CoreWeave
A landlord for AI chips: it fills rented data centres with NVIDIA's processors and leases them to AI labs for years at a time, mostly paid for with borrowed money.
The scores
Growth3/4 →
Revenue up 112% in Q2 and 168% in 2025, organically, but three customers are 72% of it and the record is under three years.
Business quality3/4 →
Top-tier rating and rising prices in a shortage, but untested by a glut and attacked by Nebius, Nscale, Google TPUs and its own customers.
Financial position2/4 ↓
The 2026 build and 2027's $6.2B of maturities need money not yet raised, while unsecured debt traded near 15% in October.
Execution record3/4 ↑
Beat or met each 2026 quarterly guide and raised the year, but the record is short and late-2025 data-centre delays slipped deliveries.
Per share capture2/4 ↓
Diluted shares up 8.9% in six months, a 35M-share ATM, three convertibles, and founders holding 72% of the votes.
Potential3/4 ↑
A large upside on a short chain: $103.7B signed and power contracted, needing only delivery, funding and customers paying.
Risk profile2/4 →
Three customers are 72% of revenue, two of the named ones private, NVIDIA the only GPU supplier, and a class action pending.
Valuation3/4 ↑
At about 3 times 2027 revenue estimates the price pays for the plan delivered, not for the 2029 earnings analysts expect.
Evidence quality3/4 →
Revenue, backlog, debt and contracts are filed, but prices, power and services figures are management's word and no counterparty names CoreWeave.
If it works 0.7× · if it fails you keep about 6%.
The Verdict
We rate CoreWeave neutral. It rents NVIDIA's chips to AI labs on contracts they must pay whether they use them or not. The business is mixed: demand is real, prices are rising and more than $100B of future rent is signed, but it is built with debt, and interest still swallows the profit. The price is fair: about right if those signed contracts are delivered. As a bet it is priced for success: the new inference service adds little, and if growth stopped the debt would take most of the value. It turns on whether lenders keep funding the build at a price it can carry.
The long caseIf AI demand keeps outrunning the power and data centres to serve it, the company with GPUs already running gets paid first, and on longer contracts.
The evidence under this - 7 claims
- The 10-Q nets variable consideration from RPO, including service credits and amounts not recognised because of delivery delays, so the figure can shrink if capacity arrives late.
- The September presentation splits the backlog 21% inside 24 months, a different cut from the 10-Q's 41%, so the near-term share depends on which definition is read.
- No contract price is disclosed; the rise is management's figure, and outside indices measure the market, not CoreWeave's book.
- Rothschild & Co Redburn's Sell case on Nebius rests on GPU rental prices falling over its horizon (C58).
- SemiAnalysis sells consulting to GPU clouds and their buyers; it is independent of CoreWeave but has a stake in the market it rates.
- The top tier is now shared: Nebius was added to Platinum, so the rating no longer sets CoreWeave apart alone.
- The same week's selling hit Oracle, Nebius and Dell and came with a hot inflation print, so part of the move was the whole AI-credit trade, not CoreWeave alone.
- Its secured, contract-backed facility DDTL 4.0 was rated A- equivalent and priced below 6% in Q1 (C45).
- Company value here excludes the September convertible and post-June draws (C23, C24), so the multiple on today's debt is higher.
- 2029 earnings are nine analysts' figure for a company that has never earned a GAAP profit.
Capitalised interest of $79M in Q2 and falling borrowing costs (C46) narrow the gap; management expects margins in the low teens by Q4 (C19).
The company
Training a large AI model takes tens of thousands of chips working as one machine for weeks. Few companies can build that themselves, so they rent it. CoreWeave buys NVIDIA's newest chips, wires them into huge clusters inside data centres it leases, and runs software that keeps them healthy and busy.
Its customers are AI labs, big tech companies and trading firms. Without it they would rent from Amazon, Microsoft or Google, from a rival specialist such as Nebius, or wait to build their own.
The evidence under this - 4 claims
- SemiAnalysis sells consulting to GPU clouds and their buyers; it is independent of CoreWeave but has a stake in the market it rates.
