Bitcoin miners are unplugging machines worth an estimated 1.5 billion dollars — not because mining broke, but because a different business came knocking. According to Miner Weekly, a newsletter by Blocksbridge Consulting, the hashrate that publicly traded miners shed in the first half of 2026 equals roughly 1.5 billion in hardware investment, even assuming a bargain price of just 20 per terahash.
By Marcus Johnson | October 4, 2026
The Hook: Machines Paid For, Then Parked
Here is the uncomfortable part for investors. The mining capacity now being switched off was already financed. Shareholders paid for those machines during the 2024 and early-2025 buying spree. At some sites, equipment was marked down within months of starting production, which raises an awkward question: how much of that investment will mining ever recover before artificial intelligence workloads take over the buildings entirely?
A earlier Miner Weekly issue estimated that public miners lost 75 EH/s of realized hashrate while their directly reported high-performance computing and AI revenue rose 52 percent quarter over quarter. Think of it like a restaurant ripping out half its kitchen to build a delivery hub — the new business is growing fast, but the old kitchen was not yet paid off.
The Evidence: Write-Downs Are Already on the Books
The accounting is now catching up with the pivot. A review by TheEnergyMag of twelve tracked companies identified approximately 1.1 billion in asset impairments and held-for-sale markdowns during the first half of 2026. IREN and Core Scientific (NASDAQ: CORZ) accounted for almost 89 percent of that total.
- IREN — reached 50 EH/s in June 2025, then recorded roughly 695 million in impairments and held-for-sale markdowns between January and June 2026, with substantial charges tied to mining assets displaced by AI conversions.
- Cipher Mining — its Black Pearl site began mining in mid-2025. By year-end, after agreeing to convert the site to high-performance computing, Cipher recorded a 96.1 million markdown on Black Pearl’s mining machines. Those same machines generated 57.9 million in revenue during 2025 — the write-down exceeded the site’s first months of income.
- TeraWulf (NASDAQ: WULF) — generated about 53 million in HPC leasing revenue in the first half while paying roughly 131 million in cash interest across the company, showing the gap between the AI revenue ramp and the cost of financing it.
To be clear, not every charge is an AI conversion cost. Core Scientific attributed its major mining impairment to deteriorating mining economics, and the broader total includes other asset categories. The honest takeaway is a costly reassessment of mining investment — not proof that every dollar was sacrificed for AI.
The Core Conflict: Less Hashrate, Tighter Money
The spending that created this mess was recent and enormous. A March 2025 Miner Weekly report counted nearly 5 billion in equipment and infrastructure spending across one reporting cycle, and companies that separately disclosed hardware purchases spent more than 3 billion in 2024 alone. That arms race helped push Bitcoin’s network into zetahash territory in 2025 — every miner buying faster machines to defend their slice of production.
Now the money environment is turning against the converters. According to an excerpt from The Information’s reporting, Société Générale and SMBC have become more selective about data center financing, while MUFG is also stepping back. Meanwhile the Federal Reserve recently raised its benchmark rate by a quarter point to a range of 3.75 to 4 percent, making new borrowing and refinancing more expensive. Miners caught mid-conversion face the squeeze from both sides: the mining revenue is shrinking before the AI revenue fully replaces it, and the lenders financing the replacement want stricter terms.
Market Implications: What This Means for Your Bitcoin
For regular Bitcoin holders, the story cuts two ways. Bitcoin trades around 84,800 as of Sunday, according to CoinGecko data. Publicly traded miners stepping back removes some of the corporate hashrate growth that crowded out smaller operators — a smaller arms race can mean steadier economics for whoever keeps mining. On the other hand, mining stocks are no longer pure Bitcoin plays, so their share prices increasingly follow AI data center sentiment rather than the Bitcoin price itself. If you own miner equities expecting Bitcoin leverage, read their conversion plans closely.
