Bitcoin Hashprice Breakeven, Hashrate Still Rising

Bitcoin's network hashrate has climbed back to roughly 956 EH/s as of Aug 8, up from an 893 EH/s trough in mid-June, and Saturday's difficulty retarget moved higher instead of lower. That's bitcoin hashprice sitting at breakeven with hashrate rising anyway, the opposite of what a mining sector this thin should be doing. Hashprice, the standard measure of daily miner revenue per unit of computing power, was around $32.10 per PH/s/day as of Aug 3, a level that sits at or below the all-in breakeven point for older, less efficient rigs. Normally, thin margins push hashrate down as marginal miners switch off. This time the aggregate number is rising. The explanation isn't that mining suddenly got more profitable. It's that the miners adding hashpower are being funded by something else entirely: AI data-center contracts.

The answer is compositional, not universal. Bitcoin's mining fleet is not one uniform business, it's a spread of operators with wildly different costs. The most efficient machines, the sub-15 J/TH generation running on power under roughly $0.08/kWh, still clear real margin even with hashprice pinned near breakeven. Higher-cost, older rigs don't. So the same $32/PH/s/day environment that's forcing weaker miners offline is simultaneously profitable enough for the strongest operators to keep expanding. Add those two forces together and the network-wide hashrate chart can rise even as the median miner's economics stay ugly. The headline number hides a split: efficient public miners scaling up, inefficient private ones scaling down.

The bigger driver isn't mining profitability at all, it's capital. The same public miners posting the strongest fleet efficiency have also become the industry's biggest AI and HPC landlords, signing hosting deals that dwarf their bitcoin revenue. Hut 8 has more than $26.6B in disclosed AI/HPC contract value, TeraWulf about $22.7B (a $19B, 20-year Anthropic lease signed this quarter plus its existing Fluidstack hosting agreements), IREN roughly $9.7B including a Microsoft deal, and CleanSpark about $6.6B. That's over $65B in committed hosting capacity across just four names. Building out the power and site infrastructure to serve those contracts means acquiring or expanding megawatts of capacity, and in most cases that capacity gets pointed at bitcoin mining first, either as a bridge activity while AI conversion is finalized or as a parallel revenue stream once it's live. The hashrate expansion is a byproduct of the AI buildout, not a bet on bitcoin mining economics improving.

That distinction matters because it changes what the hashrate chart is measuring. A rising network hashrate is normally read as a vote of confidence in bitcoin mining itself. Right now it's closer to a side effect of a much larger capital cycle. The miners driving the increase would likely be expanding their footprint even if hashprice were lower, because the AI hosting revenue is the point.

The miners you don't hear about. Smaller, private, higher-cost operators without AI contracts to lean on are the ones actually exiting or idling machines at today's hashprice. They don't show up individually in headlines, but their retreat is why the recovery in hashrate looks compositional rather than broad-based. Some of that marginal fleet keeps running anyway, even at thin or negative margin, because the capital that built it is sunk: unplugging a rig doesn't recoup the capex already spent, so a machine can keep mining at a loss on a cash basis longer than pure profitability math would suggest. That explains why hashrate hasn't fallen further, not why it's climbing.

For bitcoin's security budget, the aggregate hashrate number still looks reassuring. But the concentration underneath it matters. If a growing share of network security rides on a handful of operators whose real business model is AI hosting, bitcoin's defense becomes more dependent on decisions made in AI capital markets than on mining economics itself. For miner stocks, the read is more direct: names like Hut 8, TeraWulf, IREN and CleanSpark are increasingly AI infrastructure plays wearing a mining-stock ticker, and their share prices are likely to track AI-hosting revenue recognition and hyperscaler contract news at least as much as bitcoin's price or hashprice.

Expect the bifurcation to continue into the fourth quarter: aggregate hashrate and difficulty likely keep grinding higher as AI-diversified miners scale their footprint, while smaller and higher-cost miners keep consolidating or shutting down. The next scheduled difficulty retarget, around Aug 22, is the near-term tell for whether the Aug 8 increase was a turn or a blip. Two of the four names have already reported Q2: Hut 8 on Aug 4 and TeraWulf on Aug 5, and TeraWulf's HPC/AI revenue already overtook its bitcoin-mining revenue for the first time, an early confirmation of the diversification story rather than a mining-profitability rebound. IREN and CleanSpark's reports later this month are the remaining tell, watch whether AI-hosting revenue keeps closing the gap on mining revenue there too.

The bull case is a meaningful bitcoin price recovery, helped by a dovish Aug 12 CPI print or a September Fed cut, that lifts hashprice back above breakeven for more of the fleet, turning today's compositional recovery into a genuine one. The bear case is the opposite: bitcoin stays range-bound while difficulty keeps rising on AI-miner buildout, squeezing marginal operators harder and pushing the network's security base toward a smaller set of AI-subsidized companies. The risk to watch isn't bitcoin's price at all, it's whether any hyperscaler counterparties behind those AI contracts delay or restructure them, leaving miners that bet their hashrate expansion on that capex with a funding gap mining alone can't cover.