- The top tier is now shared: Nebius was added to Platinum, so the rating no longer sets CoreWeave apart alone.
How it makes money
Almost all of its money, 98% in the June quarter, comes from customers who reserve a block of chips for one to six years and must pay whether they use it or not, often with part paid up front. The rest is rented by the hour. Each dollar arrives month by month over the contract, while the chips and buildings are paid for at the start.
The evidence under this - 3 claims
- The 10-Q nets variable consideration from RPO, including service credits and amounts not recognised because of delivery delays, so the figure can shrink if capacity arrives late.
- The September presentation splits the backlog 21% inside 24 months, a different cut from the 10-Q's 41%, so the near-term share depends on which definition is read.
Short contracts are the priciest slice and a small part of the book; the 98% of revenue under longer commitments was priced when signed.
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Committed contracts | 98% | Take-or-pay reservations of GPU clusters for set terms, usually one to six years, often prepaid in part. |
| On-demand | 2% | Pay-as-you-go GPU capacity that customers can stop with little notice. |
Who runs it
The three founders, led by chief executive Michael Intrator, still run it and control it: their special ten-vote shares give insiders about 72% of the votes. Bonuses are paid at the board's discretion and name no sales or profit target, and their shares vest with time. Insiders bought no shares on the market in the past year and sold many, mostly on vesting days or under pre-set plans.
The evidence under this - 4 claims
The bought-to-sold ratio for Q3 2026 was 0.11 against a nine-quarter median of 0.12, ordinary for this company; same-day, same-price officer sales on vesting dates read as tax sales, not views.
Why it wins, and who could beat it
Why customers come back
Right now the scarce thing in AI is not the chip but a powered building ready to run it. CoreWeave has more of that running and contracted than any other specialist, and NVIDIA, an investor, gets it the newest systems early.
The evidence is that it is raising prices, not cutting them: about 25% in July, and outside price indices for rented chips are rising too. The test is a glut: if new supply arrives faster than demand, this edge shrinks.
The evidence under this - 4 claims
- No contract price is disclosed; the rise is management's figure, and outside indices measure the market, not CoreWeave's book.
- Rothschild & Co Redburn's Sell case on Nebius rests on GPU rental prices falling over its horizon (C58).
- Both series track the H100, a 2022 chip; CoreWeave's growth is in Blackwell and Vera Rubin systems, which neither index prices.
- The same Silicon Data page's FAQ still states $2.53, so the page carries two readings and only the dated live one is used.
- No filing gives megawatts, so every figure here is management's.
- The contracted figure moved from about 3.7 GW in the release to 4.2 GW on the call the same day, so the definition is loose.
- SemiAnalysis sells consulting to GPU clouds and their buyers; it is independent of CoreWeave but has a stake in the market it rates.
- The top tier is now shared: Nebius was added to Platinum, so the rating no longer sets CoreWeave apart alone.
The threat
The biggest threat is its own customers. Microsoft, which paid two-thirds of last year's sales, and Meta build their own data centres and rent only to cover shortfalls. Rivals sell the same thing: Nebius now holds the same top rating, and Nscale has huge Microsoft and Anthropic contracts. Google sells its own chips to labs instead of NVIDIA's.
The evidence under this - 5 claims
The rivals: NBIS, GOOGL, MSFT, META, AMZN, ORCL, IREN, and 1 more that is private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| NBIS | Nebius Group N.V.Priced against it: Yes | CoreWeave AI cloud | On sale | Yes |
| Private | NscalePriced against it: Private | CoreWeave AI cloud | On sale | Private |
| GOOGL | Google TPU capacity sold to outside labsPriced against it: Yes | CoreWeave AI cloud | On sale | Yes |
| MSFT · META | Microsoft's and Meta's own data-centre buildsPriced against it: Yes | CoreWeave AI cloud | On sale | Yes |
| AMZN · ORCL | Amazon Web Services and Oracle Cloud InfrastructurePriced against it: Yes | CoreWeave AI cloud | On sale | Yes |
| AMZN · MSFT · GOOGL | Inference platforms of the general clouds and model labsPriced against it: Yes | Managed inference | On sale | Yes |
| IREN | IRENPriced against it: Yes | - | On sale | Yes |
How big it could get
How big this could get
One research firm measured the specialist AI cloud market at more than $25B in 2025, growing about threefold in a year, and the world's energy agency expects the electricity used by AI chips to grow 30% a year. CoreWeave was about a fifth of that market. Nobody publishes how many chips are rented out in total, so the size of the prize is a range, not a count.
