I.Executive Summary
CoreWeave rents out Nvidia GPU computing power, the chips that train and run AI models, from data centres it leases and fills with equipment. Its customers are among the largest AI buyers: Microsoft supplied 67% of 2025 revenue, and OpenAI, Meta, Jane Street and Nvidia itself have signed multi-year contracts. Growth has been extraordinary: revenue rose from $229 million in 2023 to $5.13 billion in 2025, and to $2.58 billion in the second quarter of 2026 alone (+112% year on year). Revenue backlog was about $104 billion at 30 June 2026, with more than $25 billion of new commitments signed early in the third quarter.
The business is financed mostly with debt, and it loses money after interest. Second-quarter adjusted EBITDA was $1.51 billion (59% of revenue), but depreciation of $1.39 billion and net interest of $640 million turned that into a net loss of $626 million. Capital spending is guided at $35–39 billion for 2026. In mid-October 2026 the shares fell 15.5% in a day and more than 25% in a week as yields on CoreWeave's bonds rose to about 15% and its credit-default-swap spread passed 1,000 basis points (CNBC). The close on 30 October 2026 was $54.87, 60% below the 52-week high of $138.25.
The question for CoreWeave is not demand, which its backlog shows is enormous, but whether a highly leveraged company can turn contracted revenue into enough cash to service, refinance and eventually repay its debt before the GPUs it is buying today become obsolete.
CoreWeave has no positive earnings to value (2026 consensus EPS −$5.18), so our illustrative five-year projection uses a revenue multiple, then deducts debt: $12.8B of 2026 revenue growing 18–35% a year, valued at 3.0–4.5× sales in 2031, less $50–70B of net debt, with 3% annual dilution. That gives Year-5 values of $27.92 (bear), $110.74 (base, the midpoint) and $325.77 (bull), against today's $54.87: about +15.1% a year in the base case (−12.6% to +42.8%). Discounted at 10% a year, the base case is worth $68.76 today. The width of that range is the point: with about $30B of net debt, the equity is a leveraged claim on the data centres. Evidence: Weak. *All values are Pathos's illustrative estimates for educational purposes, not price targets or recommendations.
II.Business, Customers & Capacity
What it sells. CoreWeave calls itself "The Essential Cloud for AI". Customers sign contracts, typically for several years, to reserve clusters of Nvidia GPUs, networking and storage, run through CoreWeave's own software (its Kubernetes service, SUNK for Slurm-based training, and the Weights & Biases developer tools it bought in May 2025). It also sells shorter-term and on-demand capacity. All of its GPUs today are Nvidia's: "as a result of our obligations in our current customer contracts, all of the GPUs used in our infrastructure today are NVIDIA GPUs" (Q2 2026 10-Q).
Customers and concentration. "We recognized an aggregate of approximately 67% of our revenue from our top customer, Microsoft, for the year ended December 31, 2025" (2025 10-K). That share is falling as new contracts start: in the second quarter of 2026 the largest customer was 36% of revenue, the second 26% and a third 10% (the 10-Q does not name them). Disclosed commitments include OpenAI ("up to approximately $11.9 billion through October 2030", March 2025, and a further "up to approximately $6.5 billion through May 31, 2031", September 2025), Meta ("up to approximately $14.2 billion through December 2031", September 2025, and "up to approximately $21.0 billion" more in March 2026) and Jane Street ("approximately $6.0 billion", April 2026).
The Nvidia relationship. Nvidia is supplier, customer, shareholder and backstop. It owned 47.2 million Class A shares (10.7% of that class) at 15 April 2026 (proxy), bought $2.0 billion more at $87.20 a share in January 2026, and in September 2025 signed a $6.3 billion order under which "NVIDIA is obligated to purchase the residual unsold capacity through April 13, 2032" (8-K, 15 Sep 2025). That backstop lowers the risk of empty capacity, but it also ties CoreWeave's fortunes, and its financing, to one chip supplier.
