01 · Equity deep-dive — synthesized analyst desk
CRWV
$96.58 ▼ 44% off 52-wk high
NASDAQ · AI CLOUD INFRASTRUCTUREMKT CAP ≈ $52.7B52-WK $63.80 – $173.35AS OF MKT CLOSE JUN 26 ’26

Demand has never run hotter. The balance sheet has never been heavier.

CoreWeave went from a crypto-mining pivot to the fastest cloud ever to $5B in revenue, and sits on a $99.4B contracted backlog with ten customers committed to $1B+ each. It also burns billions in free cash flow, carries ~$25B of debt, and rents fast-depreciating NVIDIA GPUs to the very hyperscalers that could build their own. Is it the toll road for the AI buildout — or a levered bet that the boom outlasts the debt? Five analyst lenses, three scenarios, four time horizons.

The verdict · TL;DR
One question decides the stock: does AI-compute demand keep flowing through CoreWeave's data centers faster than its GPUs depreciate and its debt matures? The backlog (~$99B), the growth (+112% revenue) and the customer roster say indispensable toll road. The leverage (~$25B debt, deeply negative free cash flow, ~$2B/yr interest) and 67% Microsoft concentration say fragile. The upside is real, but so is the path to zero — this is a high-beta call on the durability of the AI capex cycle.
5-yr · prob-weighted
$176
+82% vs $96.58
52-week playback · where the tape sits ▼ Faded from the June rally
$96.58 · Jun 26 ’26 consensus $140 · +45%
$63.80 · 52-wk low $173.35 · 52-wk high
Price history + cone of outcomes · 2025 → 2031
HISTORICALBULLBASEBEARPROB-WTD
$400$300$200 $100$0 202520262027 202820292030 2031 $187 peak · Jun ’25 $64 · 52-wk low $176 $128$142$152 $400 $165 $28 TODAY · $96.58

Gray line = CoreWeave's actual price from the March 2025 IPO ($40) through the $187 peak (Jun ’25), the slide to a $64 52-week low, and back to $96.58 now; colored paths = synthesized scenario midpoints forward, probability-weighted (base 40% · bull 25% · bear 35%). Log-linear, mid-year marks. Wall Street 12-month consensus ≈ $140 (range $36–$303), a “Buy” skew. The bear path bends toward distress because the leverage makes the downside non-linear.

Re-weight the scenarios

Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.

35% bear 40% base 25% bull
Blended 5-yr expected $176 +82% vs $96.58
+112%
Q1’26 Revenue YoY ($2.08B)
$99.4B
Revenue Backlog (+284%)
56%
Adj. EBITDA Margin
−$740M
Q1’26 GAAP Net Loss
~$24.9B
Total Debt (ex-leases)
−$8.6B
TTM Free Cash Flow
~1 GW
Active Power (8 GW target ’30)
10
Customers Committed $1B+
02 · The panel — five ways to read the same tape

Five analyst lenses, five answers

The same fundamentals support wildly different conclusions depending on which framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target. Notably, the panel's center of gravity sits below the Street's ~$140 consensus.

Growth / Momentum PM

The Backlog Believer

A $99.4B contracted backlog (+284% YoY) against $12.5B of 2026 guided revenue is eight years of demand already signed — take-or-pay, with ten customers at $1B+. Revenue +112%, fastest cloud ever to $5B, exit-2027 ARR guided above $30B. This is the picks-and-shovels winner of the largest infrastructure build in tech history; you pay up for the only pure-play at this scale.

12-MO TARGET $185 · ~7x fwd EV/sales
Value / FCF / Quality

The Cash-Flow Skeptic

Adjusted EBITDA looks gorgeous at 56% margin — until you remember it excludes the GPU depreciation and ~$2B/yr of interest that are the whole business. Free cash flow is roughly −$8.6B trailing; there is no GAAP profit anywhere in the forecast window. The backlog is only as good as customers' ability to pay for compute that may be cheaper next year. Quality screens this out.