The evidence under this - 4 claims
The 2031 figure is Synergy's forecast, not a measurement, and its method is paywalled.
The IEA's own scenarios range widely; the base case is a projection made in April 2025, before the 2026 shortage.
- The 10-Q nets variable consideration from RPO, including service credits and amounts not recognised because of delivery delays, so the figure can shrink if capacity arrives late.
- The September presentation splits the backlog 21% inside 24 months, a different cut from the 10-Q's 41%, so the near-term share depends on which definition is read.
Who has signed
Meta has committed up to about $35B across two orders, OpenAI up to about $6.5B and the trading firm Jane Street about $6B. With others, $103.7B was signed at June 30, and management says more than $25B more was added since. None of the customers has confirmed the contracts in its own filings.
The evidence under this - 4 claims
- 'Up to' amounts are ceilings; neither OpenAI nor Meta has filed anything naming CoreWeave, so the counterparty side is unconfirmed.
- OpenAI is a private company whose ability to pay rests on revenue it is still building; a report on October 8, 2026 that its revenue trailed internal targets sent CoreWeave's shares down.
- The 10-Q nets variable consideration from RPO, including service credits and amounts not recognised because of delivery delays, so the figure can shrink if capacity arrives late.
- The September presentation splits the backlog 21% inside 24 months, a different cut from the 10-Q's 41%, so the near-term share depends on which definition is read.
The results
The numbers
Sales grew 168% in the year to Dec 2025, slower than the 736% in the year before.
Operating margin is −0.9%.
Running costs take 73¢ of each sales dollar, down from 76¢ in the year to Dec 2023.
It used $7.25B of cash in the latest year, so the $3.16B it holds lasts about 0.4 years at that pace.
It owes $18.2B more than the cash it holds, about 6.0 years of the cash its operations bring in.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-18. Fiscal years end in December.
| Dec 2023 | Dec 2024 | Dec 2025 | |
|---|---|---|---|
| Size and growth | |||
| Sales | $229M | $1.92B | $5.13B |
| Sales growth | - | +736% | +168% |
| Profit | |||
| Gross margin | 70% | 74% | 72% |
| Operating margin | −6.1% | 17% | −0.9% |
| Net margin | −259% | −45% | −23% |
| What growth costs, per sales dollar | |||
| Research and development | 9.2% | 2.9% | 6.9% |
| Sales and marketing | 5.7% | 0.9% | 2.8% |
| General and admin | 13% | 6.2% | 13% |
| Pay in shares | 6.6% | 1.6% | 12% |
| Capital spending | 13× | 454% | 201% |
| Cash | |||
| Cash from operations | $1.83B | $2.75B | $3.06B |
| Free cash flow | −$1.11B | −$5.95B | −$7.25B |
| Balance sheet | |||
| Cash and short-term investments | $217M | $1.36B | $3.16B |
| Debt | - | $7.93B | $21.4B |
| Net cash | - | −$6.57B | −$18.2B |
| Returns and shares | |||
| Shares (diluted) | - | - | 436M |
What growth costs
Growth is making it less profitable for now. Office and sales costs are small, but every new site brings rent, power and depreciation before it brings sales, and interest grows with the debt. The operating margin went from positive in 2024 to slightly negative this year; management says it turns once new sites fill.
The evidence under this - 4 claims
Six years is a judgement about how long a GPU earns rent; CoreWeave signed an A100 contract into 2029, but no Blackwell system has yet run six years, and a shorter life would raise depreciation.
The last report (11 Aug)
Quarter ended June 2026, filed figures.