Capacity. At the end of 2025 CoreWeave ran 43 data centres with over 850 MW of active power and about 3.1 GW contracted (10-K). By 30 June 2026 active power was 1.5 GW and contracted power about 3.7 GW (Q2 release); management raised its year-end target to more than 1.85 GW, cited 4.2 GW contracted at the time of the August call, and repeated a goal of at least 8 GW by 2030 (Investing.com call transcript). CoreWeave mostly leases powered shells from developers; its attempt to buy one of them, Core Scientific, in an all-stock deal ended on 30 October 2025 when Core Scientific's stockholders voted it down ("the Agreement has been terminated"); the two continue as commercial partners.
Competition. CoreWeave competes with other GPU clouds such as Nebius and with the hyperscalers, including Oracle Cloud Infrastructure. Because CoreWeave and Nebius report losses, the multiple shown is price / sales.
Peer comparison
| Company | TTM to | Market cap | Revenue | Revenue growth | Operating margin | FCF / net income | Net cash (debt) | P/S |
|---|---|---|---|---|---|---|---|---|
| CoreWeave (CRWV) | Jun 26 | $27.7B | $7.59B | +115% | −2.7% | n.m. (FCF −$13.7B) | ($46.0B) | 7.2 |
| Nebius Group (NBIS) | Jun 26 | $38.8B | $1.36B | +507%‡ | −37.4% | n.m. (FCF −$5.9B) | ($2.15B) | 52.2 |
| Oracle (ORCL) | Aug 26 | $361B | $71.8B | +22% | 34.3% | n.m. (FCF −$28.7B) | ($132.1B) | 6.0 |
Source: StockAnalysis.com trailing-twelve-month financials (S&P Global Market Intelligence data drawn from company filings), read 2 Nov 2026. Revenue growth is TTM vs prior TTM; margin is GAAP operating income / revenue; FCF / net income (operating cash flow − capex, over GAAP net income) is a Pathos calculation; P/S is trailing at the quarter-end shown, not today's price. Periods differ because fiscal years differ. n.m. = not meaningful (loss or negative denominator). Verified: figures as reported by the data provider from filings. Estimated / labelled: ‡ Nebius' growth is off a tiny base after its 2024 restructuring. Oracle is a whole company, not its cloud segment. Market cap is spot, from StockAnalysis.com on 2 Nov 2026 (US close; Asian listings at their latest 2–3 Nov local quote), so it is not at the quarter-end shown; revenue is TTM to the period shown. All amounts are in US$. Educational comparison only; it is not a ranking or a recommendation.
III.Share Price & Valuation Multiples
Price as of Nasdaq close, 30 Oct 2026 · Fundamentals: FY2025 10-K and Q2 2026 Form 10-Q · Consensus retrieved 31 Oct 2026| Metric | Value | Basis |
|---|---|---|
| Share price | $54.87 | Close, 30 Oct 2026 (Yahoo Finance chart data); IPO price $40 (28 Mar 2025) |
| 52-week range | $51.24 – $138.25 | Intraday, Yahoo Finance; all-time closing high $183.58 (20 Jun 2025) |
| Market capitalisation | $30.3B | 551.5M Class A and B shares (10-Q cover, 31 Jul 2026) × $54.87 |
| Enterprise value, ex leases | $60.4B | Plus debt principal $35.6B, less cash $5.5B (30 Jun 2026) |
| Enterprise value, incl. leases | $76.9B | Plus operating and finance lease liabilities of $16.5B |
| P/E | n/m | Net loss: TTM −$1.93B; 2026 consensus EPS −$5.18, 2027 −$4.68 (23 analysts) |
| EV / 2026 revenue (ex leases) | 4.7× | 2026 consensus $12.87B (40 analysts); guidance $12.4–13.2B |
| EV / 2027 revenue (ex leases) | 2.3× | 2027 consensus $26.47B (range $21.1–31.5B) |
| EV / adjusted EBITDA (TTM, ex leases) | 13.7× | TTM adjusted EBITDA $4.40B (Pathos calculation from releases) |
| Price / sales (TTM) | 4.0× | TTM revenue $7.59B |
| Price / book | 6.0× | Equity $5.02B at 30 Jun 2026 |
| Dividend | None | No buybacks; the company issues shares and convertibles |
With no earnings, the usual multiples do not apply, and the cheapest-looking figure, 2.3× next year's revenue, assumes revenue doubles again and ignores that the debt holders come first. For context only, the mean analyst price target compiled by Yahoo Finance (38 analysts) is $134.58, with a range from $39 to $278; several firms cut their targets in the week to 21 October 2026.
