CoreWeave, Inc. Class A Common Stock CRWV
Technology · Software - Infrastructure · Synthos Deep Dive · 2026-07-03
The Overview
CoreWeave rents out specialized computers (Nvidia GPUs) that companies use to build and run AI — think of it as a landlord for AI supercomputers. Business is booming: revenue nearly tripled last year, from about $1.9B to $5.1B, and roughly 70 cents of every sales dollar is gross profit.
The catch is huge, and it is about debt. To buy all those expensive AI computers, CoreWeave borrowed enormous amounts of money — about $30 billion, against a company that still loses money and burns cash. It's like someone earning a great salary but who took out a mortgage many times their income to buy the house: it works beautifully as long as the tenants keep paying and the bank keeps lending — and it can go very wrong fast if either stops.
Our verdict is Watch — an interesting, high-upside AI name, but too financially fragile to buy today. We want to see the cash start covering the debt first.
Here's what our three scores mean in everyday terms:
- Downside Risk 9/10 (very high — this is the headline). Loaded with debt, burns cash, and the stock swings violently (it already fell about in half over the past year). A bad quarter or a tighter lending market could be severe.
- Growth Quality 5/10 (mixed). The growth is spectacular, but it isn't yet quality growth: the company loses money and earns poor returns on all that borrowed capital.
- Exponential Potential 8/10 (high). If it survives and executes, this could genuinely multiply — it's still small relative to the AI build-out, and Nvidia is a big backer.
The one big worry: the debt. Everything depends on customers keeping their contracts and lenders staying friendly.
Putting a number on it: our fair-value estimate is $125 against a current price of $84.23 — real upside if our numbers are right.
Our summary metrics
Net-debt/EBITDA 10.5×, beta 7.1, −55% max drawdown, FCF −$7.3B, current ratio 0.31 — a leveraged, solvency-tail bet.
Revenue +168% and 70%+ gross margin, but net margin −26%, ROIC negative, moat is thin (GPU reselling on borrowed money).
~73% forward revenue CAGR, still accelerating, tiny vs AI-infra TAM, $2B Nvidia backing — genuine multibagger optionality if the debt doesn't sink it.
What does “fair value” mean?
Fair value is Synthos’s estimate of what one share is worth today, based on our model of the company’s future cash generation and the risks to it. It is not a price target or a prediction of where the stock trades next quarter — it’s the price where we think risk and reward are balanced. Above it, you’re paying for outcomes better than our base case; below it, the market is offering a margin of safety.
The model’s inputs and weightings are proprietary — they’re the product. What we publish is the output, We don’t publish a reverse-DCF cross-check for pre-profit companies — negative or missing earnings break that math — so take this number on our modeling alone.
What do the 5 tiers mean? (Core · Tactical · Watch · Hold · Avoid)
Exponential Potential
~73% forward revenue CAGR, still accelerating, tiny vs AI-infra TAM, $2B Nvidia backing — genuine multibagger optionality if the debt doesn't sink it.
What could take this further than the base case — and where the market may be underpricing it. Full forward-growth and acceleration math in Deeper analysis, §4 below.
Deeper analysis
Technicals, fundamentals, valuation, and the full expert-claim evidence panel — the detail behind the numbers above.
Reference table
| Street consensus | $131.5 (high $192 / low $67; 15 Buy · 11 Hold · 1 Sell) — context, not our anchor |
| Valuation | Unprofitable (−$2.75 EPS FY25) · EV/S 12.3× TTM · 6.0× FY26E · 3.1× FY27E · P/S 8.7× · P/B 9.1× · FMP letter rating D+ |
| Technicals | Downtrend — $81.75, −50% off 52-wk high, below 50-DMA ($109) and 200-DMA ($100), RSI 40, −48% 12-mo (QQQ +30%) |
| Conviction | Low / Split — only 3 voices, 5 claims; net +2 bullish, but the highest-skill voice (Jordi Visser, 2.0) is bearish |
| Position sizing | If owned at all: degen/satellite, ≤1% speculative sleeve — not a core holding |
What the experts actually said 16 traceable claims on CRWV · showing the highest-conviction voices
“CoreWeave's commentary is well-supported — he has tons of data backing everything they say on demand and pricing; uses it to counter the compute-bubble bears.”