In the June 2026 quarter sales rose 112% to $2,575M, and nearly all of the growth came from existing customers taking more capacity. Before depreciation, interest and tax it kept 59 cents of each sales dollar.
After those it lost $626M. The chips are written off over six years, so new capacity costs money before it earns, and interest of $640M in the quarter ate the rest. It spent $14.1B on equipment in six months against $3.7B of cash coming in.
What it did to the thesisIt confirmed the demand: sales doubled and the signed backlog kept growing. It also showed the cost: losses after interest, spending far above the cash coming in, and a margin still falling. The view stays Neutral because the funding, not the demand, decides the case.
The quarter in detail
Sales growth from the previous quarter, filed figures. The company has restated its own history, so a quarter either side of the restatement may describe a different business; n/c marks one that does not compare.
Earnings per share $-1.03 against $-1.21 expected; $-1.14 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
Six years is a judgement about how long a GPU earns rent; CoreWeave signed an A100 contract into 2029, but no Blackwell system has yet run six years, and a shorter life would raise depreciation.
What management did with the money
Every dollar goes into chips and buildings. It is raised from loans backed by customer contracts, high-interest bonds, convertible bonds and share sales, including $2B to NVIDIA and $1B to Jane Street this year. In September it set up a program to sell up to 35M new shares. It pays no dividend and buys back no stock.
Insiders in 2026Q3: 93 purchases and 862 sales, from their Form 4 filings.
The evidence under this - 4 claims
The price, and where we disagree
In short. Fair. Buyers pay for the signed contracts being delivered on borrowed money, not for the profits analysts expect years later.
The shares fell about a quarter in the week to October 16, to near their lowest in a year, even though sales per share kept rising. What moved was trust in the debt: bond yields near 15% and a credit rating well below investment grade. The chief executive told staff the backlog more than covers what the company owes.
Two tests are close. Q3 results on November 9 show whether profit before interest reaches the guided $200M to $260M, and $6.2B of debt falls due in 2027.
The evidence under this - 5 claims
- The same week's selling hit Oracle, Nebius and Dell and came with a hot inflation print, so part of the move was the whole AI-credit trade, not CoreWeave alone.
- Its secured, contract-backed facility DDTL 4.0 was rated A- equivalent and priced below 6% in Q1 (C45).
A backlog is future revenue before the cost of delivering it and before the capex still to fund; the bond market priced the same debt near 15% that week (C25).
Q1's adjusted operating income of $21M came in the middle of its $0 to $40M guide, so margins have not always beaten.
How we read the multiple
The company has never made a profit after interest, so the useful question is what the price needs. Its value includes the debt, which is larger than the shares, so the stock moves much more than the business does.
- The company, debt included, is worth about $79.1B.
- That is about 3.0 times the $26.56B of 2027 sales analysts expect.
- It is about 4.2 times the yearly billing pace guided for December.
- The share price is about 8.6 times the profit per share expected in 2029.
Against its own history. Including debt it trades near its own short history; the shares alone sit far below it, because the stock fell while debt and sales kept growing.
The evidence under this - 5 claims
- Company value here excludes the September convertible and post-June draws (C23, C24), so the multiple on today's debt is higher.
- 2029 earnings are nine analysts' figure for a company that has never earned a GAAP profit.
The vendor's company value implies about $45.2B of net debt, more than the $35.6B of debt principal less cash, so it appears to count leases; the September convertible and later draws are not in it.
The 2029 figure rests on nine analysts and the 2028 one on fifteen; estimates for a business this young move with every guide.
Capitalised interest of $79M in Q2 and falling borrowing costs (C46) narrow the gap; management expects margins in the low teens by Q4 (C19).
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on company value for each $1 of sales, cheapest first:
| Ticker | Company | company value for each $1 of sales |
|---|---|---|
| AMZN | Amazon.com | 3.7× |
| ORCL | Oracle | 7.1× |
| GOOGL | Alphabet | 9.4× |
| CRWV | CoreWeave | 10.4× |
| MSFT | Microsoft | 12.2× |
| IREN | IREN Limited | 17.8× |
| NBIS | Nebius Group | 30.8× |
The middle of the other 6 is 10.8. CoreWeave, Inc. is 4th cheapest of 7.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Including debt it sits mid-pack among cloud rivals; on shares alone it is far cheaper, because lenders own more of this company than of the big clouds.