IV.Financial Record Since 2023
Source: CoreWeave 2025 10-K and earnings releases (adjusted EBITDA margins as reported); 2026 = Yahoo Finance consensus of $12.87B (40 analysts, retrieved 21 Oct 2026), with the first-half 2026 adjusted EBITDA margin. Adjusted EBITDA is a non-GAAP measure that excludes stock compensation, depreciation and interest.
| US$ millions | FY2023 | FY2024 | FY2025 | H1 2026 |
|---|---|---|---|---|
| Revenue | 229 | 1,915 | 5,131 | 4,653 |
| Growth | n/m | 736% | 168% | 112% vs H1 25 |
| Adjusted EBITDA (non-GAAP) | n/a | 1,219 | 3,093 | 2,667 |
| Margin | n/a | 64% | 60% | 57% |
| Depreciation & amortisation | 103 | 863 | 2,454 | 2,540 |
| Operating income (GAAP) | (14) | 324 | (46) | (193) |
| Interest expense, net | (28) | (361) | (1,229) | (1,176) |
| Net income (loss) | (594) | (863) | (1,167) | (1,366) |
| Diluted EPS (US$) | (3.09) | (4.30) | (2.81) | (2.53) |
| Operating cash flow | 1,833 | 2,749 | 3,058 | 3,663 |
| Capital expenditure (cash) | (2,943) | (8,702) | (10,309) | (14,117) |
| Free cash flow | (1,110) | (5,953) | (7,251) | (10,454) |
CoreWeave has existed as a public company for only six quarters, so a five-year record is not available: the 2025 10-K gives three years, and 2024 quarters come from comparatives. Revenue grew 736% in 2024 and 168% in 2025, and 112% in the first half of 2026. Operating income has hovered around zero because depreciation grows as fast as revenue, and interest expense has grown faster: net interest was $1.23 billion in 2025, 24% of revenue. Third-quarter guidance (11 Aug 2026) is revenue of $3.45–3.60 billion, adjusted operating income of $200–260 million and interest expense of $860–940 million; management expects adjusted operating margin to reach "low teens" in the fourth quarter. Third-quarter results are expected around 9 November 2026 (StockAnalysis estimate).
Bars: quarterly revenue and net income, US$ billions. CoreWeave filed its first 10-Q after its 28 March 2025 IPO; 2024 quarters come from comparatives in later filings, so the chart covers ten quarters (Q1 2024–Q2 2026). Line: trailing P/E = closing share price at quarter-end ÷ trailing-12-month diluted EPS (sum of the last four quarters; fiscal Q4 derived as full-year minus nine-month figures). Gaps in the line: no P/E where trailing EPS was negative (Q1 FY25–Q2 FY26); no P/E before the shares began trading (Q1 FY24–Q4 FY24). Both axes share zero; the P/E scale extends below 0× only to make room for net-loss bars (P/E is never plotted below zero). Sources: Form 10-Q/10-K data via SEC XBRL company facts; Yahoo Finance price history. Hover over a quarter for exact values.
Total tangible book value, US$ billions = shareholders’ equity minus goodwill and other intangible assets, at quarter-end. CoreWeave’s equity was negative before its March 2025 IPO, when its redeemable convertible preferred stock sat outside equity; the IPO converted the preferred into common stock. Goodwill and intangibles come mainly from 2025 acquisitions (Weights & Biases). Share counts are reported on the cover page only, so per-share values are not shown. Source: balance sheets in Forms 10-Q/10-K via SEC XBRL company facts. Hover over a quarter for the components and the value per share.
V.Cash Conversion & Earnings Quality
Operating cash flow is positive and large, $6.91 billion over the twelve months to June 2026, because customers prepay (deferred revenue reached $9.7 billion) and depreciation is a non-cash charge. But cash capital spending was $20.6 billion over the same period, so free cash flow was about −$13.7 billion, financed by borrowing and new shares.
