There isn’t a doubt about it, that is the age of AI. Firms are slicing their workforces in half to take a position money circulate into {hardware}, whereas the inventory market stays close to all-time highs, principally due to FAANG. OpenClaw, a self-hosted AI agent, has extra stars on GitHub than Linux and React, whereas even Jack Dorsey is taking harsh measures to restructure Block within the face of digital, synthetic intelligence. However how a lot of this AI wave is hype, and the way most of the firms that construct its infrastructure will truly seize the income?
Public Bitcoin miners in the US have made their selection, a wide range of them committing capital to constructing out AI datacenters, and a few even making full rebrands, distancing themselves from the orange coin. Whereas there’s a full vary of AI-related pivots and statements made by public Bitcoin miners on the matter, a pair stand out as essentially the most radical.
Cypher Mining, estimated to be price round six billion {dollars} — putting it among the many largest within the nation – introduced a full rebrand away from Bitcoin and on to the AI hype prepare. Of their most up-to-date funding report titled “Rebrands to Cipher Digital to Mirror Strategic Shift Towards HPC,” the corporate defined why they “Divested 49% Stake in Alborz, Bear, and Chief Mining Websites”. Bitfarms Ltd, one other giant public miner valued at over a billion {dollars}, additionally made a full pivot to AI. The CEO, Ben Gagnon, went so far as saying “We’re not a Bitcoin firm,” as reported by Coindesk, although they did hold the ‘Bit’ within the title.
A few of these public firms predict extra greenback returns from digital intelligence than these they get from Bitcoin, a minimum of within the brief to mid time period, whereas different are others may think about it a diversification or a possibility too giant to overlook.
Kent Halliburton — Co-Founder & CEO of Sazminingexplained to Bitcoin Journal in an unique interview that “The common price to mine a bitcoin proper now could be about $87,000. The spot value of bitcoin is about $70,000. So a lot of the trade is underwater, and the general public miners are utilizing that as their excuse to pivot.” Sazmining is a personal Bitcoin miner that focuses on frontier power sources, with operations principally exterior of the US.
Halliburtonalso famous that “$87,000 is an trade common — it contains guys working old-gen rigs on grid energy in Texas. At our websites in Paraguay and Ethiopia, our purchasers are producing bitcoin on an power price foundation of $50,000 to $64,000, on 100% renewable power. That’s a ten to 30 % low cost to identify. The profitability is correct there.” It simply requires an extended funding horizon or cheaper power, neither of which seems to be actionable for American public miners who’ve dollar-denominated quarterly experiences to focus on.
On the subject of cheaper power, nevertheless, Halliburton means that public U.S. miners had the prospect to be aggressive however have didn’t reap the benefits of their assets. He minced no phrases on the subject, saying that these public firms “had the facility contracts, the land, the infrastructure — all the pieces you have to mine bitcoin cheaply — and so they’re handing it to Microsoft and Google in trade for lease checks. They went from securing the Bitcoin community to securing rack area for hyperscalers, and so they’re calling it a method. In the meantime, they’ve dumped over 15,000 bitcoin off their stability sheets to fund the transition”.
Of the most important public Bitcoin miners, IREN Restricted started its pivot to AI cloud providers in April 2025, asserting a$9.7 billion, five-year settlement with Microsoft for 200 MW of essential IT load utilizing NVIDIA GB300 GPUs. TeraWulf has executed a number of Google-backed HPC expansions by means of Fluidstack, securing 10 yr agreements for over 200 MW.
Cipher Digital accomplished its full rebrand to an 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 Fluidstack backed by Google. Hut 8 signed a 15-year, 245 MW lease with Fluidstack, additionally backed by Google, eyeing future potential extensions and a proper of first supply for over 1,000 MW. Core Scientific has expanded its HPC focus to 270 MW by means of partnerships with CoreWeave, which serves Microsoft and OpenAI workloads.
Riot Platforms is strategically evaluating an AI internet hosting growth by partnering with AMD on an operational 10-year, 25 MW lease and assessments for 600 MW of AI/HPC at its Corsicana web site, although no hyperscaler agreements have been introduced.
