Loading today’s stories
Invest1 distinct publisher3 min readUpdated
The listed cohort's realized hashrate fell 13.4 percent in six months against a 10.6 percent network decline. Two of its members now earn most of their revenue from something other than mining.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
The publicly listed bitcoin mining cohort tracked by TheEnergyMag delivered a combined realized hashrate of 368.3 EH/s in the fourth quarter of 2025, 344.4 EH/s in the first quarter of 2026 and 319.0 EH/s in the second, a 13.4 percent decline in six months [1]. Over the same stretch the network's quarterly average went from 1,071 EH/s to 957 EH/s, a 10.6 percent fall [2], which means the listed operators shrank faster than the thing they are supposedly a levered bet on [3].
The reason is on the revenue lines. Core Scientific booked $136.7 million of colocation income in the second quarter against $27.5 million from bitcoin mining, with colocation reaching 83 percent of total sales, up from 67 percent the prior quarter [5]. That is roughly five times the mining number [7] and a 16 point shift in a single quarter [4]. TeraWulf reported $31.9 million of high-performance computing lease revenue, 71 percent of its total, against $12.8 million from mining [6]. For both companies, non-mining activity has already displaced bitcoin production as the primary revenue driver [7].
Elsewhere the transition is earlier. Riot Platforms recorded $23.2 million of data-center revenue against $113.7 million from mining [8], about 17 percent of those two lines combined [5]. Bitdeer took $14 million from AI cloud services against $197.1 million from mining-related operations [9], roughly 7 percent [6]. Hut 8 and MARA showed smaller compute contributions, and Cipher and Keel Infrastructure had not yet recognised any HPC revenue at all [10].
Three months earlier the arithmetic still balanced. Core Scientific, IREN, Cipher, TeraWulf and Keel were scaling back while Bitdeer, MARA, Riot and American Bitcoin absorbed the displaced share, leaving the cohort roughly flat [19]. That offset has now failed [14]. Strip out Bitdeer, which raised realized hashrate 44 percent to 63.0 EH/s on its own SEALMINER line [11][13], and the rest of the cohort fell 21.2 percent, from 324.6 EH/s to 255.9 EH/s [12] - double the network's rate of decline [3]. Bitdeer alone is now close to a fifth of the listed cohort's hash [2], and by June reported 73 EH/s of self-mining plus 15.9 EH/s of co-mining capacity and 990 bitcoin mined in the month, 388 percent more than a year earlier [13]. One company is carrying the category's growth statistics.
Cango shows how fast the exit can be. It entered mining in late 2024, deployed 50 EH/s during 2025, then began decommissioning less efficient machines, leasing hashrate out and relocating capacity to cheaper regions [15]. Realized hashrate went from 44.8 EH/s to 31.3 EH/s to an estimated 16.5 EH/s, down 63 percent in six months [16]. Keel Infrastructure finished decommissioning all of its US bitcoin mining in the second quarter to clear ground for data-center construction, with Canadian mining continuing through the phased transition [17].
The awkward part is the gap. Replacement revenue has not fully materialised [18], so investors holding these names are paying for capacity that no longer mines and does not yet bill a tenant. This cohort's post-2021 build was a response to China removing roughly half the network's hashrate, with the bottom at 57.5 EH/s in June 2021 before recovery by December [20]; that cycle is now unwinding.
Watch whether Cipher and Keel recognise their first HPC revenue, whether Bitdeer's SEALMINER expansion continues to mask cohort contraction, and whether index and screen providers keep classifying companies like Core Scientific and TeraWulf as miners when most of the money arrives from leases [5][6][10][11].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
TheEnergyMag calculates the tracked cohort of public miners delivered a combined realized hashrate of 368.3 EH/s in Q4 2025, 344.4 EH/s in Q1 2026 and 319.0 EH/s in Q2 2026, a 13.4 percent decline over six months.
The Bitcoin network's quarterly average hashrate fell from 1,071 EH/s to 993 EH/s and then to 957 EH/s, a 10.6 percent reduction.
Public companies contracted faster than the network as a whole; listed miners reduced dedicated computational capacity more rapidly than the broader network as operators shifted electricity and data-center resources toward AI and high-performance computing, prioritising more predictable income over cryptocurrency production.
Core Scientific recorded $136.7 million in colocation income in Q2 versus $27.5 million from bitcoin mining, with colocation accounting for 83 percent of total sales, up from 67 percent in the prior quarter.
TeraWulf's high-performance computing lease revenue reached $31.9 million, or 71 percent of overall revenue, against $12.8 million from mining.
For Core Scientific and TeraWulf, non-mining activities have already surpassed bitcoin production as the primary revenue driver.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Quantitative but single-sourced and partly self-computed
The numbers are specific and internally consistent — per-quarter cohort and network hashrate, per-company revenue splits, per-company capacity — and several rest on company reported results. But the entire cluster is one article from one publisher relaying a third party's own cohort calculation, with no methodology disclosure, no independent corroboration, and at least one key figure (Cango's Q2 capacity) explicitly an estimate.
Real revenue-scale pivot, concentrated in two operators
Adoption of the AI/HPC pivot is observable in dollars and in retired capacity: Core Scientific at 83 percent colocation share, TeraWulf at 71 percent HPC share, Keel's complete US mining shutdown, Cango's capacity cut. It is not yet broad — Riot is near 17 percent, Bitdeer near 7 percent, Hut 8 and MARA smaller, and Cipher and Keel recognise no HPC revenue at all — and the article states replacement revenue has not fully materialized.
Mildly overstated at the cohort level
The framing generalises a pivot that is revenue-dominant at only two named operators, and the aggregate 'contracting faster than the network' conclusion depends on an undisclosed cohort methodology plus one estimated data point. Offsetting this, the article is restrained where it matters: it concedes replacement revenue has not fully materialized, separates advanced from early movers with figures, and explicitly notes the contraction lacks a single catalyst unlike the 2021 China ban.
Self-reported figures plus an analyst promoting its own dataset
Two incentive layers are visible in the supplied material. Company-level revenue splits are the operators' own disclosures, and firms mid-pivot have reason to foreground AI and colocation revenue over mining. The aggregate hashrate series is TheEnergyMag's proprietary calculation, republished by a fintech/crowdfunding outlet, which gives the analyst an interest in the cohort narrative it originated three months earlier and is now updating.
Directionally credible, weakly corroborated
The direction — public miners shrinking hashrate while colocation and HPC revenue rises at the most advanced operators — is supported by multiple concrete, per-company data points and is consistent across revenue and capacity evidence. Confidence is held down by single-publisher sourcing, an undisclosed cohort methodology, one estimated figure, and the absence of any mining-economics data (price, hashprice, margins) to test the stated cause.
invest
IREN hands Microsoft its first 50MW, and the $9.7B contract starts having to be true1 distinct publisher
product
Iren hands Microsoft its first 50MW at Childress, and a schedule to be judged by1 distinct publisher
product
Two auctions, one week: the disclosure gap buried in Situational Awareness's $30bn July1 distinct publisher
product
Keel turns off all four US bitcoin mines and waits until 2027 for the first data center dollar1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 16, 2026