“Nvidia deepens CoreWeave partnership with $2B investment; CoreWeave adopts Vera CPU and Bluefield storage to add 5+ gigawatts by 2030.”
“Neo-cloud category (CoreWeave, Crusoe, Nebius, Lambda) is durable — running a GPU cluster is like driving an F1 car; their GPUs are utilized 2–3x more per hour, a rare and durable operational advantage.”
“CoreWeave scaling from three regions to 28 online by year-end — delivered 11 in Q1 alone; building at a pace almost larger than some of the three big hyperscalers.”
“Coreweave is a rare pure-play on the AI trend with a strong management team; entered as a pure GPU reseller and now punches above its weight in Nvidia GPU allocation (~11%).”
“CoreWeave cut CapEx guidance from $20B to $12-14B on power-shell delays despite insatiable demand; supply, not demand, will delay high-PE AI names' orders.”
Every claim reconciles to a real claim_id in the Synthos knowledge base — this is the evidence the verdict is built on, not vibes. Management (the company itself) is shown but half-weighted; one cautionary voice is included on purpose.
Price & moving averages 12 months · 50 & 200-day averages · 52-week range
Solid line = price · dashed line = 50-day average · dotted line = 200-day average · the two thin horizontal lines mark the 52-week high and low. Price above both averages is an uptrend.
Data summary: last close $84.23, 4% below the 50-day average ($88), 8% below the 200-day average ($91) — a downtrend. 41% below the 52-week high of $143, 38% above the 52-week low of $61.
Bollinger Bands 20-day average ± 2 standard deviations
The shaded band widens when the stock gets more volatile. Riding the upper edge = strong momentum (sometimes stretched); the lower edge = weak / potentially oversold.
Data summary: price $84.23 is currently inside the band (band $77–$107).
RSI (14) momentum gauge · 0–100
Above 70 (overbought zone, shaded) = overbought, below 30 (oversold zone, shaded) = oversold. Currently 45.
MACD 12 / 26 / 9 · trend & momentum
The MACD line crossing above the signal line (bars flip to the up color) = momentum turning up; crossing below (bars flip to the down color) = turning down. Bar height = the size of that gap.
Data summary: MACD is currently below its signal line by 1.19, negative momentum.
Relative performance vs S&P 500 & its sector (XLK (sector)), set to 100 a year ago
Solid = CRWV · dashed = S&P 500 · dotted = XLK (sector). A rising line means it is beating that benchmark — the sector line shows whether it is a leader or laggard within its own group.
Forward revenue & earnings actual → estimate · "FY" = fiscal year, "E" = estimate
Darker bars = actual results, brighter = analyst estimates. Taller bars to the right = expected growth.
Key stats an RIA wants
1. What it is
CoreWeave (NASDAQ: CRWV) is a specialized "neocloud" — a cloud-computing platform purpose-built for AI. It buys Nvidia GPUs at massive scale and rents them out as high-performance compute (bare-metal and virtual), plus storage, networking and managed AI services (training, inference, VFX/rendering, "Mission Control"). Founded in 2017 as Atlantic Crypto Corporation, rebranded to CoreWeave in 2019, headquartered in Livingston, NJ; IPO'd 2025-03-28. Only ~881 employees against $5.1B of revenue — this is a capital business, not a headcount business. Fiscal year ends December 31.
Revenue mix (FY2025, from filings):
- By segment: FMP provides no product-line breakout (
seg_prodempty). Publicly, revenue is concentrated in a handful of very large customers (hyperscalers and AI labs) on long-term take-or-pay contracts — the concentration is a core risk (§11). - By geography: United States $4.80B (94%) · Non-US $0.33B (6%). Heavily US-centric, with international buildout still early.
The strategic story is a debt-and-capex-funded land grab for AI compute capacity: convert a large contracted backlog into physical data-center power (the panel cites a target of 5+ gigawatts by 2030), financed largely with borrowing collateralized against GPUs and customer contracts.
