Core Scientific, Inc./tx (CORZ) stock intrinsic value fair value estimate & owner earnings analysis

What the business is. Core Scientific spent its public life as a large bitcoin miner and is now turning itself into a landlord for AI computing: it builds and operates power-dense data center space and leases it, under long contracts, to customers who install their own chips. In 2025, 72% of revenue still came from mining bitcoin for its own account and 20% from leasing data center capacity; by the second quarter of 2026 leasing was $136.7 million of $164.2 million in revenue, while the remaining mining operation lost money (a negative 56% gross margin) and is being shut down so its power can be converted. All revenue is earned in the United States. Why past results do not set the value. On our cash-earnings measure (cash from operations, excluding working-capital swings, minus stock compensation and the capital spending needed to stay in business, with mined bitcoin counted when earned) the company lost money in every year from 2022 to 2025. Those years describe a bitcoin miner, including a bankruptcy reorganization, and not the leasing business that exists today. In 2025 we counted as maintenance only an amount equal to depreciation ($68.9 million) of the $729 million spent on property, because most of that spending built capacity already leased to CoreWeave. Large accounting losses in 2024 (a $1.45 billion warrant charge) and the first half of 2026 (a $1.08 billion warrant charge, which the company attributes to its rising share price) come mostly from revaluing warrants; they are not cash costs, and we deal with the warrants below through the share count. So instead of averaging past years, we valued the contracts the company has actually signed. What is signed. (1) CoreWeave: about 590 MW of capacity on 12-year terms, with two five-year renewal options held by the customer. At June 30, 2026 the fixed payments still owed under these leases totaled about $9.2 billion ($6.1 billion on leases already running and $3.1 billion on leases not yet started). About 437 MW was billing by mid-July 2026, and management says the remaining roughly 150 MW, at the Dalton, Georgia campus, begins delivering late in 2026 and is due to be complete in early 2027. CoreWeave is currently the only leasing customer. CoreWeave had agreed to acquire Core Scientific, but Core Scientific stockholders voted the deal down on October 30, 2025. (2) AMD: 15-year agreements announced July 28, 2026 for about 530 MW at Pecos and Hunt County, Texas; Dalton, Georgia; Muskogee, Oklahoma; and Auburn, Alabama, described as more than $14 billion of potential base contracted revenue. Pecos and Auburn are due to start in the first half of 2027 and all 530 MW by the end of 2028. Unlike the CoreWeave build, where the customer funded part of the cost, management expects Core Scientific to fund this one itself. It estimates $11-12 million per MW, about $6 billion, to be funded with project-level bonds that have not yet been issued. How we valued it. Contracted rent was treated as secured growth and nothing else was. From that rent we subtracted site operating costs (20% of rent in the base case, our estimate of second-quarter 2026 costs after removing pass-through power using 2025's split of fixed and variable lease revenue), about $120-140 million a year of corporate overhead, $72-80 million a year of stock compensation, upkeep spending of 3% of rent, the remaining CoreWeave build (about $300 million) and the full AMD build. We added back interest, because debt is subtracted separately, and discounted everything at 6%. We gave no value to capacity that is not under contract: that includes the 440 MW of grid power bought with Polaris DS for $444 million in cash in August 2026, AMD's options to expand at Hunt County and Muskogee, and the rest of the mining sites awaiting conversion. That is the largest reason our figures may look low. Our cash-earnings measure fell further in 2025 (from a loss of $51 million to a loss of $184 million), yet we assume a sharp rise from 2027. That rise comes only from leases already signed and starting on stated dates, not from any recovery in the old business. Balance sheet. There are $3.3 billion of 7.75% senior secured notes due 2031, secured by the Austin, Denton, Dalton, Marble and Muskogee campuses. We add about $0.23 billion for their above-6% cost. There are also $625 million of 0% convertible notes due 2031 (conversion price about $22.49; holders can demand repayment on December 15, 2027), treated as debt. The $460 million of 3% convertible notes due 2029 (conversion price $11.00) were worth $719 million at year-end 2025, far above face value, so we treat them as converting into about 41.8 million shares. The roughly 95 million remaining $6.81 warrants expire January 23, 2027 and were carried as a $1.81 billion current liability at June 30, 2026, so we treat them as exercised, adding about $0.65 billion in cash. That gives about 462 million diluted shares. For cash we count $1.77 billion unrestricted plus $0.78 billion restricted (a debt-service reserve for the notes plus escrow deposits), less the $444 million Polaris payment, plus the warrant cash: about $2.75 billion in total. On top of that we charge about $0.45 billion for financing the AMD build at a rate like the 7.75% notes rather than 6%. AMD also receives warrants for up to 30 million shares at a market-level strike, subject to conditions. We could not size that dilution and left it out, so it is a further risk to all three cases. Base case ($5.01 per share). CoreWeave pays as contracted through the late 2030s. AMD capacity arrives on schedule at about $6 billion and pays about $13.5 billion over 15 years, slightly below the stated figure. When each contract ends, the space is re-leased for five years at half the prior rent, because powered, connected data center shells remain useful, but nothing beyond that is credited. The business is worth about $4.1 billion. Against it sit about $4.6 billion of debt and financing costs and about $2.75 billion of cash, leaving about $2.3 billion for shareholders. Bear case (-$1.95 per share). The AMD build costs about $6.9 billion (about $13 million per MW), opens a year late, is financed at about 9%, and site costs run at 25% of rent. No lease is renewed, and nothing is earned after the contracts end. The business is worth about $1.2 billion against about $4.9 billion of debt and financing costs and $2.75 billion of cash. The company would then owe roughly $0.9 billion more than it is worth. CoreWeave supplying all current leasing revenue is a further risk this case does not fully capture. Bull case ($11.23 per share). Both builds are delivered on budget (AMD at about $11 million per MW). CoreWeave exercises both five-year renewals at the same rent, AMD renews once for five years, and site costs are 18% of rent. The business is worth about $7.0 billion against about $4.6 billion of debt and financing costs and $2.75 billion of cash, leaving about $5.2 billion. What would make this wrong. Upside: contracts for the uncommitted power, such as the Muskogee expansion or AMD's expansion options, would add value we have excluded. Downside: a failure or renegotiation by CoreWeave, AMD build costs above estimates, or project bonds priced well above 7.75% would cut value quickly, because debt is large relative to the equity. The margin-of-safety price is set far below the base case because the bear case leaves nothing for shareholders. It also depends on the AMD financing being secured on terms no worse than the existing notes.