Whatever happened to the bitcoin miners? 

The most consequential story in cryptocurrency this year has almost nothing to do with cryptocurrency. It concerns electricity, real estate and the discovery that an entire crypto mining industry had spent a decade building the wrong business on top of exactly the right asset.

If you think back a couple of years, you’ll remember that bitcoin miners were the bêtes noires of the energy debate – warehouses of electron-hungry silicon dotted all over the globe, converting megawatts into bitcoin security and coin minting, while environmentalists and grid operators fretted. What nobody fully appreciated, least of all the miners themselves, was that the truly valuable thing they had built had little to do with bitcoin.

It was the electricity plug in the wall.

Today, grid-connected power at industrial scale is the scarcest input in the AI supply chain. The hyperscalers – Microsoft, Google, Amazon and their proxies – have effectively unlimited capital and an insatiable appetite for compute, but a new data centre campus can take five to seven years to energise from scratch.

The bitcoin miners, flush with early 2020s cash from investors, already had the substations, the cooling, the land, the licences and, crucially, the interconnection agreements. They had solved AI’s hardest bottleneck by accident, years before anyone knew it was a bottleneck.

The great defection

Now, in the age of unrestrained AI expansion, a great defection is under way. An estimated $100bn in cumulative AI computing contracts has been announced across the publicly listed bitcoin mining sector, and analysts at CoinShares reckon these companies could derive as much as 70% of their revenue from AI by the end of 2026 (as reported by Coin Insider), up from roughly 30% at the start of the year. The bitcoin mining industry, as a category of public company, will soon earn less than a third of its living from bitcoin.

The individual deals are remarkable. IREN, the Australian-founded miner formerly known as Iris Energy, signed a five-year partnership with Microsoft projected to generate $1.9bn in annualised revenue at an 85% earnings before interest, tax, depreciation and amortisation margin – economics that no amount of bitcoin mining could produce, but that renting its infrastructure and interconnects to AI companies can.

TeraWulf, another bitcoin miner, has contracted some $6.7bn in revenue, potentially $16bn with extensions, from the AI cloud firm Fluidstack. Cipher, which has dropped the word “Mining” from its name, has exited most of its bitcoin operations entirely in favour of a $9.3bn contracted backlog anchored by a 300MW deal with Amazon Web Services. Bitfarms has announced it will wind down cryptocurrency mining altogether by 2027.

These are not diversifications. They are conversions, in something close to the religious sense.

The world’s most powerful landlord?

The most interesting player in all of this is Google, which has developed a mechanism of some elegance (others may describe this less charitably). Rather than buying the miners outright – which would invite tiresome attention from merger regulators – Google guarantees the lease obligations of the AI tenants in the data centre. It has guaranteed $3.2bn of Fluidstack’s obligations to TeraWulf and $1.4bn of the same tenant’s obligations to Cipher, and in exchange for these guarantees it receives warrants for shares – roughly 14% of TeraWulf, and about 5% of Cipher. The trick is that a sub-investment-grade landlord suddenly carries hyperscaler-grade credit, which unlocks the project financing to build the thing in the first place.

This is so tricksy it bears repeating: Google guarantees the tenant’s rent, which lets the data centre landlord borrow against Google’s credit rather than its own (they are new companies; their credit ratings are poor). So the data centre gets built, Google takes equity warrants for its trouble, and the whole arrangement delivers control of AI capacity without triggering a single merger filing to regulators. And so, in this scheme, Google owns no buildings, operates no campuses, but increasingly shapes who may deploy large-scale computing on the American grid. Neat, huh?

Critics have started calling it a virtual utility. The antitrust lawyers will get there eventually; they usually do, about five years too late.

Meanwhile, the equity markets have noticed. Bitcoin mining stocks have handsomely outperformed bitcoin itself this year, even as the token has fallen hard – a divergence that would have been unthinkable in any previous cycle, when miners were simply leveraged proxies for the coin. Bernstein, initiating coverage on the sector, called the miners “the power landlords of AI”, which is the correct framing. The market is now pricing contracted AI power-demand, contract backlog and delivery timelines rather than crypto hashrates and the hope of increased bitcoin prices.

A hashrate collapse

What, then, of bitcoin? Here the story darkens somewhat. Every megawatt redirected to an AI GPU cluster rather than a mining ASIC rig is a megawatt no longer securing the bitcoin network. Global bitcoin mining hashrate has fallen from about 1,160 exahashes per second in October 2025 to roughly 950 by early this year, producing three consecutive negative “difficulty” adjustments (a clever compute mitigation strategy dreamed up by Satoshi, but rarely employed – the last time being during China’s 2021 mining ban).

Public miners sold more than 25,000 BTC in the first quarter alone, liquidating treasuries to pay down debt and fund the pivot to AI energy rental businesses. The remaining miners face production costs north of $70,000 per coin against a spot price that has spent much of the year uncomfortably close to that line. Mining bitcoin has become, for the marginal operator, a risky business.

There is an irony in one of these deals. CoreWeave, the AI cloud darling now signing multibillion-dollar compute deals with OpenAI, began life as an ethereum mining operation. The path from token harvesting to critical AI infrastructure is not a detour. It may be the industry’s actual destiny, visible only in retrospect.

And that is the real lesson. Capital is perfectly agnostic. The machines do not care what they compute, the substations do not care what they power, and the institutional money that flooded into crypto over the past two years was never sentimental about decentralisation or sound money or any of the hotly debated ideology of permissionless transactions and private money. It was, as always, hunting for return, and when a better use for the underlying atoms appeared – electrons, land, transformers – it repriced the entire sector in about 18 months and moved on.

Bitcoin’s security budget now competes with chatbots for electricity, and the chatbots are winning, because Microsoft pays in guaranteed dollars and the bitcoin protocol pays in what is rather sarcastically (but accurately) called “hopium”.

The miners, to their credit, saw which way the current was flowing. Whether bitcoin can afford their departure is a question the protocol will answer in its own time, one downward difficulty adjustment at a time, perhaps even to a point where the security of the network itself might be at risk.

Steven Boykey Sidley is a professor of practice (ex-JBS, University of Johannesburg), a partner at Bridge Capital and a columnist-at-large at Daily Maverick, Daily Friend and Currency. His new book “It’s Mine: How the Crypto Industry is Redefining Ownership” is published by Maverick451 in South Africa and Legend Times Group in the UK/EU, available now.

ALSO READ:

• How AI is impacting crypto
• Why South Africa is tipped for a data centre surge
• When crypto faith breeds financial fiction

_{Top image collage: Rawpixel; Currency.}