2. The expert thesis — thin and split (traceable)
This is NOT a high-conviction name in the Synthos KB. Total traceable claims: 5, across just 3 voices, net +2 bullish — and, crucially, the highest-skill voice is bearish. Treat this section as thin, and let the fundamentals/quant lead the verdict.
- The bull case is essentially "Nvidia is all-in." Both bullish claims are Jensen Huang / Nvidia (conviction 85): Nvidia deepened the partnership with a $2B investment, and CoreWeave is adopting Nvidia's Vera CPU and Bluefield storage to help add 5+ gigawatts of AI capacity by 2030 (
jensen_huang-4I7XBgPyRLM:00ab19a076;jensen_huang_ai-4I7XBgPyRLM:fb9967edcf). Honest weighting: Huang is talking his own book — CoreWeave is one of Nvidia's largest GPU customers and an investee, so his endorsement is a supplier promoting a customer. Real signal (Nvidia is putting money in), but conflicted. - The bear case comes from the sharpest voice. Jordi Visser (selection skill 2.0, the highest in this file), bearish, conviction 60 (
jordi_visser_m-XQsVmAFd34U:38a2a79fb1, 2025-12-02): "CoreWeave disappointed investors; be very wary of the companies actually spending the AI capex." This is the precise risk our balance-sheet analysis independently surfaces — the money is being spent, the returns are unproven, and the spenders (not the pick-and-shovel sellers like Nvidia) carry the risk.
Honest composite note. Two bullish claims from a conflicted supplier vs one bearish claim from the highest-skill independent voice is not a conviction signal. The verdict below is fundamentals- and quant-driven, with the KB read as a caution flag, not a green light.
3. Synthos scores & the Bull / Base / Bear cases
The one-glance judgment — three scores, 0–10, each anchored to real metrics (not probabilities we can't honestly calibrate):
| Score | 0–10 | The read |
|---|---|---|
| Downside Risk (lower = safer) | 9 · Very High | Net-debt/EBITDA 10.5×, total debt $29.8B, FCF −$7.3B, current ratio 0.31, beta 7.1, max drawdown −55%. Customer-concentrated, capex-cyclical, and dependent on continued cheap financing. A solvency-tail bet, not a wobble. |
| Growth Quality | 5 · Mixed | Revenue +168% and gross margin ~72% are elite — but net margin −26%, ROIC/ROE negative, EBIT barely positive only because D&A is added back, and the "moat" is renting Nvidia chips on borrowed money. Spectacular growth, poor quality. |
| Exponential Potential | 8 · High | ~73% revenue CAGR FY25→FY30E ($5.1B → ~$80B est), still accelerating, market cap a fraction of the multi-hundred-billion AI-infra TAM, and a $2B Nvidia anchor. The upside magnitude is real — if it survives the leverage. |
The three cases (our own scenario model — assumptions shown; each target is a ~12–24-month fair value). We deliberately do not attach probabilities. With a company this leveraged and this early, the honest output is a range, and the range is enormous by design — that width is itself the thesis. Because the equity sits on top of ~$26B of net debt (and rising), small changes in the revenue-multiple assumption swing the equity value violently.
| Case | Key assumptions | Fair value |
|---|---|---|
| Bull | 5GW buildout executes; backlog converts; FY27 revenue ~$30B; market keeps paying a growth EV/S ~6×; net debt plateaus ~$40B; leverage becomes a feature (cheap capacity moat). | ~$260 (+217%) |
| Base (our anchor) | Revenue ramps roughly to consensus (~$24B FY27), the market applies a disciplined ~4.5× EV/S to a leveraged, lower-return GPU-cloud, net debt ~$40B, modest dilution (~540M shares). | ~$125 (+54%) |
| Bear | Demand air-pocket / GPU oversupply / a large contract slips or a lender tightens; FY27 revenue ~$16B, multiple compresses to ~3.5× EV/S, net debt still $42B — the equity is a thin, distressed sliver. | ~$26 (−68%) |
Synthos fair value = the base case, ~$125 (+54%), with the full $26–$260 span as the honest range. Note the asymmetry: the base implies +54%, but the bear implies −68%, and the bear is a balance-sheet outcome, not a valuation wobble. Our base sits just below the Street's $131.5 consensus, while our bear is far below the Street's $67 low — because we take the leverage seriously. This is a tracked call — the Forecaster Scorecard grades it once it matures.