Private or foreign rivals that cannot be priced: Crusoe, Lambda, Together AI.
The bet: what it pays if it works
The bet is the new service that sells AI answers rather than chip time. If it grows to the size management's start suggests, it adds only a small slice to a company valued in the tens of billions, so it does not pay much on its own.
If it never grows, shareholders keep the chip-rental business, valued at this year's sales like the cheapest big cloud, which barely covers what the company owes.
| Managed inference revenue in two to three years | Percentage points of 2028 sales needed | Valued like its peers | Like the cheapest |
|---|---|---|---|
| $500.00M | about 1 at 100% kept | 0.7× | 0.6× |
| $1.00B (the big win) | about 2 at 100% kept | 0.7× | 0.6× |
| $2.00B | about 5 at 100% kept | 0.9× | 0.7× |
To pay 2× (a Fair bet) it would need about $10.26B of Managed inference revenue (about 24 percentage points of 2028 sales); to pay 5× (a Cheap bet), about $32.22B of Managed inference revenue (about 76 percentage points of 2028 sales), at the same multiple.
Each row is Managed inference alone, plus the rest of the company (CoreWeave AI cloud) counted at about $47.36B.
The steps it has to take
- Done Managed inference launched, with booked revenue above $100M a year
- In progress Managed inference at $250M a year by the end of 2026 due 2027-02
- Not started CoreWeave reports managed inference revenue in a filing due 2027-12
- Not started Managed inference at $1B a year due 2028-12
The detail behind the label
The evidence under this - 6 claims
- No filing gives a revenue figure for either line, and booked revenue is contracted rather than billed.
- Model labs and the general clouds sell the same per-token answers, often from their own models.
- No filing gives megawatts, so every figure here is management's.
- The contracted figure moved from about 3.7 GW in the release to 4.2 GW on the call the same day, so the definition is loose.
- Analysts' 2026 average of $12.88B sits at the guide's midpoint, so the raise is already in estimates (C28).
- Capex near three times guided revenue assumes third-party data centres arrive on time, which they did not in late 2025 (C49).
The vendor's company value implies about $45.2B of net debt, more than the $35.6B of debt principal less cash, so it appears to count leases; the September convertible and later draws are not in it.
Where we disagree with the market
Most analysts see the leading AI cloud with a huge signed backlog; bond buyers price its debt like a borrower in trouble.
The contracts are real and the prices good; the shares turn on refinancing next year's debt without paying 15% for it.
Bulls treat the backlog as if it were cash, but rent arrives over years while the chips are paid for now. Bears treat the debt as if customers could walk away, but most contracts must be paid regardless.
Against us: the company's average borrowing cost has fallen every year, to about 8% now, and its contract-backed loans have won strong ratings. If that continues, the funding worry fades.
What analysts say, and the evidence
The evidence under this - 6 claims
- The same week's selling hit Oracle, Nebius and Dell and came with a hot inflation print, so part of the move was the whole AI-credit trade, not CoreWeave alone.
- Its secured, contract-backed facility DDTL 4.0 was rated A- equivalent and priced below 6% in Q1 (C45).
- The 10-Q nets variable consideration from RPO, including service credits and amounts not recognised because of delivery delays, so the figure can shrink if capacity arrives late.
- The September presentation splits the backlog 21% inside 24 months, a different cut from the 10-Q's 41%, so the near-term share depends on which definition is read.
The 2031 and 2032 unsecured notes yielded near 15% in October (C25), so new unsecured money would cost far more than the average.
- The newer DDTL 5.5 facility was rated Ba2 and BB+, below investment grade, and priced at SOFR plus 5.5% (C24).
- S&P rates the company itself B+ (C25).
