CoreWeave, Inc. (CRWV) stock intrinsic value fair value estimate & owner earnings analysis

CoreWeave rents GPU-based AI computing capacity to customers under multi-year committed contracts, plus some pay-as-you-go use. Revenue grew from $229M in 2023 to $1.9B in 2024 and $5.1B in 2025, and contracted backlog reached $60.7B at December 31, 2025 (up from $15.1B a year earlier), with a weighted-average contract length of about five years. 43% of that backlog is scheduled for recognition through 2027, 38% in the following two years, and the remainder between 2030 and 2032. The key question is what the shares earn after the cost of keeping the GPU fleet running. We measure earnings as operating cash flow, less the swing in working capital (customers prepay large amounts, which is timing rather than profit: $4.2B of prepayments in 2025 alone), less stock compensation ($630M in 2025), less the cost of replacing the fleet as it wears out. Because capex ($10.3B in 2025) is many times revenue growth, we treated fleet replacement as equal to depreciation ($2.5B in 2025); the usual growth/maintenance split would have wrongly counted zero replacement cost. On that basis earnings per share were $0.02 in 2023, $0.41 in 2024 and $(2.11) in 2025. Operating income was $(46)M in 2025 and $(193)M in the first half of 2026 despite revenue more than doubling to $4.7B, so the growth has not yet turned into profit for shareholders. Debt rose to $21.4B at year-end 2025 (from $7.9B), and interest paid was $869M in 2025 and $806M in the first half of 2026 alone. What each case assumes. All three cases credit only the contracted backlog, delivered on its disclosed schedule (roughly $13B a year in 2026 and 2027, $11.5B a year in 2028 and 2029, and about $3.8B a year in 2030 to 2032), and nothing beyond it, because no renewal, new customer or new contract is yet signed. Backlog growth beyond what is signed was left out, as were any margin improvements not yet visible in the results. - Base (-$17 per share): backlog converts to shareholder earnings at the recent margin of about -18% of revenue (2025 was -18%; the first half of 2026 was about -18%). Discounted at 6%, that is about $(9.2)B, or roughly $(17) per diluted share on about 551M shares. This says that on the evidence so far the company owes more to lenders and to the cost of replacing its fleet than the contracts earn, by roughly that amount. - Bear (-$28 per share): margin worsens to -30% of revenue, since capital spending is growing faster than revenue (first-half 2026 capex was $14.1B against $10.3B for all of 2025) and interest costs rose 240% in 2025. Two-thirds of 2025 revenue came from one customer, so a problem at that customer would hit hard. - Bull ($4 per share): the contracts earn the 2024 margin of about 4.6% of revenue, the best the company has shown after replacement cost. Even this modest outcome is worth only about $2.3B. Growth-path check: owner earnings fell from $89M to $(920)M in the last fiscal year; the base case assumes no improvement, consistent with the trailing rate, since the first half of 2026 shows no improvement. What would make this wrong: if newer contracts carry materially better pricing than the existing fleet, or if borrowing costs fall, earnings after replacement costs could turn solidly positive. We have not seen that in any reported period. Revenue is concentrated: one customer was 67% of 2025 revenue and 68% of receivables. The reported figures were reconciled to the cash flow statement; about $247M of 2025 operating cash flow is other non-cash items not itemised here. On current evidence these figures support only a low margin-of-safety price, and it applies only once the business shows positive earnings after replacement costs.