CoinShares: Listed BTC miners fell below cash breakeven in

CoinShares’ Q2 2026 mining readout argues listed bitcoin miners slipped below cash breakeven in aggregate and are reallocating power and capex toward AI and high-performance computing data-center revenue. The report estimates AI/HPC profits run about three times higher than BTC mining and projects at least 35 EH/s of listed hashrate could be minimized or shut this quarter.

Key Takeaways

CoinShares said Q2 2026 conditions “pushed the listed mining sector below cash breakeven in aggregate.”
On CoinShares’ estimates, AI/HPC generated about US$1.5 million in annualized profits per MW versus about US$0.5 million for BTC mining.
CoinShares projected “at least 35 EH/s” of listed miners could be leaving, minimizing, or halting mining this quarter, about ~4.7% of current network hashrate.
The next BTC mining difficulty adjustment scheduled for Saturday (19) was expected to rise by more than 5% to over 134 trillion hashes.

CoinShares: Listed Miners Dip Below Cash Breakeven as AI/HPC Becomes the Priority

CoinShares’ Q2 2026 mining report framed the listed mining cohort as operating on the wrong side of the cash line. The firm said Q2 conditions “pushed the listed mining sector below cash breakeven in aggregate,” even as BTC’s price later rebounded from the US$58,400 level it held at the end of the quarter.

The mechanism CoinShares is pointing at is straightforward: when the block subsidy dominates revenue and the cost base is mostly fixed power plus hardware depreciation, small moves in revenue per unit of compute can flip the sector from marginally profitable to cash-negative. CoinShares also tied the pressure to fee weakness, noting transaction fees in Q2 were “persistently below 1% of block rewards,” leaving miners with little buffer when economics tighten.

That backdrop is why the report treats the AI pivot as more than a narrative hedge. CoinShares wrote that BTC mining “is not dying,” but argued operators are shifting focus “to capture what is a lucrative and seemingly secular tailwind” in AI/HPC.

The Numbers Behind the Pivot: US$1.5M/MW vs US$0.5M/MW and a Potential 35 EH/s Shift

CoinShares’ core claim is a relative-return problem, not a philosophical one. It estimated AI/HPC generates annualized profits of approximately US$1.5 million per MW, compared with about US$0.5 million for BTC mining. MW here is the power capacity miners control, which is the real scarce input they can redeploy between hashing and data-center workloads.

For traders, the key translation layer is how that profit gap flows into operational decisions. Hash price is the revenue generated per unit of computing power, and CoinShares said hash price sank to record lows in H1 2026, leaving “many ASIC mining rigs performing below breakeven.” An ASIC is the specialized hardware built to do proof-of-work hashing efficiently. When hash price compresses, the least efficient ASICs get turned off first, and new ASIC orders stop making sense.

CoinShares said some miners were “happily paying for the privilege” of canceling orders for new ASIC rigs, a detail that matters because it implies a capex freeze rather than a temporary throttle. It also laid out a near-term hashrate implication: “at least 35 EH/s” (exahashes per second, where 1 EH/s equals 10^18 hashes per second) could leave listed operators as they minimize or halt mining in favor of AI/HPC. CoinShares pegged that at roughly 4.7% of current network hashrate, large enough to matter at the margin without implying the network is breaking.

The report also pointed to how public markets are already pricing the split. Miners with “contracted AI or HPC capacity” traded at an average 12.9x EV/NTM versus 3.7x for miners without such contracts. EV/NTM is enterprise value divided by next-twelve-months expected results, and the dispersion is the equity market’s way of saying predictable data-center cash flows are being valued differently than cyclical block-subsidy exposure.

Mining’s Near-Term Squeeze: Difficulty Set to Jump Above 134T as Fees Stay Thin

The immediate headwind CoinShares flagged is difficulty. Mining difficulty adjustment is the periodic retargeting that keeps block times roughly constant by making it harder or easier to find a block depending on how much hashrate is online. The next adjustment scheduled for Saturday (19) was expected to increase difficulty by more than 5% to over 134 trillion hashes.

A higher difficulty number means the same fleet produces fewer coins for the same power burn, unless BTC price or fees rise enough to offset it. CoinShares’ point about fees staying thin matters here because it removes the usual “maybe fees save the day” argument. With fees “persistently below 1% of block rewards” in Q2, miners are still mostly living on subsidy economics.

CoinShares also framed the macro tape as unhelpful in the same week. BTC’s fiat price fell Tuesday as it became clear the U.S. Senate’s digital asset market structure bill, the CLARITY Act, lacked votes to clear its first procedural hurdle. The report’s broader claim is that even if BTC rebounds, “a BTC recovery is unlikely to reverse the AI transition,” because AI/HPC revenue is expected to accelerate through H2 2026.

There is one partial release valve for miners that stay committed to hashing. CoinShares said ASIC prices are “likely to soften” as demand falls, which “should improve upgrade economics for miners that remain committed.” That is less a bullish catalyst than a redistribution mechanism: weaker hands stop ordering, secondary hardware gets cheaper, and the remaining operators can lower their cost per hash.

What Comes Next for CoinShares: BTC miners pivot to AI/HPC

The next clean datapoint is the Saturday (19) difficulty adjustment itself. If difficulty rises more than 5% to above 134T as expected, the question is whether network hashrate follows CoinShares’ implied direction afterward, or whether miners absorb the hit and keep machines online.

CoinShares’ own follow-up will matter more than most earnings calls. The firm said AI/HPC revenue “should accelerate through H2 2026,” with the run rate expected to “more than double” by the time it publishes its Q3 retrospective. That retrospective is also where the projected ≥35 EH/s listed-miner pullback should start to show up in disclosures, either as explicit mining curtailments or as power capacity reallocated to data-center contracts.

The report’s numbers are directionally clear, but the packet does not include a direct link to CoinShares’ full methodology for the US$1.5M/MW versus US$0.5M/MW profit estimates. That leaves room for the usual modeling risk: contract terms, utilization, and capex assumptions can swing per-MW profitability quickly, especially when AI/HPC revenue is tied to customer concentration and buildout timelines.

My Read: The AI Pivot Looks Like a Structural Re-Rating, Not a One-Quarter Trade

The threshold that matters is whether CoinShares’ per-MW profit gap holds up once the Q3 retrospective lands. If AI/HPC really clears roughly US$1.5 million per MW in annualized profits against roughly US$0.5 million for mining, miners are not “diversifying,” they are rationally reallocating their scarce input, power, toward the higher-return line.

The real test is whether the network absorbs a listed-miner pullback of ≥35 EH/s while difficulty is still rising toward >134T and fees remain stuck below 1% of block rewards. If that combination persists, the AI/HPC pivot stops looking like a hedge against a bad quarter and starts looking like the new base case for how listed miners are valued and how much incremental hashrate they are willing to fund.

Sources

CoinGeek