What could go wrong, and how we would know
The risks
- Debt it cannot roll cheaply. With $6.2B due in 2027 and tens of billions still to borrow, refinancing near 15% would hand most of the contracts' profit to lenders.
- A big customer stumbles. Three customers pay 72% of sales; OpenAI and Jane Street are private, and one failing to pay would leave chips idle and debt unpaid.
- Prices fall when supply arrives. July's 25% rise came during a shortage; a wave of new capacity in 2027 could push contract prices back down.
- Chips wear out faster. Profit assumes each chip earns rent for six years; if new generations make them obsolete sooner, write-offs rise.
- One supplier, more shares. Every chip comes from NVIDIA, and a 35M-share sale program and convertible bonds can dilute shareholders.
The evidence under this - 8 claims
- The same week's selling hit Oracle, Nebius and Dell and came with a hot inflation print, so part of the move was the whole AI-credit trade, not CoreWeave alone.
- Its secured, contract-backed facility DDTL 4.0 was rated A- equivalent and priced below 6% in Q1 (C45).
- 'Up to' amounts are ceilings; neither OpenAI nor Meta has filed anything naming CoreWeave, so the counterparty side is unconfirmed.
- OpenAI is a private company whose ability to pay rests on revenue it is still building; a report on October 8, 2026 that its revenue trailed internal targets sent CoreWeave's shares down.
- No contract price is disclosed; the rise is management's figure, and outside indices measure the market, not CoreWeave's book.
- Rothschild & Co Redburn's Sell case on Nebius rests on GPU rental prices falling over its horizon (C58).
Six years is a judgement about how long a GPU earns rent; CoreWeave signed an A100 contract into 2029, but no Blackwell system has yet run six years, and a shorter life would raise depreciation.
What we still do not know
- Which customers are behind the 36% and 26% of sales, and how much of the backlog each holds.
- What price per chip-hour the large contracts actually pay; no customer or index publishes it.
- What the rating agencies would need to see before calling the company's own debt investment grade.
- Whether the new inference service becomes a real business; its figures so far come only from the call.
The evidence under this - 4 claims
- No contract price is disclosed; the rise is management's figure, and outside indices measure the market, not CoreWeave's book.
- Rothschild & Co Redburn's Sell case on Nebius rests on GPU rental prices falling over its horizon (C58).
- The newer DDTL 5.5 facility was rated Ba2 and BB+, below investment grade, and priced at SOFR plus 5.5% (C24).
- S&P rates the company itself B+ (C25).
- No filing gives a revenue figure for either line, and booked revenue is contracted rather than billed.
- Model labs and the general clouds sell the same per-token answers, often from their own models.
What would change our view
- Neutral→AttractiveFunding cost falls and the guide holds
It needs all of:
- ○The company's value, debt included, falls to 7 times its past year's sales, with this year's targets still intact.checked daily
- ○New borrowing over the next two reports costs 8% or less, or a rating agency raises the company's credit rating.watched as it happens
- Neutral→UnattractiveA customer falters or the guide misses
It needs any of:
- ○Profit before interest, as the company measures it, comes in below $200M for the September quarter.checked at the next results, due by 9 Nov
- ○A customer paying more than a tenth of sales cuts, delays or fails to pay for capacity it committed to.watched as it happens
We would also be wrong if
- The yearly billing pace at the end of 2026 comes in below $18.5B.
- New contracts are signed below July's prices.
What is coming
- 9 November 2026 Q3 2026 results: Whether revenue and adjusted operating income land inside the guide, interest stays at or below $940M and the 2026 guide holds.
Every dated event ahead
- 9 November 2026 Q3 2026 results: Whether revenue and adjusted operating income land inside the guide, interest stays at or below $940M and the 2026 guide holds.
- 2 March 2027 Q4 and full-year 2026 results and the 2026 10-K: Year-end run-rate against $18.5B to $19.5B, the 2027 capex and funding plan, and the first auditor opinion on internal control.
- 31 December 2027 2027 debt maturities fall due: Whether $6.2B due in 2027 is repaid from contract cash or refinanced, and at what cost.
