AI is outbidding Bitcoin mining for energy and capital, and the price of Bitcoin's security has never asked what security is worth

Written By Saša Čoh, CEO of NiceHash

In my first letter I made an uncomfortable argument: Bitcoin's wealth has concentrated at the top of its stack, in ETFs, treasury companies, and exchanges, while the mining layer that secures that wealth concentrates into a handful of coordination points at the bottom, its revenue squeezed by the protocol's own halving schedule. The holders and the foundation barely touch. And unlike a gold mine, a Bitcoin mine is a data center with options.

This letter is about who is now exercising those options on the mine's behalf, and about a fact that most holders have never examined: what the price of Bitcoin's security actually contains.

The other bidder for Bitcoin's army

AI has taken the spotlight from Bitcoin. Less noticed is that it is also taking Bitcoin's inputs.

Mining runs on four things: energy, capital, industrial sites, and chips. AI now outbids it for the most valuable forms of all four. Every marginal megawatt of firm, new generation and grid interconnect is courted by hyperscalers carrying the deepest balance sheets in corporate history. Investment capital that once funded mining expansion flows to AI infrastructure almost by default: a mining company raises money today by explaining its AI strategy, not its hashrate roadmap. Existing miners, the marginal sellers of powered land and industrial data centers, are converting capacity into multi-year, dollar-denominated compute contracts that make volatile, halving-compressed block rewards look unbankable by comparison. And the same leading-edge foundries that fabricate mining ASICs now allocate their most advanced capacity to AI accelerators first; mining silicon queues behind them, at prices set by the bidder ahead.

For fourteen years, Bitcoin's security budget competed only against its own issuance schedule. Now it competes for energy, capital, and silicon against the best-funded buyer the infrastructure market has ever seen. Hashrate does not have to be attacked to decline. It just has to be outbid: one kilowatt-hour, one financing round, one wafer, one converted site at a time.

Mining's one card, and what it does not buy

Mining does hold one structural card in this fight, and it deserves to be named. Mining is the most flexible industrial load on the grid: it switches on when energy is abundant and vanishes the moment the grid tightens, absorbing surplus renewables, monetizing stranded and curtailed power, selling demand response back to the system. It is not a battery (it stores nothing), but it does for grid balance what a battery does: soaks up excess, disappears in scarcity. Grid operators have started writing that difference into their rules, placing flexible mining load in a preferred position while lining up always-on data centers for curtailment first. AI cannot play this role at scale; training runs and service guarantees demand firm, around-the-clock power. This flexibility is why mining will not simply be priced off the grid: it can live on energy AI cannot use.

But be precise about what flexibility buys. It lowers mining's cost floor. It does not raise mining's revenue ceiling. Interruptibility lets the security layer survive on the grid's margins; whether it thrives there depends entirely on whether anyone is paying for the security it produces. And note what each conversion destroys: a mine that becomes an AI campus removes both a unit of Bitcoin's security and a unit of the grid's flexibility, two public goods traded away in the same transaction.

So flexibility answers whether mining can survive. It does not answer who pays for security. And that is the deeper thing AI exposes: not merely a threat to mining margins, but the absence of any market through which Bitcoin's holders can express demand for the security they depend on. To see why, at present, almost nobody pays for it, look at how security is priced.

Bitcoin has two prices

Bitcoin has two prices: related, but not the same thing at all.

The first is the one everyone watches, the price of the asset. The second is the price of hashrate: what one unit of the computing power securing the network earns per day. And here is what most holders have never examined: the reference price of hashrate is not discovered, it is calculated. Arithmetic, not an auction. The block subsidy plus average transaction fees, multiplied by the blocks produced per day, divided across the network's total hashrate, converted to dollars by the price of Bitcoin. Every term in that formula is a protocol variable or the asset's price.

Notice what is absent: demand for security. The formula tells miners what mining earned. It never asks what protection is worth to the people being protected. Today, the price of Bitcoin's security is almost entirely a shadow of the asset's price and the protocol's issuance schedule, which means holders fund the network's defense only implicitly, through price appreciation flowing into a shrinking subsidy. Each halving cuts the subsidy term in half, mechanically draining the calculated component of hashrate's price. Two sources of demand can fill the space it leaves: transaction fees, and explicit demand for hashrate itself.

And the two are not interchangeable. Transaction fees compensate miners for block space; they do not express the value holders place on the resilience and independence of the security layer itself. A congested mempool pays for inclusion. Nobody in that queue is paying for decentralization.

This is the real meaning of the title above. The market prices the asset every second, with extraordinary sophistication. The security underneath it, nobody buys. Miners produce it speculatively, paid by a formula, and the holders consume it for free. That arrangement was tolerable while the subsidy was rich and the mine had nowhere else to go. The first condition is expiring on a published schedule. The second expired when AI arrived.

Around that calculated reference, a market price already forms: hashrate trading at a premium or discount to its formula value. That premium is small today. But it is the only part of hashrate's price that reflects what anyone actually pays for security, and as the subsidy marches toward zero, it is the part that must grow. Bitcoin's second price has to graduate from a formula into a market.