Bitcoin Mining Profitability Squeezed as Miners Pivot Hard to AI and Data Centers
LTB Bitcoin mining profitability is getting squeezed again just as more public miners shove money, power, and ambition into AI and data centers. The brief bounce in economics looks like a false dawn, not a turn in the weather.
- Difficulty is rising again
- BTC price weakness is crushing margins
- AI and HPC are becoming the escape hatch
The improving economics of block reward mining on the Bitcoin network didn’t last long. The next difficulty adjustment on May 15 is forecast to rise nearly 3 points to 136.1 trillion hashes, just as BTC slipped from about $82,000 to around $79,500 after U.S. inflation data. For miners, that’s a nasty little double-whammy: higher competition for blocks and a weaker dollar price for the reward they’re trying to earn.
Difficulty is Bitcoin’s built-in throttle. When more computing power joins the network, the protocol makes it harder to find the next block, so the average block time stays around 10 minutes. That’s great for keeping the system stable. It’s not so great for miners trying to keep the lights on and the spreadsheets out of therapy. Higher difficulty means more hashpower, more electricity, and more hardware just to keep earning the same slice of the block reward.
And that slice is getting thinner in fiat terms. The current average all-in cost of mining one BTC is roughly $3,000 higher than the coin’s present price, according to the numbers cited here. That’s the sort of margin compression that turns a growth story into an exasperated boardroom discussion about “strategic optionality,” which is corporate language for “we need a new plan and fast.”
The result is a broader shift across public mining companies away from pure Bitcoin mining and toward artificial intelligence, high-performance computing (HPC), and data center infrastructure. That’s not some betrayal of the orange coin. It’s what happens when electricity, land, and grid access become more valuable selling compute to AI customers than grinding out ever-less-profitable sats.
Bitcoin miners pivoting to AI is no longer a side storyline. It’s becoming the main event. Sun sets quickly on BTC mining economics false dawn
Bitdeer and Cango are already trimming the old model
Bitdeer offers a clean example of the new reality. Its April AI cloud revenue rose 60% month-over-month to $69 million, while its GPU count nearly doubled from 2,128 to 4,184 in the same month. It still self-mined 783 BTC in April, up from 661 in March, but it ended the month with just 73 BTC on hand.
That mix says plenty. Bitdeer is still mining, but it’s also clearly building a business around AI cloud services and GPU infrastructure. In plain English: the rigs that once mostly chased block rewards are being pointed toward customers who will pay for compute in a less volatile way than the Bitcoin network’s next difficulty adjustment.
Cango is following a similar path, though more cautiously. Its self-mining hashrate fell to 20.4 EH/s as of April 30, down from 27.9 EH/s in February. The company reported 230 BTC mined in April, though it didn’t disclose a March figure.
That’s a meaningful pullback. Hashrate, for newer readers, is the total computing power used to secure the network and compete for mining rewards. When a company reduces its self-mining hashrate, it is effectively stepping back from the hardest part of the business. Not necessarily exiting, but absolutely telegraphing that the old playbook is getting less attractive by the month.
MARA is still mining, but the priorities are changing
MARA, one of the biggest public Bitcoin miners, is showing the same tension between commitment and pragmatism. Its Q1 revenue fell 18% year-over-year to $174.6 million, while net losses more than doubled to $1.26 billion. The company’s BTC treasury also dropped from 53,822 BTC to 35,303 BTC, and it cut 15% of staff tied to mining-related work.
For readers unfamiliar with the term, a BTC treasury is the amount of Bitcoin a company holds on its balance sheet. It’s the corporate stash. When that number falls, it can reflect sales, reallocation, or other balance-sheet changes. Either way, it tells you the firm is managing risk and liquidity very differently than a faithful “stack forever” maximist would prefer.
Instead of chasing huge new ASIC orders, MARA bought 2.4 EH of used next-gen miners and acquired Long Ridge Energy & Power’s 505 MW gas plant plus 1,600 acres for an AI/data center campus. That’s not a subtle hint. That’s a neon sign.
MARA CEO Fred Thiel has made the underlying logic fairly clear:
“It is the operational foundation on which we are building …”
“supporting revenue continuity during transition.”
That transition is exactly what’s happening. MARA CFO Salman Khan put it in even blunter terms:
“What are the skills that we are missing that we need to add to get there, and what are the skills that we do not need?”
That’s not the language of a company planning to worship block rewards forever. That’s the language of a firm trying to figure out which parts of the mining workforce, mining stack, and mining identity still make sense once the AI money starts looking fatter than the hash reward.
CleanSpark’s numbers are ugly, and the market knows it
CleanSpark’s latest figures are a reminder that Bitcoin mining profitability can turn nasty fast when costs outrun revenue. The company’s Q2 net loss was $378.3 million, and it booked a $263 million write-down on its BTC treasury of 13,561 BTC.
The more painful part is the operating math. CleanSpark’s average all-in cost to mine one BTC was $103,440, while average revenue per coin was only $75,827. That means mining cost represented 136.4% of BTC revenue per coin. That’s not just a squeeze; that’s a business model getting kneecapped by reality.
Still, CleanSpark isn’t pretending mining is dead. It produced 640 BTC in April, down slightly from 658 in March, and CEO Matt Schultz says the company is making progress on securing its first AI customer. CFO Gary Vecchiarelli framed the mining side like this:
“mining is really our functional currency going forward.”
That’s a useful line because it captures the new role of mining for several public companies. Bitcoin mining is increasingly being treated as a cash engine, not the final destination. It can fund expansion, but it may not be the business that survives in its original form.