There is also a political shadow over the whole sector. New York announced a pause on state environmental permits for new hyperscale data centers, Massachusetts added approval requirements, and Chicago’s mayor proposed a year-long moratorium. Existing mining sites could actually become more valuable where they offer usable power access — but conversions still need technical upgrades and regulatory sign-offs, and delays are costly when the old revenue stream has already been retired.
The Verdict
AI may ultimately be the better use of these sites — more power per dollar, more diversified customers. But the first half of 2026 shows the transition is not free. Investors financed the machines being unplugged, the write-downs are landing now, and the financing for what comes next is getting pricier. The return on the AI pivot will depend on three things: how much value is recovered from old equipment, how much the rebuild costs, and how long lenders must be paid before the new capacity starts earning. For now, the Great Unplug is a reminder that even in crypto, there is no shortcut past the bill.
The cryptocurrency market remains highly volatile. This article is for informational purposes only and does not constitute financial advice.
1.5 billion in machines just parked. shareholders paid for those rigs during the 2024 buying spree and now they sit idle so the sites can serve ai customers. brutal read
welcome to the pivot. 75 eh/s gone while ai revenue jumped 52 percent qoq. market will cheer this right up until it doesnt
+52 percent ai growth against a 1.5b paper writeoff, the market will do the math eventually and it is not kind to hash rate stories
that 20 per terahash price assumption feels generous tbh. secondhand auction prices on older models are well below that, so the real writeoff is probably bigger than 1.5b
do you have actual auction data for that? genuinely asking, trying to size my mara position
still waiting on someone to post actual secondhand rig listings to settle this. until then 1.5b reads like a floor, not a ceiling
rigs_andy liquidation auctions tell the story, older s19 batches cleared well under 10 per th months ago. 1.5b flatters the balance sheets if anything
cleared under 10 per th months ago and its worse now. ai buyers take the whole site, nobody is bidding on jacks and fans separately
^ this. 20 per terahash is the optimistic case, actual secondhand pricing on older rigs makes the real writeoff worse than 1.5b
agreed, secondhand s19s cleared around 8 per th in spring. at that pricing the markdown is closer to 2b and nobody models it
second this. below 10 per th the 96m figure starts looking like the optimistic scenario, not the floor
The restaurant analogy in this piece is doing a lot of work. Shareholders funded a hash rate story and ended up with a datacenter landlord.
Milo Berg and landlord tenants sign 10 year contracts. these machines depreciated in months, the ai leases are the only durable asset left on those sites
the black pearl numbers are the cleanest version of this. 96.1m markdown on machines that pulled 57.9m of revenue the entire time they ran
joule_per_th 57.9M lifetime revenue against a 96M markdown is a tombstone inscription. growth deck did its job tho, retail bought the pivot slide by the gallon
57.9m of lifetime revenue against a 96m markdown is the whole industry in one sentence. the growth story was the actual product
thats the math nobody wants printed on a slide. the black pearl machines never paid for themselves even before the pivot
the 57.9m lifetime revenue number makes it worse. those rigs were underwater before the ai pivot was even a slide deck
iren and corz being 89 percent of the 1.1b in impairments is the part nobody flags. this pivot story is basically two companies
so anyone reading this as an industry wide markdown is doing the bear case work for free. two balance sheets are not the whole sector
If the AI leases come with 10 year take-or-pay terms this math works. If they are 2 year trials with break clauses, that 1.5b of shelved rigs comes right back online at a loss.
20 per terahash valuation on gear that cleared under 10 at auction means the real number is closer to 750m. the ai leases better carry the whole load
secondhand auction comps are the only real price discovery in this whole writeoff debate. site level AI deals clear fine, standalone rigs are nearly unbid
the unbid part is what gets me. listing a rig and getting zero asks means the writeoff isnt conservative, its just honest
Unplugging 1.5b of machines to lease the space back at AI margins only works if the contracts stick. Every miner is now a landlord with a single tenant problem.
Bjorn Li single tenant problem is the phrase investors will repeat in a year. an AI lease is only as good as the counterparty and hyperscalers know they hold all the leverage