- Adjusted EBITDA flatters. Second-quarter adjusted EBITDA of $1,510M compares with a net loss of $626M; the gap is depreciation ($1,393M), net interest ($640M), stock compensation ($165M) and tax ($62M). For a company whose main asset is GPUs with a useful life of a few years, depreciation is a real cost.
- Prepayments support cash flow. Deferred revenue rose $1.37 billion in the first half of 2026 and contributed most of the gap between operating cash flow ($3.66B) and the net loss (−$1.37B) after depreciation.
- Stock compensation was $630M in 2025 (12% of revenue), mostly from awards that vested at the IPO, and $318M in the first half of 2026.
- Capex runs ahead of revenue. Management put second-quarter capital expenditure at $9.4 billion on the August call, more than three times quarterly revenue. Guidance of $35–39 billion for 2026 compares with $14.1 billion paid in the first half.
VI.Balance Sheet, Debt & Financing
| US$ millions | Dec 2024 | Dec 2025 | Jun 2026 |
|---|---|---|---|
| Cash and equivalents | 1,361 | 3,127 | 5,524 |
| Property and equipment, net | 11,915 | 30,557 | 46,736 |
| Debt, carrying value (recourse + non-recourse) | 7,926 | 21,373 | 35,068 |
| Operating lease liabilities | 2,602 | 8,195 | 16,319 |
| Deferred revenue (current + non-current) | 4,100 | 8,200 | 9,700 |
| Shareholders' equity | (414) | 3,335 | 5,024 |
| Goodwill and intangibles | 25 | 1,336 | 1,346 |
| Facility (30 Jun 2026) | Maturity | Effective rate | US$ millions |
|---|---|---|---|
| DDTL 1.0 – 5.0 term loans (recourse) | 2028–2031 | 9–15% | 10,806 |
| DDTL 4.0 (non-recourse) | Mar 2032 | 7% | 2,837 |
| Senior notes 2030–2032 (incl. euro notes) | 2030–2032 | 9–10% | 10,029 |
| Convertible notes 2031 and 2032 | 2031–2032 | 2% | 6,588 |
| OEM and software financing, Magnetar loan | 2026–2030 | 9–12% | 5,291 |
| Total principal | 35,551 | ||
| Added since: 2.875% convertible notes due 2033 (22 Sep 2026) | Apr 2033 | 2.875% coupon | 4,200 |
The balance sheet is the heart of the investment case. Debt principal rose from $21.6 billion at the end of 2025 to $35.6 billion at 30 June 2026, plus $4.2 billion of convertible notes in September. Scheduled repayments are $4.4 billion for the rest of 2026 and $6.2 billion in 2027 (10-Q). Interest: contractual interest was $592 million in the second quarter, and management guided $860–940 million for the third. Much of the debt is secured on GPUs and specific customer contracts (the DDTL facilities), and CoreWeave's corporate credit rating is B+ at S&P (CNBC), below investment grade. In mid-October 2026 the 9.75% notes due 2031 yielded about 15–16% (CNBC), so new unsecured borrowing would now cost far more than the 2% convertible coupons of 2025–26. Management's answer is the backlog: "We are confident in our backlog which more than supports all of our debt obligations" (CEO memo, quoted by CNBC, 14 Oct 2026).
VII.Capital Allocation, Dilution & Insiders
| US$ millions | FY2023 | FY2024 | FY2025 | H1 2026 |
|---|---|---|---|---|
| Share repurchases | – | – | – | – |
| Stock issued for cash (IPO, private placements) | n/a | n/a | 1,491 (IPO) | 2,982 |
| Stock-based compensation | 15 | 31 | 630 | 318 |
| Diluted weighted shares (M) | 192 | 218 | 436 | 539 |
- Dilution is the business model so far. 551.5 million Class A and B shares were outstanding at 31 July 2026; StockAnalysis puts the one-year increase at 76%, mostly from the conversion of preferred stock at the IPO and later share sales. Convertible notes add potential shares: the 2032 notes convert at about $119.60 and the 2033 notes at about $97.85 a share, both above today's price; capped calls offset part of the dilution.