12-MO TARGET $92 · below today on cash risk
Bear / Disruption Skeptic

The Obsolescence Bear

This is a levered bet on depreciating assets. ~$25B of debt funds GPUs with a 3–4 year useful life against 5-year contracts and 5-year bonds — an asset-liability mismatch that breaks if utilization or residual values disappoint. 67% of 2025 revenue was Microsoft; the hyperscalers are customers and the most likely disruptors as they build in-house. A refi wall in a higher-rate world ends it.

12-MO TARGET $48 · multiple collapses
Moat / Competitive Strategy

The Toll-Road Operator

The moat isn't the chips — anyone can buy NVIDIA. It's the scale, the power contracts (~3.5 GW), the data-center operating know-how, and privileged NVIDIA allocation that let CoreWeave deploy Blackwell faster than rivals can. That's a real, if narrow, advantage with a finite shelf life: durable enough to earn the toll for a few years, not wide enough to guarantee the equity compounds through a downturn.

12-MO TARGET $135 · in line with consensus
Quant / Technical

The Tape Reader

Beta 2.39, ~9.4% of float short, a fresh Nasdaq-100 entrant (Jun ’26) that just round-tripped a 50%+ rally and faded 17% off the highs. The chart is a momentum vehicle, not a value one — it trades on AI-capex sentiment and index flows, not DCF. Respect the trend in both directions; position size for 2.4x market volatility and expect violent two-way moves around every print.

12-MO TARGET $115 · volatility-driven
03 · Wall Street's read

Wall Street 12-month price targets

What the sell-side expects over the next year. Bars are sorted low to high and colored by rating; the dashed line is today's $96.58. The dispersion here is the story — few stocks carry a target range this wide.

Consensus ≈ $140 (+45%) · selected names, full range $36–$303
BUYHOLDSELL
HSBC $36 Mizuho $52 DA Davidson $100 Goldman Sachs $120 Macquarie $125 Morgan Stanley $138 Citi $158 Cantor Fitzgerald $170 Wedbush $205 Rosenblatt $250 TODAY · $96.58

Sell-side 12-month targets — a selection of the ~37–41 firms covering CoreWeave; the mean is ≈ $140 (about +45% above today), rated “Buy/Moderate Buy,” but the absolute range runs $36 (HSBC) to as high as $303. The dashed line marks today's $96.58. Note how the bears (HSBC, Mizuho) sit far below the stock while the bulls (Rosenblatt, Wedbush) more than double it — the disagreement is about survival and multiple, not a few points of growth. Firms, ratings, and targets illustrative and as-of varying 2025–2026 dates.

04 · Price scenarios — 1 / 2 / 3 / 5 years

Where the buildout leads

Synthesized scenario midpoints (mid-year). Returns shown vs. today's $96.58. These are illustrative frameworks, not predictions — five-year outcomes hinge almost entirely on whether the AI-capex cycle keeps feeding the backlog faster than the debt and depreciation bite. The bear tail is genuinely near-zero given the leverage.

