Public Bitcoin miner dumps Bitcoin for AI, historic mistake
Definitely, that is the period of AI. Corporations are reducing their workforces in half to make use of money stream to put money into {hardware}, whereas the inventory market stays close to all-time highs, thanks largely to FAANG. OpenClaw, a self-hosted AI agent, has extra stars on GitHub than Linux or React, however even Jack Dorsey is taking powerful steps to rebuild Block within the face of digital synthetic intelligence. However how a lot of this AI wave is hype, and the way most of the corporations constructing its infrastructure will really profit?
Public Bitcoin miners within the US are making their very own selections, with some investing cash in constructing AI knowledge facilities or utterly rebranding themselves to distance themselves from Orangecoin. Whereas there are numerous AI-related focuses and statements made by public Bitcoin miners on this challenge, a number of stand out as probably the most excessive.
Cypher Mining, estimated to be value about $6 billion and one of many largest within the nation, has introduced a whole rebrand away from Bitcoin and in direction of AI hype. In its newest funding report, titled “Rebranding to Cipher Digital to Replicate Strategic Shift to HPC,” the corporate defined why it “offered its 49% stake in Alborz, Bear, and Chief Mining Websites.” Bitfarms Ltd, one other giant public miner valued at over $1 billion, has additionally pivoted utterly to AI. Coindesk reported that CEO Ben Gagnon went so far as to say, “We’re now not a Bitcoin firm,” however the firm stored the “Bit” in its identify.
A few of these publicly traded corporations anticipate the returns from digital intelligence to be larger than these from Bitcoin, a minimum of within the brief to medium time period, whereas others see it as a chance to diversify or be too huge to cross up.
“The typical price to mine Bitcoin proper now’s about $87,000. The spot worth of Bitcoin is about $70,000. So a lot of the trade is underwater, and public miners are utilizing that as an excuse to pivot,” Kent Halliburton — co-founder and CEO of SazMining, defined in an unique interview with Bitcoin Journal. Sazmining is a personal Bitcoin miner that makes a speciality of frontier power sources and operates primarily outdoors the US.
Halliburton additionally stated, “$87,000 is an trade common and contains these working older technology rigs on the Texas energy grid. At our areas in Paraguay and Ethiopia, our clients obtain 100% “We’re producing Bitcoin for $50,000 to $64,000 based mostly on power prices, which is a spot low cost of 10% to 30%. Profitability is true there.” That might require an extended funding horizon or cheaper power, neither of which appear viable for U.S. public miners concentrating on quarterly stories in greenback phrases.
However as regards to cheaper power, Halliburton means that public miners in the US had the chance to be aggressive, however have been unable to capitalize on that useful resource. He did not say something on the subject, saying that these publicly traded corporations “had all the things they wanted to mine Bitcoin cheaply – energy contracts, land, infrastructure – however they’re handing it over to Microsoft and Google in trade for lease checks. They’re doing all the things from securing the Bitcoin community to securing rack house for hyperscalers, and so they name it a technique. In the meantime, they’re releasing over 15,000 Bitcoin from their steadiness sheets to fund the transition.”
Among the many largest public Bitcoin miners, IREN Restricted started pivoting to AI cloud providers in April 2025, saying a $9.7 billion, five-year settlement with Microsoft for 200 MW of crucial IT workloads utilizing NVIDIA GB300 GPUs. TeraWulf has accomplished a number of Google-backed HPC expansions by way of Fluidstack, securing a 10-year contract for over 200 MW.
Cipher Digital has accomplished a whole rebrand to HPC landlord with 600 MW of contracted capability, together with a 15-year 300 MW lease with AWS and a 10-year 300 MW lease with Google-backed Fluidstack. Hut 8 has signed a 15-year, 245 MW lease with Fluidstack, additionally backed by Google, with potential for future extensions and first provide rights of over 1,000 MW. Core Scientific has expanded its HPC focus to 270 MW by way of a partnership with Microsoft and CoreWeave to service OpenAI workloads.
Riot Platforms is strategically evaluating AI internet hosting enlargement by partnering with AMD for a 10-year 25 MW operational lease and 600 MW AI/HPC analysis on the Corsicana website, however no hyperscaler settlement has been introduced.