- The equity plan grows automatically by up to 5% of shares outstanding each January until 2035 (10-Q); 63 million shares were available at 30 June 2026.
- Insiders have sold heavily. The IPO lockup ended in mid-August 2025. Insiders sold more than $1 billion in the third quarter of 2025 (Bloomberg, 10 Oct 2025); by June 2026 the founders had sold more than $2.3 billion and Magnetar more than $5.5 billion (Bloomberg, 9 Jun 2026), mostly under pre-arranged 10b5-1 plans. Dataroma's insider summary shows no open-market purchases and 1,933 sales worth about $3.63 billion in the six months to 21 October 2026.
- Control is concentrated. Class B shares carry ten votes each; directors and executive officers held about 72% of the voting power at 15 April 2026, CEO Michael Intrator alone 38.7% (proxy). Class B converts automatically to Class A by 31 May 2032 at the latest.
VIII.Five-Year Revenue-Multiple Projection (Illustrative)
| Input | Value | Type and basis |
|---|---|---|
| Share price (Class A) | $54.87 | Source: Nasdaq close, 30 Oct 2026 |
| 2026 / 2027 consensus EPS | −$5.18 / −$4.68 | Source: Yahoo Finance, retrieved 31 Oct 2026 (23 analysts each; 2027 range −$8.75 to −$1.08). Negative, so no P/E |
| 2026 / 2027 consensus revenue | $12.87B / $26.47B | Source: Yahoo Finance, 40 analysts, retrieved 31 Oct 2026 (2027 range $21.1–31.5B). Context |
| 2026 revenue (start) | $12.8B | Company guidance: $12.4–13.2B (Q2 2026 call, 11 Aug 2026); first half reported $4.65B. Midpoint used |
| Revenue growth to 2031, a year | Bear 18% · Base 22.5% · Bull 35% | Assumption: 2026 revenue of $12.8B (midpoint of the $12.4–13.2B guidance; consensus $12.87B) growing 18–35% a year to 2031: the low end is roughly the 2027 consensus ($26.4B) and then little growth (backlog delivered, few renewals); the high end needs the $104B backlog plus continued new contracts. |
| EV/sales in 2031 | Bear 3.0× · Base 3.7× · Bull 4.5× | Assumption: 3.0–4.5× EV/sales in 2031: about 5–7.5× EBITDA at a 60% adjusted EBITDA margin, the range of capital-heavy data-centre and telecom infrastructure rather than software. Year-5 net debt (ex leases) $70B bear / $50B bull, against about $30B at 30 Jun 2026; share count diluted 3% a year (convertibles, stock pay). |
| Net debt in 2031 (ex leases) | Bear $70B · Base $59B · Bull $50B | Assumption. Today: $35.6B of debt principal less $5.5B cash at 30 Jun 2026 (10-Q), before the $4.2B convertible of Sep 2026; 2026 capex guide $35–39B. Bear = borrowing keeps pace with build-out; bull = free cash flow starts repaying debt |
| Diluted shares in 2031 | 639M | Calculation: 551.5M Class A + B at 31 Jul 2026 (10-Q cover) diluted 3% a year (assumption: convertibles, stock pay, 5% evergreen plan reserve) |
| Required return (present value) | 10% a year | Assumption |
| US$ | Bear | Base | Bull |
|---|---|---|---|
| 2031 revenue | $29.3B | $35.4B | $57.4B |
| × EV/sales | 3.0× | 3.7× | 4.5× |
| = Enterprise value, 2031 | $87.8B | $130.0B | $258.3B |
| − Net debt, 2031 | $70.0B | $59.2B | $50.0B |
| = Equity value, 2031 | $17.8B | $70.8B | $208.3B |
| ÷ Diluted shares (M) | 639 | 639 | 639 |
| Year-5 value per share | $27.92 | $110.74 | $325.77 |
| Implied annual return (5-yr CAGR) | −12.6% | +15.1% | +42.8% |
| Present value at 10% a year | $17.34 | $68.76 | $202.28 |
Why the range is so wide. Leverage magnifies everything. Today's equity (about $30.3B at $54.87) sits on top of about $30B of net debt and $16.3B of data-centre lease liabilities, so the enterprise is worth roughly $60.3B before leases: about 2.3× 2027 consensus revenue. A small change in what the business is worth in 2031 becomes a large change in what is left for shareholders. The table below holds base-case revenue ($35.4B) and varies the two assumptions that matter most.