MARA Holdings is diversifying into AI by means of a three way partnership with Starwood Capital’s Starwood Digital Ventures, focusing on 1 GW of near-term IT capability expandable to over 2.5 GW for hyperscale and AI workloads, with Starwood main financing and tenant sourcing, however with out named hyperscaler contracts but.
CleanSpark is pursuing a pivot to AI by buying Texas land and energy for AI/HPC, together with 447 acres in Brazoria County for 300–600 MW potential and an Austin County web site contributing to 890 MW mixture, funded by Bitcoin gross sales, with tenant discussions ongoing however no hyperscaler leases disclosed.
So the AI gold rush is right here, there’s little doubt about it, many of those public miners apparently see a possibility to construct out the infrastructure of — what’s certainly— a profound technological development. However historical past has not been sort to those that construct the infrastructure of a brand new period, not in the long run anyway. It tends to be a really high-risk, medium-reward type of guess. How most of the firms that constructed the railroads — for instance — are nonetheless round right this moment? Or, with out going again that far, are you able to title any firm that constructed out web fiber strains within the late 90’s and 2000’s?
There’s a lengthy record of railroad bankruptcies from the late 1800’s, which even led to a monetary disaster in what’s known as the Panic of 1873, many overleveraged in debt to fund build-outs for which there was not sufficient demand but. After the panic, J.P. Morgan led a consolidation of bankrupt railroad firms, resolving debt disputes and bringing their actual property belongings underneath new possession. It was they who ended up capturing the upside of the railway build-out.
And simply across the nook of the century, the dot com bubble of the 2000’s left a graveyard of fiber line infrastructure firms that have been additionally, ultimately, purchased out by hyper scalers like Google and Meta throughout the publish crash consolidation, for pennies on the greenback.
Whereas each the railway and fiber line build-outs total helped scale commerce for the world in unbelievable methods — demonstrating the general knowledge of the markets — most particular person firms concerned didn’t survive the method, and enterprise capitalists wanting on the AI increase right this moment are conscious of this dynamic.
The Capex vs Income AI Hole
Numerous investor teams are beginning to query the place the returns on this huge infrastructure spending will come from. In an October 2025 report titled “AI: In a bubble?”, GoldmanSachs took a argued that, whereas the investments to date might be supported by large tech income, the valuations of a few of these firms have been beginning to get “frothy”.
David Chan at Sequoia has been declaring the rising hole between AI-driven income and capital expenditures (Capex) since 2023, resulting in a broadly reported variety of a $600 billion hole between them. Capex spending commitments in 2026 are north of $700 for the hyper scalers, however the place are the returns?
OpenAI’s $20 billion annual recurring income (ARR) is spectacular for a brand new firm, however that represents “roughly 3% of the projected 2026 hyperscaler capex whole” as reported by FuturumGroup, who famous that “Anthropic’s $9 billion run price, whereas exhibiting 9x year-over-year development, occupies an analogous place. Your complete cohort of pure-play AI distributors – together with Cohere ($150 million ARR), Mistral (~$400 million), Perplexity ($148 million annualized), and others – doubtless accounts for lower than $35 billion in projected mixed 2026 income.”
Skepticism about the place the worth of AI will truly be captured has additionally been aired by VC’s like Chamath Palihapitiya. He was a outstanding investor in Groq, an organization constructing customized silicon for the AI age, which was licensed by NVIDIA in a $20 billion deal final yr, and was a Fb insider by means of the corporate’s rise to turn into a hyperscaler. If he has his doubts concerning the profitability of constructing the railroads of synthetic intelligence, then maybe there’s one thing price giving a really shut have a look at.
Palihapitiya additionally argued in a latest All In Podcast that firms may quickly begin to notice they’re exposing their commerce secrets and techniques to cloud AI, preferring as an alternative to self-host. Constructing out in-house GPU farms may seem to be a little bit of a aspect quest, however can you actually danger your commerce secrets and techniques with AI suppliers who prepare on person information? In spite of everything, new variations of fashions educated on that information could have it of their data base, uncovered to the world. And even when company agreements to not prepare on company information turn into the norm, a really excessive belief relationship could be shaped, posing a systemic danger to sure firms, a danger that the information may get leaked or seen by the flawed insiders contained in the cloud AI supplier firms.
