Bitcoin Miners Are Bidding Against AI for the Same Megawatts

Bitcoin is sitting near $77,000 today, down roughly two-fifths from the $125,000 high it printed within the past year. That is the part everyone can see. The less visible part is that mining companies spent 2026 setting record hash rates anyway, and are now running on what analysts call the thinnest margins in years. The two facts belong together, but not for the reason you might guess. It is not really a crypto story at all. It is an electricity story.

A bitcoin miner is, at bottom, a power company with expensive computers bolted on. It buys megawatts, pushes them through chips that grind hashes, and each hash is a lottery ticket on a share of the block reward. Electricity is 60% to 85% of what the whole business costs, so the only question that ultimately matters is how cheaply a miner can buy power. That is the asset: cheap, located, interruptible electricity. Everything else is noise.

The deal that makes mining work

The energy relationship is the part the headlines usually get backward. For years mining was cast as an electricity hog racing a small country's consumption. What that framing misses is that miners are the least valuable electricity customer on the grid — which is exactly why grids want them.

A miner can power down almost instantly and without damaging its machines, so utilities use mining as a flexible valve. When wind and solar overproduce and nobody can use the power, miners buy it at near-zero or negative prices; when the grid is stressed and homes need every watt, miners shut off and, in many markets, get paid to do it. Each side gives the other something no one else will. That is why the industry's generation mix has flipped to majority sustainable — about 52% of mining electricity now comes from zero-emission sources — not out of virtue, but because stranded, intermittent power is the cheapest power there is.

The payments for switching off have become real money. Riot Platforms earned $30.6 million in electricity-curtailment credits in the third quarter of 2025, up 147% year over year. In Texas, grid services account for an estimated 10% to 15% of miner profits. Think of curtailable power as a subsidy baked into the model: miners accept being the grid's shock absorber in exchange for electricity that costs a fraction of the industrial average.

The price floor is a knife's edge

That energy bargain is the first half of the balancing act. The second half is the block reward. Since the April 2024 halving cut new issuance in half, the revenue each machine can earn — "hashprice" — has collapsed. In 2026 it sits at roughly $35 per petahash per day, described as the lowest in years. The cause is arithmetic, not sentiment: every new ASIC adds compute, difficulty climbs to keep block times steady, and the same fixed reward gets split across more machines. Energy efficiency gains feed straight back into more hash rate, which pushes per-machine revenue back down. The industry has been running on this treadmill for years, and it never gets easier.

The result is a break-even that hangs on fractions of a cent per kilowatt-hour. The newest, most efficient machines can still profit on power at around $0.10 to $0.12 per kilowatt-hour. Older-generation ASICs need below $0.055, which is below standard industrial rates. With bitcoin near $77,000, the industry's average production cost is about $74,000 to $77,000 a coin and all-in costs exceed $100,000 for many operators. Miners without the cheapest power and the newest chips are, today, operating in the red. This is a system that balances itself the hard way: when price falls far enough that marginal machines lose money, they shut down, hash rate drops, difficulty "blinks lower," and margins recover just enough to invite the next wave of machines back in.

The rival bidder for the same watts

Here is the change that makes this moment different from past downturns. Bitcoin is no longer the only, or even the best, customer for the megawatts miners hold. AI is. Analysts put AI workload revenue at roughly $25 per kilowatt-hour against about $1 for bitcoin mining — a tenfold gap on the same physical infrastructure. That disparity is why public mining companies have announced more than $70 billion in cumulative AI and high-performance-computing contracts, and why Core Scientific alone has a 12-year, $8.7 billion hosting deal with CoreWeave across 500 megawatts, an AMD partnership for up to 2.5 gigawatts, and Pecos, Texas plans stretching to 1.5 gigawatts. Bitfarms converted a site and says it plans to exit crypto mining entirely by 2027. Analysts project listed miners could derive up to 70% of revenue from AI by the end of this year, up from about 30% at the start.

Read that as a repricing of what these companies are. What they have never really owned is a special claim on bitcoin — they own energized land, power contracts, and the patience to interconnect that a hyperscaler desperately needs and cannot conjure quickly. Blockchain mining got first claim on that scarce resource because it was willing to run on stranded, flexible power no one else wanted. Now there is a richer tenant knocking.

What this means for the trade

For a retail investor, the practical consequence is that "bitcoin mining stock" is quietly becoming a misleading label. A miner's share price now moves on power contracts, AI counterparties, and hosting revenue as much as — by end of 2026, potentially more than — the price of bitcoin itself. If you bought a miner to get leveraged bitcoin exposure, you are buying an asset that is steadily de-levering from bitcoin. If you are evaluating a miner on its energy asset, the honest question is which tenant clears the megawatts, and today that answer is AI.

The structural insight cuts the other way too. The stranded-power, demand-response relationship is a genuine moat — it is why these firms can source electricity at all and why the business has survived repeated price collapses. But it is a thin-margin moat. A miner's profitability is bounded on one side by the block reward split across the whole network, and on the other by whatever AI will pay for the same watts. Bitcoin mining is not going away — it is the flexible, interruptible, cheap-end buyer that no grid can easily replace. But it is no longer the marginal bidder for the electricity that matters. The balancing act has resolved: the same megawatts are now an auction, and AI is currently paying more.

The lasting lesson is to stop reading these companies through the coin price alone. Bitcoin tells you how much the mining tenant can pay. The energy market tells you who is actually going to get the building. Right now those two answers are diverging, and the divergence is the trade.