| EV/sales 2031 ↓ · Net debt 2031 → | $50B | $60B | $70B |
|---|---|---|---|
| 3.0× | $87 (+10%) | $72 (+6%) | $56 (+1%) |
| 3.7× | $126 (+18%) | $110 (+15%) | $95 (+12%) |
| 4.5× | $170 (+25%) | $155 (+23%) | $139 (+21%) |
IX.Return vs. Today's Price
Pathos illustrative Year-5 value range. Red line: today's price. Dashed green line: the base-case Year-5 value discounted at 10% a year.
At $54.87 the base case implies about +15.1% a year (−12.6% bear, +42.8% bull), with no dividend. Discounted at 10% a year, the base Year-5 value is worth $68.76 today, +25% versus the price. Put the other way: a 10% annual return needs a Year-5 value of about $88.37, which at base-case revenue and net debt requires the market to pay about 3.3× 2031 sales. The arithmetic is asymmetric: the bear case loses about 49% of today's price, the bull case multiplies it about 5.9 times, and a deeper bear (slower growth, a lower multiple or more debt) could leave little or nothing for shareholders. Evidence grade: Weak.
X.Superinvestor Holders (Dataroma)
Which of the portfolio managers that Dataroma tracks own CRWV, according to their latest 13F filings? The figures below come only from Dataroma’s CRWV ownership page and activity page.
Source: Dataroma · retrieved 31 Oct 2026 · 13F period Q2 2026 (holdings at 30 Jun 2026)Dataroma shows 1 tracked superinvestor with a position at 30 Jun 2026. The largest weight belongs to David Tepper - Appaloosa Management, at 1.44% of the manager’s reported portfolio. Dataroma also summarises insider trading on its CRWV page: no open-market purchases and 1,933 sales worth about $3.63 billion by insiders in the six months to 21 October 2026 (from Form 4 filings). Dataroma also shows full exits in Q2 2026: Viking Global Investors (Sell 100%).
XI.Key Risks for Value Investors
Leverage and refinancing
$35.6 billion of debt principal (June 2026) plus $4.2 billion since, $10.6 billion due in 2026–27, and bond yields near 15% in October 2026. If capital markets stay closed or costly, growth would have to slow, and in a severe case the equity could be worth little after creditors are paid.
Customer concentration and counterparty risk
Microsoft was 67% of 2025 revenue; three customers made up 72% of second-quarter 2026 revenue. Several large customers (OpenAI, Jane Street) are private companies whose ability to pay depends on their own funding. Contract values are stated as "up to" amounts and depend on CoreWeave delivering capacity on time.
Technology obsolescence
GPUs lose value quickly as Nvidia releases new generations (Rubin is being deployed in the second half of 2026). If contracts are not renewed at similar prices, older equipment may earn too little to cover its depreciation and the debt secured on it.
Competition and pricing
Hyperscalers (Microsoft Azure, Amazon, Google, Oracle) are both customers and competitors, and other GPU clouds such as Nebius and Lambda price aggressively; public on-demand H100 prices in October 2026 ranged from about $3.99 (Lambda) to $6.16 (CoreWeave) a GPU-hour (GPU Cost Compare, list prices). HSBC argues CoreWeave's return on invested capital (8.9% a year, 2026–30 estimates) is below its cost of borrowing.
Execution and power
Revenue depends on data centres being delivered on time with power. Management cited supply-chain complexity in the second quarter; delays defer revenue while interest continues.
Governance and insider selling
Founders control about 72% of the vote through Class B shares and have sold billions of dollars of stock since the lockup ended, while outside shareholders fund growth through dilution.
XII.Advice to Management
Pathos's long-term owner principles judge management by the per-share value it builds over many years, not by quarterly results. Applied to CoreWeave's figures in this note, we would offer the following recommendations.