Since we wrote this
- 18 OctThesis writtenNeutral, medium convictionDemand, prices and contracts are real and filed, but the equity's value turns on refinancing a debt-funded build that lenders priced near distress in October.
Also checked 1 time with nothing new to act on.
Sources 28
Sources - 28
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-03-02Item 1 business and competition; the competition risk factor naming AWS, Google Cloud, Azure and Oracle; Microsoft at 67% of 2025 revenue; the OpenAI and first Meta order forms; supplier concentration; the six-year depreciation life; 2025 statementsbacksC01C02C10C11C14C16C40C55C56C60
- S02primarycompany filing2026-08-12Q2 statements; committed contracts at 98% of revenue; customers A, B and C; RPO and its timing; the March 2026 Meta and April 2026 Jane Street commitments; debt by facility and maturity; leases signed and not commenced; liquidity; share classes and dilutive securities; material weaknesses; litigation; subsequent financingbacksC01C02C07C09C11C12C13C15C16C20C21C22C24C29C48C51C54C56
- S03primarycompany release2026-08-11adjusted EBITDA and adjusted operating income for Q2, backlog of about $104B, more than $25B of commitments added in early Q3, active and contracted powerbacksC08C17C48C50
- S04primarytranscript2026-08-11CoreWeave Q2 2026 earnings call transcript (FMP, via scripts/transcript_fetch.py)Q3 and raised 2026 guidance, the July price change, contribution margins on Q2 deals, managed inference and non-GPU ARR, contracted power, financing raised; analysts from JPMorgan, Deutsche Bank, Evercore ISI, Barclays, Wells Fargo and Cantor
In the past few months since its launch, booked ARR for our managed inference platform has grown from $1 million to more than $100 million.
backsC04C18C19C47C50 - S05primarycompany release2026-02-262025 revenue, adjusted EBITDA, backlog of $66.8B, active power above 850 MW and contracted power of about 3.1 GWread, but no claim rests on it
- S06primarytranscript2026-02-26CoreWeave Q4 2025 earnings call transcriptthe first 2026 guidance, the Q1 guide, the resolution of the Q3 2025 data-centre delays and capex above guidance
For full year 2026, we expect revenue of $12 billion to $13 billion, representing approximately 140% growth year-over-year at the midpoint.
backsC19C48C49 - S07primarytranscript2026-05-07CoreWeave Q1 2026 earnings call transcriptthe reaffirmed 2026 guide, the Q2 guide, the A- equivalent rating on the $8.5B DDTL 4.0 facility priced at a cost below 6%
We remain firmly on track to reach or exceed our target of more than 1.7 gigawatts by the end of 2026.
backsC45C48 - S08primarycompany release2026-09-17Q3 short-dated contracts near $40M a megawatt a year, prepayments on about 70% of Q2 deals, the backlog's timing, weighted cost of debt by period, quarterly adjusted operating income and capex, the contribution-margin pathbacksC04C06C46
- S09primarycompany release2026-09-22$3.7B of notes, the conversion price, the capped call and its cap, net proceedsbacksC23
- S10primarycompany filing2026-09-17an at-the-market program for up to 35,000,000 Class A shares including forward sales, and the stated use of proceedsbacksC23C59
- S11primarycompany filing2026-04-22ownership and voting power at April 15, 2026, salaries, the discretionary bonus at 100% of target, time-vesting RSUs, named executive officersbacksC30C52C53
- S12primaryindustry data2026-09-23SemiAnalysis, ClusterMAX 3.0 rating of GPU clouds SemiAnalysisCoreWeave Platinum for the third time; Nebius joins it at Platinum. SemiAnalysis sells consulting to clouds and their buyers, so it is independent of CoreWeave but not of the market
CoreWeave remains the industry’s sharpest, sturdiest, most proactive neocloud.
backsC03C36 - S13primaryindustry data2026-04-02Synergy Research Group, neocloud market forecast Synergy Research Groupthe measured size of the neocloud market in 2025, a forecast near $400B by 2031, and CoreWeave named the most direct challenger to the hyperscale clouds
neocloud revenues are scaling at an unprecedented pace, reaching $9 billion in Q4-up 223% year-over-year-and exceeding $25 billion for full-year 2025.
