A Math Professor Says AI Is Bitcoin's Real Security Threat. The Mining Math Backs Him Up

I've spent the last eighteen months watching bitcoin miners rebrand themselves as AI companies. Every earnings call is the same theater. The press release says data center. The 10-K says GPUs. Somewhere in the footnotes, the actual hash rate shows up like an awkward cousin at Thanksgiving.

So when Micah Warren, an associate professor of mathematics at the University of Oregon, started talking about game theory and bitcoin's security model, I paid attention. He wrote a book called Bitcoin: A Game-Theoretic Analysis. His argument isn't the usual quantum panic or the tired "governments will ban it" routine. It's simpler, and it's worse. Miners have alternatives now. Alternatives change incentives. Incentives are the entire ballgame.

The Mechanics Nobody Wants To Model

Here's the thing about bitcoin's security that gets glossed over in every whitepaper summary. There's no cryptographic force field. There's a payout structure. A miner spends electricity to extend the chain, and the protocol hands them 3.125 BTC plus whatever fees are sitting in the block. At a $90,000 bitcoin, that's roughly $280,000 per block, every ten minutes, forever, until it isn't.

Warren's point is that this payout structure is doing the work people think math is doing. Attacking the network has to be less profitable than extending it. Classic reorgs and double-spends fail that test today, and they'll keep failing it as long as the honest chain is the profitable chain. The cost of rewriting six blocks on a network hashing near a zetahash per second is measured in millions of dollars of hardware and power, with no guarantee you win. Naturally, nobody tries.

But the incentive design assumes miners are somewhat captive. That their capital, their buildings, their electrical interconnects really only have one productive use. That assumption broke sometime around 2023, when Core Scientific signed a multi-billion dollar compute contract with CoreWeave and every other listed miner looked at the per-megawatt math and got religion.

Because that's the whole story. An AI data center tenant can pay two to four times more per megawatt-hour than a bitcoin mining rig does. Same substation. Same transformers. Same land. Triple the revenue. If you're a miner with a board and a share price and a fiduciary duty, the choice isn't hard. It's arithmetic.

Warren also walks through the uglier scenarios. Feather forks, where a minority of miners keeps blocks or transactions off the main chain to censor them. Government stakes in publicly traded mining firms, which turns a shareholder register into a policy lever. North Korea and Iran running hash where they can. And the one nobody likes talking about at conferences, miner collusion as a bigger structural risk than any rogue attacker with a rented ASIC farm.

Then there's the demand side choke point. Institutional buyers want coins with clean provenance. No sanctioned addresses, no mixer history, no awkward chain of custody. Which seems like an even stronger argument for a two-tier market where freshly mined bitcoin trades at a premium and everything else gets discounted like a salvage title car. Fungibility is the property nobody prices until it's gone.

What This Actually Does To The Market

Pull the camera back and the security budget debate stops being academic.

The April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC. The 2028 halving takes it to 1.5625. Transaction fees are supposed to fill that gap. But fees are lumpy and spiky and demand-dependent, and they don't show up on a schedule. A miner can't model them the way they model a ten-year power purchase agreement.

So if AI compute keeps pulling marginal capital out of mining, hash rate growth flattens. Flatter hash rate means a lower absolute cost to attack, even if the percentage cost stays the same. That's not a tomorrow problem. It's a 2028 to 2032 problem, and markets don't price those until they're forced to.

Who wins? Power producers with interconnection queues, obviously. Firms that locked in long-term contracts before the AI bidding war started. And GPU cloud operators, who get stranded mining sites with cheap electricity for pennies on the dollar of what building greenfield would cost.

Who loses? Anyone who bought a mining stock in 2021 for the bitcoin beta and got an AI infrastructure company instead. That's not a bad outcome, but it's a different asset. Read the segment reporting before you assume you own what you think you own.

What I'd Actually Watch

Spare me the roadmap. Nobody knows how this resolves, and the people who claim they do are selling something.

But you can watch a few real numbers instead of vibes. First, fee revenue as a share of total miner revenue. If that's climbing while hash rate stalls, the network is doing exactly what it's supposed to. If it's flat and hash rate is flat, that's a problem wearing a nice suit.

Second, miner capex disclosure. Companies that are converting existing sites to HPC and keeping the mining fleet running are hedged. Companies that can't stop announcing new AI contracts while quietly shelving hash rate expansion are telling you something.

Third, watch the strange marriage of convenience between treasuries and miners. Corporate bitcoin treasuries and governments holding hash both change the game in ways the original design never contemplated. Warren's whole framework is about who has what to gain. Add a state balance sheet to that equation and the answers get less pretty.

I've seen enough hype cycles to know that "bitcoin is doomed" sells better than "bitcoin is fine, it's just changing shape." This is the second one. The network isn't breaking. The security model is getting repriced, slowly, by a market that values megawatts more than it values ideology. That's not a catastrophe. It's a bill coming due, and the industry has about three years to figure out who's paying it.

Warren's real contribution isn't a prediction. It's a reminder that bitcoin's resilience has never come from the code being clever. It comes from enough people finding it more profitable to play nice. The day that stops being true, no amount of cryptography saves you.