1 Year

Mid-2027
Bull$185+92%
Base$128+33%
Bear$62−36%
Prob-wtd$119+23%

2 Years

Mid-2028
Bull$248+157%
Base$142+47%
Bear$50−48%
Prob-wtd$136+41%

3 Years

Mid-2029
Bull$305+216%
Base$152+57%
Bear$40−59%
Prob-wtd$151+56%

5 Years

Mid-2031
Bull$400+314%
Base$165+71%
Bear$28−71%
Prob-wtd$176+82%
Bull case — show the assumptions & math
CoreWeave becomes the default AI-compute utility: revenue compounds ~30%+ toward ~$50B by 2031, the long-term 25–30% operating-margin target is hit, the buildout phase ends so real free cash flow finally appears, and a falling cost of capital lets the debt be refinanced cheaply. The market keeps paying a scarcity premium for the only pure-play at scale.
Revenue ≈ $50B by 2031 × ~5x EV/sales, net of debt → ≈ $400 · 5-yr price CAGR ≈ +33%/yr
Base case — show the assumptions & math
Demand stays strong but the economics stay heavy: revenue grinds to ~$35–40B, adjusted EBITDA scales but GAAP profit stays thin under depreciation and ~$2B+/yr interest, ongoing share issuance dilutes holders, and the EV/sales multiple compresses as growth normalizes. A good business that struggles to turn into a compounding equity.
Revenue ≈ $37B by 2031 × ~3x EV/sales, net of debt & dilution → ≈ $165 · 5-yr price CAGR ≈ +11%/yr
Bear case — show the assumptions & math
The AI-capex cycle cools or GPU utilization and residual values disappoint. The asset-liability mismatch bites: 3–4 year chips against 5-year debt, a refinancing wall in a higher-rate world, and hyperscalers pulling workloads in-house. Equity absorbs the leverage on the way down — the multiple collapses and dilution accelerates. In a true distress scenario the outcome is worse than the midpoint below.
Adj. EBITDA stalls, refi terms harden, multiple collapses → ≈ $28 · 5-yr price CAGR ≈ −22%/yr (tail: lower)
05 · Follow the cash

Revenue, capex, free cash flow & debt ($B)

Where the money actually goes — and the chart that explains the whole debate. Capex towers over revenue, free cash flow runs deeply negative (below the zero line), and debt has climbed from ~$2B to ~$30B in three years. The bull sees pre-funded growth; the bear sees a furnace.

Annual revenue, capex, free cash flow & total debt · 2023 → 2026E
REVENUECAPEXFREE CASH FLOWTOTAL DEBT
$30$20$10$0−$10−$20 2023202420252026E

CoreWeave's engine in one view, and the crux of the stock: revenue (sky) is compounding fast, but capex (clay) runs several times larger every year, so free cash flow (olive, below the zero line) is deeply negative — roughly −$8.6B trailing and widening with the buildout. Total debt (slate) has gone from ~$2B (2023) to ~$30B (2026E). The bull says this is pre-funded, take-or-pay growth that inflects to cash once the build slows; the bear says the capex and debt bars never stop climbing and the contracts can't outrun the interest. Capex reflects total capital deployed (cash plus finance-leased equipment); FCF is approximate (operating cash flow less capex); 2026E figures are guidance midpoints. Directionally — capex >> revenue, FCF deeply negative, debt soaring — the picture holds on any basis.

06 · Earnings power — the metric the targets use

Adjusted EBITDA ladder ($B)

GAAP EPS is negative across the entire forecast, so the price targets are built on EV/EBITDA, not P/E. Here's the adjusted-EBITDA ladder underneath them — with the crucial caveat the bull case tends to skip.

Adjusted EBITDA · reported vs. estimated, 2024 → 2031E
REPORTEDESTIMATE
$0$7$14$21$28 202420252026E2027E2028E2029E2030E2031E $1.2 $3.1 $7.0 $11.5 $15.5 $19.0 $22.5 $26.0

Adjusted EBITDA scales beautifully — from ~$1.2B (2024) toward ~$26B (2031E). But here's the catch the targets ride on: adjusted EBITDA deliberately excludes the GPU depreciation and the ~$2B+/yr of interest expense that are the defining costs of this business. Strip those back in and CoreWeave is loss-making on a GAAP basis throughout the window — reported EPS was about −$3.20 in 2025 and is estimated near −$4.60 in 2026E, improving to roughly −$1.20 in 2027E and only approaching breakeven later. The gap between this soaring olive ladder and those persistent GAAP losses is the bear case. Gray = reported, olive = estimates; figures illustrative, built from guidance and consensus.

07 · Growth scorecard

Everything is growing — including what it owes

Q1 FY26, year-over-year. Most scorecards show only the good lines; this one is honest about all of them. The fastest-growing bars are as much what CoreWeave spends and owes as what it earns — and that tension is the entire stock.