4. Exponential Potential
Synthos separates compounders (durable high returns on capital) from exponentials (accelerating multi-baggers-from-here). CRWV is a textbook exponential on the top line — and a textbook risk on the balance sheet:
- Forward growth: revenue CAGR FY25→FY30E ~73% ($5.13B → ~$79.6B est, 16 analysts). EPS is estimated to cross to positive around FY28 (−$3.43 FY26E → −$1.06 FY27E → +$3.31 FY28E → +$14.91 FY30E est) — labeled estimates, and the outer-year figures carry wide dispersion and internally inconsistent EBITDA lines, so weight them lightly.
- Acceleration (the 2nd derivative) is still positive: revenue +168% (FY25) with Q1'26 +112% YoY — decelerating in percentage terms (inevitable off a bigger base) but adding enormous absolute dollars. Estimates imply ~$12.7B FY26E → ~$25B FY27E → ~$40B FY28E: still roughly doubling. This is the rare name where the acceleration is real.
- Room to run: at a $44.6B market cap against an AI-infrastructure TAM measured in the hundreds of billions, there is genuine multibagger headroom — the binding constraint is not demand or TAM, it's financing and returns on capital.
- Reinvestment runway: enormous but unproven — capex $10.3B FY25 (2.0× revenue), funded by debt, with FCF −$7.3B. The exponential only pays off if that reinvestment eventually throws off cash faster than the interest bill ($1.23B FY25, ~24% of revenue) compounds.
Exponential Potential: High (8/10) — but it is call-option upside, not compounder certainty. Own it (if at all) understanding you are underwriting both a genuine multibagger and a real chance of a distressed sliver. That is why it lives in the degen sleeve, never the core.
5. Financials (real numbers — FMP annual/quarterly)
- Revenue: FY25 $5.13B, +168% (FY24 $1.92B, itself +737% on FY23 $229M). Q1'26 $2.08B, +112% YoY (vs $982M). Explosive and still accelerating in absolute terms.
- Margins: gross ~72% FY25 (69.4% TTM) — genuinely high. But operating margin ≈ −2.6% TTM and net margin −26% TTM: the D&A on the GPU fleet ($2.45B FY25) and the interest bill swamp the gross profit. EBITDA is positive ($2.47B FY25) but that's before the depreciation that is the real economic cost of a chip fleet.
- Earnings: net loss −$1.17B FY25 (EPS −$2.75); Q1'26 net loss −$740M (EPS −$1.40). Losses are widening in dollars as the buildout scales.
- Cash flow (the crux): operating CF +$3.06B FY25, but capex −$10.3B → FCF −$7.25B (FY24 was −$5.95B). The company consumes billions of external cash per year. This is the single most important number in the whole note — until FCF inflects, the thesis rests on capital-market access.
- Balance sheet: total debt $29.8B, net debt $25.9B, net-debt/EBITDA 10.5×; PP&E $38.8B (the GPU/data-center fleet); current ratio 0.31 (current liabilities $16.4B vs current assets $7.5B — a working-capital stretch); equity just turned positive ($3.34B, from −$414M a year earlier). Interest expense $1.23B FY25.
6. Valuation — priced in or room?
You cannot value CRWV on earnings (it has none), so the market runs on EV/Sales, and the picture depends entirely on which year you use: 12.3× TTM → 6.0× FY26E → 3.1× FY27E. If the revenue ramp is real, the stock is cheap on FY27 sales; if the ramp slips, 12× trailing sales on an unprofitable, over-levered balance sheet is dangerous. That is the whole debate in one line. FMP's model-driven letter rating is D+ (overall score 1/6, red across DCF, ROE, ROA, D/E, P/E, P/B) — reflecting the negative returns and leverage, not the growth. Our base-case ~$125 sits just under the Street's $131.5 consensus; we don't give the bull multiple as much credit as the top-end $192 target, and we take the bear (a balance-sheet outcome) far more seriously than the Street's $67 floor. Not a value stock; a leveraged growth option priced on faith in the ramp.