MARA Holdings is diversifying into AI by way of a three way partnership with Starwood Capital’s Starwood Digital Ventures, concentrating on 1 GW of IT capability within the close to time period, scalable to greater than 2.5 GW for hyperscale and AI workloads, with Starwood main funding and tenant procurement, however no hyperscaler deal has been finalized but.
CleanSpark pursues AI pivot by buying Texas land and energy for AI/HPC. It contains 447 acres of land in Brazoria County with a possible of 300-600 MW and a website in Austin County contributing 890 MW of complete energy technology, and is being funded by way of Bitcoin gross sales, with tenant negotiations underway however no hyperscaler leases disclosed.
The AI gold rush is right here, there is no such thing as a doubt about it. Many of those public miners undoubtedly appear to see a chance to construct infrastructure for profound know-how tendencies. However in any case, in the long term, historical past has not been form to these constructing the infrastructure of a brand new period. It tends to be a really excessive threat, medium reward sort of wager. For instance, how most of the corporations that constructed railroads are nonetheless round immediately? Or, with out going that far again, are you able to identify the businesses that constructed fiber optic web traces within the late 90s and 2000s?
There’s a lengthy listing of railroad bankruptcies from the late 1800s that even led to a monetary disaster referred to as the Panic of 1873, lots of which took on extreme debt to finance building that was not but in enough demand. After the panic, JPMorgan led the consolidation of bankrupt railroad corporations, resolved debt disputes, and positioned actual property property beneath new possession. They have been those who lastly took benefit of railroad building.
And simply because the century was coming to an finish, the dot-com bubble of the 2000s left behind a graveyard of fiber-optic infrastructure corporations that have been ultimately acquired by hyperscalers like Google and Meta for $1 throughout post-crash consolidation.
Whereas the development of railroads and fiber optic traces collectively helped develop international commerce in unbelievable methods, demonstrating the knowledge of the market as a complete, most particular person corporations concerned didn’t survive the method, and enterprise capitalists targeted on immediately’s AI increase are conscious of this dynamic.
Hole between capital funding and income AI
Varied investor teams are starting to query the place the income for this large infrastructure spending will come from. Goldman Sachs argued in an October 2025 report titled “AI: Is We in a Bubble?” that whereas earlier investments might have been supported by sturdy technology-related returns, valuations for some corporations are beginning to “bubbly.”
Sequoia’s David Chan factors out that the hole between AI income and capital expenditures (Capex) has widened since 2023, with a extensively reported distinction of $600 billion. The hyperscaler’s 2026 capex dedication is over $700, however the place is the return?
Whereas OpenAI’s $20 billion in annual recurring income (ARR) is spectacular for a startup, FuturumGroup stories that it represents “roughly 3% of complete hyperscaler capex in 2026,” and “Anthropic’s $9 billion run fee places it in an identical place, with 9x year-over-year progress. Cohere ($150 million) So is all the cohort of pure-play AI distributors, together with . ARR), Mistral (roughly $400 million), and Perplexity ($148 million yearly), amongst others, will most likely account for lower than $35 billion of complete projected income in 2026. ”
Skepticism about the place the worth of AI can really be captured can be expressed by enterprise capitalists like Chamath Palihapitiya. He’s a distinguished investor in Groq, an organization growing customized silicon for the AI period that was licensed by NVIDIA in a $20 billion deal final yr, and has been a Fb insider all through the corporate’s progress right into a hyperscaler. If he has doubts in regards to the profitability of synthetic intelligence railway building, then maybe there’s something value trying into very intently.
Palihapitiya additionally argued on a latest All In Podcast that corporations will quickly notice that they’re exposing their commerce secrets and techniques to cloud AI and should select to self-host as a substitute. Constructing an in-house GPU farm might look like a little bit of a aspect quest, however are you actually prepared to threat your commerce secrets and techniques by working with an AI supplier that trains on consumer knowledge? In any case, new variations of fashions skilled on that knowledge can have it integrated into their data base and uncovered to the world. And even when company agreements to not use company knowledge for coaching change into the norm, it creates a really excessive degree of belief, making a systemic threat for a given firm: the danger that the information could possibly be leaked or considered by the incorrect insiders inside the cloud AI supplier firm.