Set a public ceiling on leverage and a path to repay it
Debt principal was $35.6 billion at 30 June 2026, before a further $4.2 billion of convertible notes in September, and interest expense of $640 million in the second quarter was more than twelve times the operating loss. In October 2026 the yields on CoreWeave's 2031 notes passed 15%. We would publish a maximum ratio of net debt to contracted EBITDA, and the date by which operating cash flow is expected to cover capital spending, so owners and lenders can judge progress against a stated limit rather than the next financing.
Report returns on each data-centre cohort
Capital spending is guided at $35–39 billion for 2026 against $12.4–13.2 billion of revenue, and the GPUs bought today may be superseded within a few years. Adjusted EBITDA (59% of revenue in the second quarter) leaves out the $1.4 billion of quarterly depreciation that measures that wear. We would report, for each year's deployments, the capital invested, contracted revenue, cash returned to date and expected payback, and compare it with the 7–15% effective interest rates on its term loans and senior notes (10-Q).
Reduce dependence on one customer and disclose renewals
Microsoft provided 67% of 2025 revenue and the largest customer was still 36% in the second quarter of 2026. Most revenue backlog ($103.7 billion of performance obligations) runs out within about six and a half years. We would disclose concentration by named customer where contracts allow, the share of backlog due for renewal in each year, and how renewal prices compare with original contract prices.
Align founders' selling with owners' interests
Founders and early holders have sold several billion dollars of stock since the 2025 lockup ended (more than $2.3 billion by the founders by June 2026, Bloomberg), while the company issued new shares and convertible notes to fund growth. Class B shares give the founders about 72% of the vote. We would ask the board to cap annual founder sales under 10b5-1 plans while the company is a net issuer of equity, and to publish the policy.
Measure success per share, after stock pay
Shares outstanding rose about 76% in the year to July 2026 (StockAnalysis), stock compensation was $630 million in 2025, and the equity plan adds up to 5% a year to its reserve automatically. Revenue growth means little to owners if it is matched by dilution. We would make revenue, cash flow and debt per diluted share the headline measures, and count stock pay as an expense in every adjusted figure.
Opinion, not advice. These recommendations are Pathos's opinion, based on long-term owner principles and the figures in this note; they are not investment advice.
XIII.Conclusion
CoreWeave is an extraordinary growth story and a financially fragile one. It has won contracts from the largest buyers of AI computing in the world and turned them into about $104 billion of backlog, but it has paid for its build-out mostly with debt that now costs far more to refinance, and it has yet to earn a profit after interest. Our revenue-multiple projection gives a wide range because the shareholders own what is left after the lenders: a bear case below today's price, a bull case several times higher. For a value-minded reader the deciding evidence is not revenue growth but cash: whether operating cash flow catches up with capital spending, and whether debt can be refinanced on reasonable terms before the first large maturities.
Pathos view (illustrative). CoreWeave has built a large, fast-growing business on contracts with some of the strongest companies in the world, financed largely with debt. On our revenue-multiple projection the base case implies about +15.1% a year (Year-5 range $27–325), but the bear case loses money and a harsher one could leave shareholders little. Evidence: Weak. Revenue, backlog and debt are reported in filings, but earnings are negative, every valuation input is an assumption, and with about $30B of net debt small changes in the multiple swing the equity value a lot. Points worth watching: (a) whether adjusted operating margin reaches the low teens in Q4 2026 as guided, (b) the cost and maturity of new debt as $10.6B falls due in 2026–27, (c) how much of revenue still comes from Microsoft as Meta, OpenAI and others ramp, (d) whether contracts are renewed at similar prices as GPUs age. This is an educational illustration, not a recommendation to buy, sell or hold.
XIV.Sources & Notes
As-of dates. Share price, market capitalisation and price-based multiples use the Nasdaq close of $54.87 on 30 October 2026. Consensus estimates and Dataroma holdings were retrieved on 31 October 2026. Annual figures are from the 2025 Form 10-K via SEC XBRL data; the latest quarter is from the 11 August 2026 results release and the 10-Q for the quarter ended 30 June 2026. Third-quarter 2026 results had not been reported. Ratios and TTM figures are Pathos calculations unless attributed. Verified = taken from a filing or company release; reported = from a named news source; estimate = analyst consensus or Pathos assumption.