backsC34 - S14primaryindustry data2026-10-17Silicon Data, H100 Rental Price Index (SDH100RT) Silicon Datathe neo-cloud H100 rental index at $2.82 an hour, up 6.4% in seven days; the same page's FAQ text still says $2.53, so the live reading is the one used
As of Oct 17, 2026 2.82 USD / GPU-hour 6.4% (7D)
backsC05 - S15primarygovernment data2025-04-10the base-case growth of electricity used by accelerated servers
Electricity consumption in accelerated servers, which is mainly driven by AI technology adoption (Box 2.1), is projected to grow by 30% annually in the Base Case
backsC35 - S16primaryindustry data2026-04-02SemiAnalysis, The Great GPU Shortage SemiAnalysisthe one-year contract price series, a different measure from the on-demand index; read for the Nebius thesis on 2026-10-02
H100 1-year GPU rental contract pricing has shot up almost 40% from a low of $1.70/hr/GPU in October 2025 to $2.35/hr/GPU by March 2026.
backsC05 - S17primarycompany filing2026-04-30the competition risk factor naming CoreWeave first among specialised AI cloudsbacksC36
- S18primarycompany filing2026-09-18Nscale Form S-1 NscaleMicrosoft statements of work up to about $43.8B and Anthropic up to about $44.6BbacksC37
- S19primarycompany release2026-04-06a large lab taking Google TPU capacity from 2027; read for the Nebius thesis on 2026-10-02
We have signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity that we expect to come online starting in 2027.
backsC38 - S20primarycompany filing2026-07-30Meta's own capital spending and its third-party capacity commitments, which name no providerbacksC39
- S21primarycompany filing2026-07-29Microsoft's own data-centre build and finance-lease commitments; it never names CoreWeavebacksC39
- S22secondarynews2026-10-15the memo; unsecured 2032 debt below 80 cents on the dollar; backlog of $99.4B at Q1. reuters.com itself refused this repository's fetcher, so the syndicated copy was read
The company's deal backlog is large enough to more than support its debt obligations, Intrator told employees in a memo that was seen by Reuters
backsC08C25C44 - S23secondarynews2026-10-14investingLive, 'CoreWeave bonds trade like distressed debt as AI spending outpaces balance sheets' investingLivebond yields on October 14, the S&P B+ rating, the 16% share fall that day; it reports CNBC's story, which refused this repository's fetcher
Its 9.625% 2032 notes yielded about 14.8%, briefly touching 15%, and its 9.75% 2031 notes yielded around 14.9%
backsC25 - S24secondarynews2026-10-14Brannin McBee on investment-grade ratings and insurer capital
The CoreWeave co-founder stated in an interview that winning investment-grade ratings for more of its chip-backed debt would let CoreWeave borrow from insurers, whose capital he described as effectively unlimited.
backsC45 - S25secondarynews2026-09-21the stated bear case on the neoclouds; read for the Nebius thesis on 2026-10-02
The firm flagged three converging pressures: falling GPU rental prices, hyperscalers scaling captive fleets, and rising financing costs on multi-billion-dollar buildouts.
backsC58 - S26primaryvendor data2026-10-18Research pack, frozen at content/CRWV/research-pack.jsonprice and range, the 126-day decomposition, market value and company value, consensus estimates, the analyst split and its series, short interest, the insider window and base rate, the filed cohortbacksC26C27C28C29C31C32C33
- S27primarycompany filing2026-10-18Form 4 filings, read for the Rule 10b5-1 flag via scripts/insider_fetch.py784 open-market sales and no purchases in the year to June 30, 2026; 45.9% of shares sold under plans; Magnetar's discretionary sales in AugustbacksC31
- S28primarycompany release2026-08-10the facility's size, SOFR plus 5.50% pricing, Ba2 and BB+ ratings, a five-year maturity on contracts averaging three years, and more than $30B of capital raised in 2026backsC24