For a similar purpose, there may be additionally the query of whether or not the market essentially needs cloud AI. Would you rent a private assistant if you happen to knew the information you shared with it could find yourself on the web? Most likely not, however that is what’s occurring with AI. Actually, the Southern District of New York lately dominated that customers are usually not entitled to client-attorney privilege when receiving authorized help from an AI chatbot, that means that confidential discussions with the AI may be legally subpoenaed and used in opposition to the shopper in courtroom. This illustrates the dangers related to blindly trusting AI. Some speculate that new sorts of phrases and contracts will should be created to help this use case. However this case illustrates the basic components of the demand for AI. In different phrases, individuals need dependable humanoid intelligence, digital or in any other case.
AI loyalty and belief
Ah, “belief,” the ever-present, virtually supernatural phrase that does a lot to hold the load of the world. However what’s belief? Basically, it’s predictability, an individual’s confidence that one other individual, system, or AI will behave in a sure manner in a dependable, predictable, and optimistic method towards his or her profit. Nonetheless, if the AI is hosted within the cloud, no such ensures may be made. Knowledge principally leaves the consumer’s machine and is processed within the “cloud”, and what occurs there may be past our human comprehension. Actually, “the cloud” comes with authorized dangers that may impede your loyalty as a consumer in sure eventualities. So maybe the general public is fascinated by OpenClaw.
In latest weeks, a brand new open supply challenge on the earth of AI has taken the tech trade by storm. With 289,000 stars on GitHub, it has extra stars than Linux has, regardless of supporting software program infrastructure around the globe, and greater than React, one of many world’s hottest net growth languages. And it is solely been stay for what number of weeks? How can this occur? Why do individuals prefer it a lot?
Nicely, there are most likely two causes. It feels extra like a human assistant than a chatbot. It updates robotically, remembers what you have an interest in, writes a diary and develops based mostly in your preferences. However most significantly, you possibly can host it by yourself machine. Folks have been shopping for Mac minis en masse to run OpenClaw, mixed with Claude Max API token plans for about $200 per 30 days. Some argue that it is a revolution in self-hosting, despite the fact that the above setup nonetheless depends on the cloud. However what’s actually occurring right here is that OpenClaw seems to be loyal, remembers you, and is “inside” your PC. This isn’t a chat interface the place the context window ultimately grows unmanageably giant and dies a small dying, changed by a brand new chat tab. OpenClaw will not be a chatbot. It is a sort of AI entity that customers construct relationships with. And good relationships are constructed on belief.
So what does all this must do with public Bitcoin miners? Maybe self-hosted AI is the longer term, and China’s AI mannequin is more and more lean and able to operating on machines which can be removed from innovative, maybe with stress to innovate as a consequence of sanctions on specialised AI {hardware} reminiscent of high-end Nvidia chips. Every kind of open supply instruments for managing and internet hosting fashions regionally are being launched and improved recurrently. If historical past is any information, the mass manufacturing of AI {hardware} will result in the commoditization of highly effective computer systems that may attain finish customers’ houses and course of AI.
Actually, Apple, the worst FAANG of all AI merchandise ever launched, might find yourself being one of many largest winners of the AI race. why? As a result of the consumer {hardware} is best. Trendy Macs don’t distinguish between RAM and VRAM. It is a drawback seen with all different computer systems that depend on GPUs, reminiscent of Nvidia. This limits the dimensions and velocity of fashions that may be self-hosted. As a substitute, trendy Mac machines have all their RAM built-in, permitting customers to run highly effective fashions regionally that can’t simply be run on non-Apple {hardware}. Self-hosted AI is the longer term.
Subsequently, public Bitcoin miners who have been pursuing medium-term fiat earnings might have simply fallen right into a entice. It is the identical entice that the giants of the dot-com bubble fell into. It is the identical entice that the giants of the economic age who constructed the railroads fell into. The infrastructure that powers the longer term does not essentially make a revenue.
The put up Public Bitcoin Miners Are Dumping Bitcoin for AI, a Historic Mistake initially appeared in Bitcoin Journal and is written by Juan Galt.