Year-over-year growth by metric · Q1 FY26
BUSINESSFRONTIERCOST OF GROWTH
Adjusted EBITDA +91% Interest expense +103% Revenue +112% Shares outstanding +122% Revenue backlog +284% Capex (Q1 YoY) +305%

Revenue +112% and adjusted EBITDA +91% (olive) are genuinely elite, and the backlog +284% (clay) is the demand signal the bulls cite. But interest expense +103%, shares outstanding +122% and capex +305% (terracotta) are growing just as fast or faster — the cost of financing the build is compounding alongside the business. A company can outgrow its costs, or its costs can outgrow it; which one CoreWeave is, is exactly what's unsettled. Frontier and cost figures are Q1-FY26 YoY; illustrative.

08 · The debate

Bull vs. Bear

The entire valuation argument compresses into one disagreement: is CoreWeave the indispensable toll road for the AI buildout, or a levered bet on depreciating chips that the boom must keep outrunning?

▲ THE BULL CASE

  • An unprecedented backlog. $99.4B of contracted, largely take-or-pay revenue (+284% YoY) — roughly eight years of 2026 sales already signed, with ten customers committed to $1B+ each and ~98% of revenue under contract.
  • Hyper-growth at real scale. Revenue +112% to $2.08B in Q1; the fastest cloud business ever to $5B; management guides exit-2027 annualized revenue above $30B and 1.7 GW online by year-end 2026.
  • Privileged NVIDIA access. NVIDIA is partner and $2B investor; CoreWeave gets among the earliest Blackwell deployments, holds MLPerf records, and earns NVIDIA's top ClusterMAX/Exemplar ratings — a real deployment-speed edge.
  • Cost of capital is falling. A first-of-its-kind $8.5B investment-grade-backed debt facility, an S&P outlook upgrade to positive, and ~80 bps off the cost of debt year-to-date — the financing engine is improving, not seizing.
  • Customers are diversifying. OpenAI (~$22B), Meta (~$35B), Anthropic, Jane Street ($6B) and others are reducing reliance on Microsoft and proving demand is industry-wide, not one-account-deep.
  • A scarce, hard-to-replicate asset. ~3.5 GW of contracted power, deep data-center operating expertise, and Nasdaq-100 membership — you can buy the chips, but not the power, sites and allocation overnight.

▼ THE BEAR CASE

  • The balance sheet is the thesis-killer. ~$25B of debt (closer to ~$35B including leases), debt-to-equity above 7x, a current ratio near 0.3, and interest expense of ~$536M per quarter and rising. Little room for error.
  • Cash burn, not cash flow. Free cash flow is roughly −$8.6B trailing; 2026 capex of $31–35B dwarfs ~$12.5B of revenue; there is no GAAP profit anywhere in the forecast window.
  • An asset-liability mismatch. GPUs depreciate over 3–4 years but fund 5-year contracts and 5-year bonds. If utilization, pricing, or residual values disappoint, the math inverts — and a refinancing wall looms in a higher-rate world.
  • Dangerous concentration. Microsoft was ~67% of 2025 revenue. The hyperscalers are CoreWeave's biggest customers and its most capable disruptors as they build in-house silicon and capacity.
  • Dilution and circular optics. Shares outstanding rose ~122% YoY; NVIDIA funding a key customer/partner invites “circular financing” scrutiny; insiders have been selling.
  • Overhangs stacking up. A securities class-action, sharp margin compression (adjusted operating margin fell from 17% to ~1%), and a contested Core Scientific acquisition all cloud the picture.
09 · Risk map

Risk map — likelihood × impact

Where each risk sits over a 3–5 year horizon, not just how big it is. The hot upper-right corner — likely and high-impact — is the one that decides the stock: for CoreWeave, that's the debt. Note how many serious risks cluster in the high-impact column.

Low impact
Medium impact
High impact
Likely
  • Power & cooling cost creep
  • Customer concentration
  • Shareholder dilution
  • Debt service & refinancing
Possible
  • GPU obsolescence
  • NVIDIA dependence
  • Hyperscaler disintermediation
  • AI demand air-pocket
Tail
  • AI-capex bust
  • Credit-market freeze

Debt service & refinancing

Likely × High

~$25B of debt and ~$2B/yr of interest must be rolled as facilities mature; if rates stay high or sentiment turns, refinancing terms harden and equity absorbs the squeeze.