IREN is moving hardest toward AI infrastructure
If there’s a company acting like it has already decided where the future is headed, it’s IREN. The firm announced an acquisition of Mirantis in an all-share deal worth $625 million, along with a $2.6 billion debt offering for AI expansion. Its AI Cloud Services revenue doubled to $33.6 million, while mining revenue came in at $111.2 million.
IREN also wrote off $140.4 million in mining gear and had 5,800 Bitmain S21 Pro ASICs up for sale. At the same time, it secured a new AI cloud contract with Nvidia worth $3.4 billion. That’s not hedging. That’s a full-bodied pivot with a side of “please notice we’re not just a miner anymore.”
IREN co-founder and co-CEO Daniel Roberts said the Mirantis deal:
“builds on our existing capabilities and strengthens how that compute is deployed, managed and operated for customers.”
That’s the key phrase here: deployed, managed, operated for customers. The business is no longer just about owning gear and finding cheap power. It’s about turning electricity, cooling, and infrastructure into contracted revenue. AI and HPC are attractive because they can bring more predictable cash flow, longer-term deals, and less dependence on the brutal boom-bust cycle of BTC price and network difficulty.
TeraWulf is barely hiding the exit ramp
TeraWulf may be the most honest about where this is going. Its digital asset revenue fell to just $13 million, while HPC lease revenue reached $21 million and overtook mining revenue. The company booked a $427.6 million net loss in Q1 and has said it may be out of mining “certainly by the next halving” in 2028.
CEO Paul Prager put it bluntly:
“Mining served its purpose.”
CFO Patrick Fleury added:
“Beyond this year … I think we will likely be … out of that business, certainly by the next halving.”
That’s a remarkable admission. The next Bitcoin halving, when block rewards are cut again, has always been a pressure test for miners. If a public mining company is already talking about being out of the business by then, it tells you the economics are not just weak — they’re forcing a strategic rewrite.
Why the AI/HPC pivot keeps spreading
The attraction of AI and HPC is pretty straightforward. These businesses can monetize power, land, and grid interconnection with less dependence on BTC’s dollar price. They can sign contracts. They can build recurring revenue. They can often sell the same underlying infrastructure at better margins than self-mining, especially when difficulty spikes and BTC price softens at the same time.
That doesn’t mean AI is magic or risk-free. It isn’t. AI demand can be frothy, capex can get bloated, and not every mining facility is automatically suitable for enterprise-grade compute. But compared with the chaos of Bitcoin mining economics, AI and HPC can look like a civilized alternative. Less roulette wheel, more utility bill with a contract attached.
There’s also a broader capital-markets angle. Public miners are under pressure from investors to show revenue diversification and reduce reliance on a single volatile asset. That’s especially true when write-downs, impairments, and treasury swings make reported losses look catastrophic, even when some of that pain is non-cash accounting noise. In other words, the numbers can be ugly for real operational reasons, but they can also look even uglier because balance-sheet marks and depreciation slap on additional bruises.
What this means for Bitcoin
This shift does not mean Bitcoin itself is in trouble. The network still works. Blocks are still being mined. The incentive system still routes real-world energy into monetary security, which remains one of Bitcoin’s most powerful design features.
But the mining sector is changing shape fast. Some public companies are becoming infrastructure businesses first and Bitcoin miners second. That has a few implications:
1. Mining may become more concentrated in fewer dedicated operators.
If marginal miners get squeezed out, hashrate can consolidate around firms with the cheapest power, the best sites, or the strongest balance sheets.
2. The network may depend more on miners who truly want Bitcoin exposure.
That’s not necessarily bad. If anything, it can strengthen the ideological alignment of the remaining miners. The tourists leave; the stubborn ones stay.
3. The public mining stock trade is becoming something else.
These companies are no longer just proxies for Bitcoin price. They’re increasingly bets on grid access, power markets, AI demand, and data center execution.
4. The next halving could be brutal for weak operators.
If margins are already thin today, a future reward cut could accelerate exits, mergers, and more desperate pivots into non-mining revenue.
That said, Bitcoin maximalists shouldn’t panic about miners wanting to diversify. Mining has always been a cutthroat business, and the market is simply exposing which firms were really in love with Bitcoin and which ones were in love with cheap megawatts. Those are not always the same people.
The old “mine forever” narrative is getting tested by basic economics. And right now, economics is winning.
Key questions and takeaways
Why are Bitcoin mining economics weakening again?
BTC’s dollar price slipped while network difficulty is set to rise, which makes it harder and more expensive to earn the same block rewards.
What does Bitcoin mining difficulty mean?
It’s the network’s automatic adjustment that keeps block production steady when more computing power joins the race.
Why are miners pivoting to AI and HPC?
Because AI cloud services and data center contracts can offer more stable, often more profitable revenue than self-mining when margins are under pressure.
Which miners are moving fastest?
IREN looks the most aggressive, while MARA, CleanSpark, TeraWulf, Bitdeer, and Cango are all adjusting their exposure in different ways.
Is Bitcoin mining disappearing?
No, but for some public companies it is becoming a bridge business rather than the main destination.
Does this hurt Bitcoin?
Not automatically. Bitcoin still benefits from proof-of-work security, but the industry may become more concentrated among operators who are truly committed to mining.
What happens at the next halving?
The block reward gets cut again, which usually puts more pressure on miner profitability and can force weaker players to exit or reinvent themselves.
Bitcoin mining still matters. It secures the network, monetizes stranded energy, and remains one of the purest examples of market-driven infrastructure. But the easy money era is over, and a lot of public miners are responding by becoming something broader: power, land, and compute companies that happen to mine Bitcoin while it still makes sense.
That’s not the collapse of mining. It’s the market telling everyone exactly what their megawatts are worth. And right now, AI is bidding higher.
