determine viability. If Proto’s efficiency lagged behind Bitmain’s S21 or MicroBT’s M63, then taking delivery would have been a losing proposition from day one. Core Scientific’s decision to pay a penalty rather than operate the chips is an explicit admission that the chips were not competitive.
Second, the timing is damning. Block announced the Proto chip in early 2025 with considerable fanfare. Core Scientific’s termination came within months of final delivery. This rapid reversal suggests a failure not in theory but in engineering validation. I recall a similar lesson from my Zilliqa days: a sharding implementation that looked flawless in simulation but uncovered a race condition during mainnet stress testing. The temptation is to blame the test environment, but the system was the problem. The code betrays when we do—when we rush to market before our technology has been battle-tested against real-world conditions.
Third, look at the broader fallout for Block. This mining chip failure is one node in a constellation of failed crypto experiments under Dorsey’s leadership: Tidal (music platform acquired for $297 million, later written down), TBD (decentralized identity project, shut down), Bitkey (self-custody wallet, discontinued), and Bitchat (encrypted messaging, never launched). Combined with $200 million in regulatory fines from the CFPB and state regulators over Cash App’s fraud handling, and a 68% decline in Block’s stock price over five years, the pattern is clear. Block has been trying to buy or build its way into crypto relevance, but product after product has failed to gain traction. Burnout is the tax on innovation—and Block has been paying it in compound interest.
Contrarian
The easy takeaway is that Bitcoin mining is dying, that AI is eating its lunch, and that hardware incumbents like Bitmain are invincible. That narrative misses the deeper truth.
What Core Scientific did is not a rejection of Bitcoin—it is a rational allocation of capital in a market where margins are razor-thin. The company’s pivot to AI computing is a survival move, not an ideological shift. Mining remains profitable for those who operate with the most efficient chips and lowest electricity costs. But efficiency is a moving target, and the barrier to entry is rising. Block’s failure is not evidence that mining is dead; it is evidence that mining hardware is a commodity business that punishes late entrants with subpar engineering.
The contrarian angle is this: perhaps we should celebrate Core Scientific’s decision rather than mourn it. By paying $41.9 million to walk away from a product that didn’t meet its standards, the company demonstrated integrity. In an industry where hype often masks reality, walking away from sunk costs is a sign of discipline. Code betrays when we do—but honesty in the face of failure is an antidote.
Furthermore, the shift to AI infrastructure is not a betrayal of crypto’s mission. Bitcoin miners own unique assets: access to cheap power, existing data center infrastructure, and expertise in managing high-density computing loads. Repurposing those assets for AI is simply running the same business model with a different end customer. The network effect of Bitcoin still exists, but the marginal dollar is now chasing AI margins, not mining margins. That is a market signal, not a moral judgment.
Takeaway
The Block-Core Scientific saga is a parable for our industry. It reminds us that technology does not respect narratives—it respects physics, economics, and engineering rigor. Jack Dorsey’s vision for a decentralized hardware ecosystem was noble, but nobility does not turn a flawed chip into a profitable one. The $41.9 million penalty is the price of learning that lesson.
As we watch Core Scientific rebrand itself as an AI data center operator, let’s not mistake adaptation for abandonment. Mining will survive, but it will be leaner, more concentrated, and less forgiving. The question we must ask ourselves is not whether Bitcoin mining is dying, but whether we have the courage to admit when our own code—our own products—betray the ideals we claim to serve.
Burnout is the tax on innovation. But it is also the tuition for wisdom. And wisdom, unlike a 3-nanometer chip, cannot be manufactured. It must be earned.