Loading today’s stories
Invest1 distinct publisher2 min readPublished
Peter Schiff says AI bids against Bitcoin for capital, power and space. The contract arithmetic suggests that at the level of the individual site, the argument is already over.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
A miner in the fourth quarter of 2025 spent about $79,995 to produce a coin that was worth $68,000 to $70,000, on CoinShares figures cited by Cryptopolitan [8][9]. At that spread, an AI tenancy is the only use of the building that clears. Hyperscale Data's week is the cash-flow version of the same problem: roughly 150.5 BTC moved for about $9.6 million [13], which is near $63,800 a coin, some 20 percent under what the industry paid to mine one [3]. Coin gets sold below cost to fund the thing that replaces the coin.
What the AI tenant is actually buying is duration. The 12-year, $10.2 billion contract sitting inside Core Scientific's total [7] works out at about $850 million a year of contracted revenue [4], booked a decade ahead against a commodity that was underwater when the ink dried [8]. The concentration is the part worth pricing: about 61 percent of the more than $70 billion CoinShares counts across the industry sits in three contracts [2].
One figure in circulation does not survive a check. The same comparison is used to say operators lost around $19,000 a coin [10]. Take a $68,000 to $70,000 price off a $79,995 cost and the gap is $10,000 to $12,000 [5]. Anyone sizing miner distress off the larger number is working with roughly double what the inputs support, which matters if you are deciding whether these companies signed from strength or from the edge.
Schiff's electricity leg is the one that holds, and not because of price. The International Energy Agency expects data-center consumption to move from 485 terawatt-hours to 950 by 2030 [19], an increase of 465 terawatt-hours, close to a doubling [6]; Galaxy cites Goldman Sachs forecasts of 45 gigawatts of US data-center demand by the same year [17]. The work itself is heavier per unit: about 2.9 watt-hours for a ChatGPT query against 0.3 for a Google search [18], near ten times [7]. US electricity demand is expected to rise roughly 50 percent this decade on data centers alone [20]. Electrons allocated that way do not come back to hashrate because the coin recovers.
Schiff's phrasing gives the promoters too much credit for coherence. He says they are "trying to hitch Bitcoin to the AI wagon" [2]. The wagon is being loaded with the miners' land and their transformers, and the fare is being paid in their coins.
Ranked by verification strength, evidence, and original report placement.
IREN signed a $9.7 billion deal with Microsoft.
TeraWulf signed a $19 billion agreement with Anthropic.
More than 15,000 BTC were sold by public miners after the peak of their treasury holdings.
Cipher Digital and Hyperscale Data are among the companies selling Bitcoin to fund AI data centers.
Peter Schiff argued in a post on X that AI is a competitor to Bitcoin rather than an aid, competing for three scarce resources: capital, electricity and data-center capacity.
Schiff wrote: "Bitcoin pumpers are trying to hitch Bitcoin to the AI wagon, hoping investors will see it as part of the AI trade."
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single-publisher aggregation with unreconciled figures
Every claim rests on one crypto-trade article that summarises third-party research (CoinShares, Galaxy, Coinbase Institutional, IEA, Goldman Sachs) and its own prior reporting without linking primary filings, releases or datasets. Two material numbers are internally inconsistent: Core Scientific's contract appears as both >$14B and 12-year/$10.2B, and the stated ~$19,000 loss per BTC contradicts the article's own cost and price inputs. The article itself warns contract values are company-reported and non-comparable, yet uses them additively. Company-level disclosures (Cipher's loss, Hyperscale's conversion, TeraWulf's revenue crossover) are specific and checkable, which keeps this above the floor.
Contracted capacity and treasury sales already in motion
Adoption of the miner-to-AI-landlord model is documented at multiple companies rather than asserted: named multi-year leases with Microsoft, Anthropic and CoreWeave-type counterparties, a CoinShares tally of more than $70 billion in announced AI/HPC deals, TeraWulf's HPC revenue overtaking mining income in Q1 2026, more than 15,000 BTC liquidated by public miners, and Hyperscale Data's week-long conversion. The score is held below the top band because all of it is reported by one publisher, contract values are company-reported potential revenue, and no independent evidence of energised megawatts actually delivered under these leases is supplied.
Direction sound, individual numbers inflated
The underlying shift — miners monetising energised land for AI while mining margins are negative — is supported by the source's own company-level disclosures, so this is not a manufactured story. Overstatement sits in the numbers and the framing: a per-coin loss inflated to roughly $19,000 from an implied $10,000-$12,000, a Core Scientific value quoted at up to $14B in one place and $10.2B in another, an additive treatment of non-comparable company-reported contract values, and an unattributed 'up to 70% of income from AI' projection. Modestly positive rather than large.
Interested principals throughout the citation chain
The source text itself flags the main conflicts. Schiff is described as a long-standing gold advocate arguing a bearish Bitcoin case; the supporting research comes from market participants with exposure to the narrative (CoinShares, Galaxy, Coinbase Institutional); the contract values are explicitly company-reported potential revenue disclosed by miners that benefit from being repriced as AI infrastructure; and the publisher cites its own earlier coverage and closes with a newsletter subscription pitch. No compensation, sponsorship or holdings disclosures are provided, so the intensity of these incentives cannot be quantified further.
Low - one publisher, unreconciled arithmetic
Confidence is limited by cluster structure as much as content: a single publisher, no primary documents, and two demonstrable internal inconsistencies. The structural direction (negative mining margins, treasury sales, capacity leased to AI counterparties) is corroborated by several independent company-level disclosures within the piece, which supports a moderate rather than minimal score, but individual dollar figures should not be relied on without a second source.
invest
IREN hands Microsoft its first 50MW, and the $9.7B contract starts having to be true1 distinct publisher
invest
The AI moat is now a balance sheet, so price the financing and not the model1 distinct publisher
product
Nebius funds $4.5bn of AI capacity on terms that pay lenders mostly in stock2 distinct publishers
product
Washington's secret AI test is coming for open weights, and release dates go with it2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 24, 2026