- CoreWeave, Inc., Form 10-K for 2025 (filed 2 Mar 2026) and SEC XBRL company facts: revenue, income, cash flow, customer concentration ("approximately 67% of our revenue from our top customer, Microsoft"), OpenAI and Meta commitments, data centres and power, risk factors. Verified.
- CoreWeave Form 10-Q for the quarter ended 30 Jun 2026 (filed 12 Aug 2026): customer percentages, RPO of $103.7 billion and its timing, Meta $21.0 billion order, Jane Street, Nvidia GPUs and private placement, debt table and maturities, interest expense, equity plan, cover-page share count. Verified.
- CoreWeave Q2 2026 earnings release (Form 8-K, Ex. 99.1, 11 Aug 2026): revenue, adjusted EBITDA and operating income, net loss, backlog ("approximately $104 billion"), power, balance sheet and cash flow. Verified. Q4 2025 release (26 Feb 2026): 2025 adjusted EBITDA $3,093M and backlog $66.8B. Verified.
- CoreWeave Q2 2026 outlook presentation and the Q2 2026 call transcript (Investing.com): guidance for Q3 and 2026 (revenue, adjusted operating income, interest, capex $35–39B, exit ARR), power targets, accrued capex. Company guidance (estimate).
- CoreWeave Forms 8-K: 15 Sep 2025 (Nvidia $6.3B capacity order); 30 Oct 2025 (Core Scientific merger terminated); 10 Aug 2026 ($2.6B DDTL 5.5 facility); 22 Sep 2026 ($4.2B 2.875% convertible notes due 2033, conversion price about $97.85). Verified.
- CoreWeave proxy statement (DEF 14A, 22 Apr 2026): beneficial ownership and voting power at 15 Apr 2026 (Nvidia 10.66% of Class A; directors and officers 72.32% of votes). Verified.
- Yahoo Finance, CRWV analysis (retrieved 31 Oct 2026): 2026–2027 consensus EPS and revenue, ranges and analyst counts; price history: closes since the 28 Mar 2025 listing, 52-week range. Estimate (consensus).
- StockAnalysis, CRWV statistics and forecast (S&P Global data, retrieved 21 Oct 2026): share-count change, employees, analyst price targets, next earnings date estimate. Reported.
- CNBC, 14 Oct 2026 (15.5% share fall, bond yields, CEO memo, S&P B+ rating), 15 Oct 2026 (CDS spread above 1,000 bp) and 20 Oct 2026 (HSBC $39 target, ROIC view); Reuters via US News, 16 Oct 2026. Reported.
- Bloomberg, 10 Oct 2025 (insider sales after the lockup) and 9 Jun 2026 via Yahoo Finance (founder and Magnetar sales). Reported.
- GPU Cost Compare, H100 cloud prices (list prices verified 20 Oct 2026): on-demand GPU-hour prices by provider. Reported (list prices, not contract prices).
- SEC XBRL company facts (SEC EDGAR XBRL API, retrieved 31 Oct 2026): quarterly revenue, net income and diluted EPS from Forms 10-Q and 10-K (fiscal Q4 derived as full year minus nine months), and quarter-end equity, goodwill, intangible assets, preferred stock and shares outstanding. Used for the quarterly charts in Section IV.
- Yahoo Finance, CRWV daily price history (retrieved 31 Oct 2026): split-adjusted quarter-end closing prices for the trailing P/E line.
- Dataroma, CRWV superinvestor ownership page and activity page (retrieved 31 Oct 2026): ownership count, rank, % of all portfolios, hold price, holders, % of portfolio, shares, value and recent activity for 13F period Q2 2026 (holdings at 30 Jun 2026). 13F data are filed up to 45 days after quarter-end and cover long US-listed positions only.
Not verified or not used: the identity of the 26% and 10% customers in Q2 2026 (not named); contract prices and renewal terms; the amount drawn on DDTL 5.5 and any debt raised after 22 Sep 2026; long-term consensus EPS (none positive is published for 2026–27, and we found no free source for 2028+); Q3 2026 results (due around 9 November 2026); adjusted EBITDA for 2023.