Hyperscaler disintermediation

Possible × High

Microsoft, Google or Amazon scale their own AI capacity and in-house silicon, pulling workloads back and shrinking CoreWeave's addressable demand.

AI demand air-pocket

Possible × High

A pause in frontier-model capex — or cheaper, more efficient models — softens utilization and pricing just as the largest tranches of capacity come online.

AI-capex bust

Tail × High

The broader AI infrastructure cycle deflates like a classic capex bubble; contracted backlog is only as good as customers' ability and willingness to pay.

Credit-market freeze

Tail × High

A shock closes high-yield and structured-debt markets; a business dependent on continuous external financing cannot fund its build and is forced to retrench.

Customer concentration

Likely × Medium

Microsoft was ~67% of 2025 revenue; renegotiation, delay or loss of a top account would dent revenue and the backlog's credibility at once.

Shareholder dilution

Likely × Medium

Shares rose ~122% YoY; funding the build with continued equity issuance steadily erodes per-share value even if the business performs.

GPU obsolescence

Possible × Medium

Faster NVIDIA generations and a 3–4 year useful life compress residual values, raising depreciation and the risk of stranded, sub-economic hardware.

NVIDIA dependence

Possible × Medium

Allocation, pricing and roadmap all run through one supplier who is also an investor — a privileged relationship today, a single point of failure if it shifts.

Power & cooling cost creep

Likely × Low

Rising energy and data-center operating costs gradually pressure margins, though they're manageable relative to the financing risks above.

10 · Glossary

Plain-English glossary

The terms that recur throughout this dossier, defined without jargon.

Backlog / RPO
Remaining performance obligations — revenue under signed contracts that hasn't been recognized yet. CoreWeave's ~$99.4B backlog is its order book, much of it take-or-pay.
Take-or-pay
A contract where the customer pays for committed capacity whether or not they use it. It makes future revenue more predictable — but only as good as the customer's solvency.
Adjusted EBITDA
Earnings before interest, taxes, depreciation & amortization, with further add-backs (e.g. stock-based comp). For CoreWeave it flatters the picture because it excludes the GPU depreciation and interest that are the core costs.
Free cash flow (FCF)
Operating cash flow minus capital expenditure — the cash a business actually generates after building. CoreWeave's is deeply negative (~−$8.6B trailing) during its buildout.
EV / EBITDA
Enterprise value (market cap plus net debt) divided by EBITDA. The multiple the price targets lean on here, since there's no positive GAAP earnings to anchor a P/E.
EV / Sales
Enterprise value divided by revenue — a valuation yardstick used for fast-growing, not-yet-profitable companies. A high multiple prices in years of future growth.
Neocloud
A new class of specialized cloud providers (CoreWeave, Nebius, Lambda) built specifically for GPU-accelerated AI workloads, rather than general-purpose computing like the big three hyperscalers.
Disintermediation
When a customer cuts out the middleman — here, hyperscalers building their own AI capacity and silicon instead of renting from CoreWeave, removing it from the value chain.
DDTL facility
A Delayed Draw Term Loan — committed debt a company can draw down over time as it needs capital. CoreWeave's $8.5B investment-grade-backed facility is a key funding source for its build.
GPU obsolescence
The risk that AI chips lose value fast as newer, faster generations arrive. With a 3–4 year useful life, residual values and depreciation assumptions drive much of the bear case.
Gigawatt (GW)
A unit of electrical power capacity. CoreWeave measures scale in GW of data-center power (~1 GW active, ~3.5 GW contracted, 8 GW targeted by 2030) — power, not chips, is the binding constraint.
Probability-weighted
A single expected value blending the bull, base and bear targets by how likely each is. The re-weighter at the top lets you set those odds yourself; the dotted chart line is the result.