7. Technicals (from the tech block)
- Trend: down. $81.75 sits below both the 50-DMA ($108.7) and 200-DMA ($100.2) — a bearish configuration (price under both, 50 under 200). MACD −4.7 (negative).
- Location: −50.5% off the 52-week high ($165.2), only +26.6% off the 52-week low ($64.55), with a max drawdown of −55.5% since peak. This is a broken-momentum chart, not a leadership uptrend.
- Momentum: RSI(14) 40 — weak, but not yet oversold (<30), so no mechanical bounce signal.
- Relative strength (the tell, and it's ugly): CRWV −47.6% over 12 months vs QQQ +30.3% and SPY +20.6% — massive underperformance of both the market and its own tech index. Even the 3-month (+4.2% vs QQQ +22.0%) lags badly.
- Read: technicals do not confirm any bull case — the market has been de-rating this stock hard for a year. A Watch verdict is consistent with the chart: wait for the downtrend to stabilize (reclaim of the 200-DMA) and a fundamental FCF tell before engaging.
8. Moat & competitive position
CoreWeave's advantages are real but thin and rentable: (1) early scale and preferential Nvidia access — first in line for new GPUs, reinforced by Nvidia's $2B stake and co-engineering (Vera CPU, Bluefield); (2) purpose-built AI-cloud software/orchestration (fleet/node lifecycle controllers, Tensorizer, Mission Control) that is genuinely more optimized for AI workloads than a general-purpose cloud; (3) contracted backlog giving revenue visibility. The problem: the core asset — Nvidia GPUs — is available to everyone with capital, and CoreWeave competes against hyperscalers with vastly deeper balance sheets (Microsoft Azure, AWS, Google Cloud) and a growing field of neocloud peers. Its edge is speed and specialization funded by leverage, which is not a durable moat if capital gets expensive.
Peer set (FMP-supplied, market cap): the list is a generic "Software — Infrastructure" basket — Fortinet $114B, Synopsys $84B, Motorola Solutions $70B, Block $47B, Infosys $45B, Autodesk $44B, Roblox $40B, Workday $35B, Strategy $30B, Zscaler $24B — none is a true neocloud comp. CRWV's real competitors are the hyperscalers and private neoclouds, which don't appear here; read this peer basket as sector-tagging noise, not a valuation anchor.
9. Management, capital allocation & guidance
- Capital allocation: the entire strategy is the capital allocation — borrow aggressively, buy GPUs, sign long-term contracts, repeat. It is a high-wire, debt-funded land grab. Whether it's brilliant or reckless depends on the returns on that ~$10B/yr of capex, which are not yet proven (ROIC negative today).
- Leadership: co-founder Michael Intrator remains CEO; a dual-class structure (Class A/B) concentrates founder control — normal for a recent IPO, but it means public holders have limited governance leverage on a very risky balance-sheet strategy.
- Insider activity: the most recent Form 4 cluster (filed 2026-07-02) is Chief Strategy Officer Brian Venturo — mostly RSU vesting / option conversions (routine), including one open-market sale of 65,493 shares at $95.69. Modest and largely mechanical; not an alarming discretionary-selling cluster, but worth monitoring given the stock's fragility.
- Guidance gap: no management voice is distilled into the KB yet (last claim 2026-01-26). Forward figures here are analyst consensus (FMP), labeled estimates — not company guidance we've reconciled.
10. Catalysts & what to watch
- Next earnings: 2026-08-11 (Q2'26; Street EPS −$1.06, revenue ~$2.56B). The lines that matter: contracted backlog / remaining performance obligations, capex guidance, and any commentary on financing costs and terms.
- Free-cash-flow trajectory: the make-or-break tell — any sign FCF burn is narrowing as contracts convert would materially de-risk the thesis.
- Financing events: new debt raises, their interest rates, and covenant terms — with $29.8B of debt, the cost and availability of capital is the stock.
