Bitcoin Miners Are Now Underwater — and They're Pivoting to AI to Stay Afloat

The math just stopped working

Here's a number that should make any bitcoin bull pause: JPMorgan figures it costs about $78,000 to mine a single bitcoin right now. Bitcoin is trading around $62,500. That's not a rounding error — it's a gap of more than $15,000 per coin, and miners are on the wrong side of it.

Normally bitcoin's "production cost" — basically the electricity, hardware, and overhead it takes to mine one coin — acts like a soft floor. When prices dip near it, weaker miners switch off, the network gets less competitive, and the economics rebalance. What's notable now is that price has fallen below that cost. JPMorgan's own word for it is blunt: mining economics have "worsened."

So what does that mean if you hold bitcoin or mining stocks? In the short run, miners are bleeding. They're either selling coins at a loss to keep the lights on, or finding another way to make money off the one thing they actually own a lot of: power and real estate wired for heavy compute.

When the operators closest to the economics stop expanding hashrate and start building GPU clusters, that's a signal worth more than any price target.

The pivot hiding inside the May numbers

Four big miners — Bitdeer, BitFuFu, Canaan, and CleanSpark — pumped out a combined 1,859 BTC in May. But the headline number buries the real story: most of them are quietly restructuring around AI rather than around mining more bitcoin.

Look at the tells. CleanSpark mined 671 coins but sold 654 of them (404 at spot, 250 through options), netting just 17 BTC of growth — and it's hiring bankers and data-center execs to fund an AI buildout in Georgia and Texas. Bitdeer's monthly output has nearly quintupled in a year, yet it held just 171 BTC at month-end because it keeps converting mined coins into cash to feed its AI strategy, including new Nvidia GB300 clusters. Both companies are keeping their core mining hashrate (the raw computing power pointed at bitcoin) essentially flat while spending elsewhere.

Not everyone's running the same play — BitFuFu actually tripled its self-mining output, calling it a deliberate bet on cheap coins at these prices. But the pattern among the biggest operators is clear, and that's the "so what": the people who run mining companies are voting with their balance sheets, and many are voting against mining more bitcoin at these prices. They'd rather rent their power and chips to AI customers. When the operators closest to the economics stop expanding hashrate and start building GPU clusters, that's a signal worth more than any price target.

Why Texas — and idle megawatts — are the whole game

The pivot only works if these companies have something AI buyers actually want, and it turns out they do: power. IREN, for example, sits on a roughly 6 gigawatt global power portfolio and is reportedly using only about 10% of it. That's a mountain of unused capacity in a world where AI data centers are starved for electricity — Jefferies thinks IREN's AI cloud business could outpace its data-center leasing, and sees 30% upside in the stock.

Texas is where this collides with policy. Data centers now account for nearly 90% of the 438 GW of large-load demand on the state's grid, and ERCOT (the body that runs that grid) is rolling out a new allocation framework that could favor miners who've already converted into data-center operators. Translation: the companies that pivoted early may get first dibs on the power connections everyone now wants.

For investors, this reframes what a "bitcoin miner" even is. The ones that survive a sub-cost stretch may not be the ones grinding out the most coins — they'll be the ones sitting on cheap, grid-connected power they can re-rent to AI. What to watch: whether bitcoin climbs back above that ~$78,000 production cost, how quickly idle gigawatts like IREN's get filled, and whether the Texas framework actually lands in miners' favor. The hashrate race is fading. The megawatt race is on.