There are additionally questions on whether or not the market essentially desires cloud AI for a similar causes. Would you rent a private assistant when you knew the information you share with them would find yourself on the web? Most likely not, however that’s what’s taking place with AI. The truth is, the U.S. Southern District of New York lately dominated that customers should not have client-attorney privilege when getting authorized assist from AI chatbots, and thus, delicate discussions with AI might be legally subpoenaed and used in opposition to the purchasers in a courtroom of regulation, an indication of the dangers concerned with trusting AI blindly. Some speculate that new sorts of phrases and agreements will must be shaped to help this use case. However the authorized case factors to a elementary aspect of the demand for AI: individuals need humanoid intelligence, digital or in any other case, that they’ll belief.
AI Loyalty and Belief
Ah, “Belief”, that ubiquitous, virtually supernatural phrase that does a lot work to hold the load of the world. However what’s belief? Basically, it’s predictability, one particular person’s confidence that one other human, system, or AI will behave in a sure manner, in a dependable, predictable, and constructive manner in the direction of one’s pursuits. AI, when hosted within the cloud, nevertheless, cannot give such assurances; the information is essentially leaving the person’s machine to be processed by “the cloud,” and what occurs up there’s past us mortals. The truth is, “the cloud” has authorized dangers which may stop it from being loyal to you as a person in sure eventualities. Therefore, maybe the general public’s fascination with OpenClaw.
In latest weeks, a brand new open supply venture within the AI world has taken the tech trade by storm. 289,000 stars on GitHub, greater than Linux has gotten regardless of supporting the software program infrastructure of the world, greater than React, one of the vital well-liked net growth languages on the earth. And it’s solely been dwell for what, weeks? How might this be? Why do individuals prefer it a lot?
Properly, arguably two causes. It feels extra like a human assistant than a chatbot; it updates itself, remembers what you have an interest in, journals, and develops round your preferences. However most necessary of all, you may host it in your machine. Individuals have been buyingMac minis in droves to run OpenClaw, pairing it up with Claude Max API token plans of about $200 a month. Some argue it is a revolution in self-hosting, although the above setup continues to be depending on the cloud. However what’s truly taking place right here is that OpenClaw seems loyal, it remembers you, it’s “in your house” in your PC. It’s not a chat interface whose context window will ultimately turn into an excessive amount of for it to handle, ending in a small loss of life, changed by a brand new chat tab. OpenClaw just isn’t a chatbot; it’s an AI entity of kinds that customers create a relationship with. And good relationships are constructed on belief.
So what does all of this should do with public Bitcoin Miners? Properly, maybe self-hosted AI is the longer term, Chinese language AI fashions are more and more leaner and may run on machines removed from the innovative, arguably pressured into innovation by sanctions on specialised AI {hardware} like high-end Nvidia chips. Open supply instruments of all types that handle and host fashions domestically are commonly launched and improved, and if historical past is any information, the mass manufacturing of AI {hardware} will result in the commoditization of highly effective computer systems that can make it to finish customers’ houses, and may deal with AI.
The truth is, Apple, the FAANG that has had the worst AI merchandise deployed up to now, could find yourself changing into one of many largest winners of the AI race. Why? As a result of their person {hardware} is great. Current Macs don’t have a distinction between RAM and VRAM, a problem all different computer systems depending on GPUs, corresponding to Nvidia, have. This limits the dimensions and pace of fashions that may be self-hosted. As a substitute, all RAM is unified within the newest Mac machines, letting customers run highly effective fashions domestically that don’t simply run on non-Apple {hardware}. Self-hosted AI is the longer term.
And thus, public Bitcoin miners, within the pursuit of mid-term fiat features, might need simply fallen for a lure. The identical lure the giants of the dot-com bubble fell for. The identical lure that the titans of the commercial period, who constructed the railroads, fell for. The infrastructure that runs the longer term doesn’t essentially seize the features.