- Customer concentration: any disclosure on top-customer mix, renewals, or a large contract win/loss.
- Nvidia relationship: further investment, GPU allocation, or Vera/Bluefield deployment milestones (the bull's whole basis).
Thesis tripwires (what would change the call): a financing round at a punitive rate or with tight covenants; a top-customer contract loss or renewal at worse terms; capex rising while revenue growth slows (returns deteriorating); or FCF burn widening rather than narrowing. Any one of these pushes Watch toward Avoid.
11. Key risks
- Leverage / solvency (the dominant risk): $29.8B debt, net-debt/EBITDA 10.5×, current ratio 0.31, FCF −$7.3B. The business runs on continuous access to cheap capital; a rate spike or credit-market freeze is existential, not cosmetic. This is exactly the risk Jordi Visser flags (
jordi_visser_m-XQsVmAFd34U:38a2a79fb1). - Customer concentration: revenue leans on a few very large customers; one non-renewal materially impairs the model.
- Technology / residual-value risk: GPUs depreciate fast and can be leapfrogged by new Nvidia generations — the $38.8B fleet could lose value faster than modeled, and D&A already swamps operating income.
- Cyclicality / AI-capex air-pocket: if hyperscaler AI capex pauses, an over-supplied, over-levered neocloud is the most exposed link in the chain.
- Valuation de-rating: 12× trailing sales with no profits leaves no cushion; the −48% 12-month move shows how fast this de-rates.
- Conflicted bull signal: the positive KB view is Nvidia (supplier + investor) talking its book, not independent conviction.
12. Verdict, position sizing & monitoring
Watch. CoreWeave is a genuine AI-infrastructure exponential — ~73% forward revenue CAGR, ~72% gross margin, Nvidia's $2B backing, and a base case (~$125) that implies +54%. But the case against buying today is stronger: net-debt/EBITDA 10.5×, FCF −$7.3B, current ratio 0.31, beta 7.1, a −48% 12-month chart in a downtrend below both moving averages, D+ quant rating, and — the tie-breaker — the single highest-skill voice in our knowledge base is explicitly bearish on exactly this "AI-capex spender" profile. The bear case is a −68% balance-sheet outcome, not a valuation wobble. That asymmetry (a solvency-shaped downside against faith-in-the-ramp upside) is a Watch, not a Buy.
- What would move it to Buy — Tactical: two consecutive prints showing FCF burn narrowing, financing raised at reasonable rates, and the technical downtrend stabilizing (reclaim of the 200-DMA). At that point it earns a tiny (≤1%) degen/satellite position — never a core weight.
- Sizing today: zero core; ≤1% only for those explicitly running a speculative AI-infra sleeve who can stomach a total loss.
- Monitoring: re-underwrite on the tripwires in §10; formal re-score each earnings print (next 2026-08-11). This verdict is logged as a tracked Synthos call as of 2026-07-03 at $81.75.
- Single biggest risk: the balance sheet — $29.8B of debt against negative free cash flow means a hostile financing market is an existential, not a cosmetic, event.
Provenance & disclosures
- Traceability: 5 KB claims, breadth 3 voices, top skill 2.0 (Jordi Visser — bearish), last claim 2026-01-26 — all reconciled to real
claim_ids (cited inline). The two bullish claims are Nvidia (Jensen Huang), a supplier/investor talking its book. Fabricated conviction is structurally impossible (claim-ID reconciliation); coverage here is thin and split, so the verdict is fundamentals/quant-driven. - Data as-of: fundamentals 2026-03-31 (Q1'26) · estimates & prices 2026-07-02/03 · expert claims through 2026-01-26. Forward figures are analyst consensus (FMP), labeled as estimates; outer-year estimates carry wide dispersion and internally inconsistent EBITDA lines — weighted lightly.
- Not investment advice. Independent research, educational and informational only, never personalized. Hypothetical/forward figures are labeled; the only performance numbers Synthos will headline are the live, real-money Flagship's.
- Version: 2026-07-03. Prior versions available via the deep-dive version dropdown ("based on the info at the time").