CoreWeave, Inc. (CRWV)

The panel's take

Verdict: SELL · Conviction: MEDIUM · Last price: USD 86.48 (as of 2026-07-01)

CoreWeave is the batch's most bearish call, and the reason is the balance sheet as much as the chart. The growth is real — Q1 revenue more than doubled (+112%) to $2.08B, and the contracted backlog swelled to $99.4B (OpenAI, Meta, Microsoft, Anthropic) — but this is a leveraged, depreciating-asset business, not a software story. It carries roughly $25-30B of debt against a thin equity cushion, interest expense of $536M in a single quarter consumed about a quarter of revenue, adjusted operating margin collapsed from 17% to ~1%, the GAAP loss widened to $740M, free cash flow ran about -$4.7B in the quarter, and a $4.2B debt maturity looms in 2026 against only ~$2-3B of cash (a 0.31 current ratio). The tape reflects it: down ~53% from the ~$185 post-IPO high, below its trend, on a fresh Supertrend sell — a Stage-4 decline the macro lens calls "a falling knife." Then, on July 1, the catalyst that dropped it another ~13%: Meta — a top-three backlog customer — is reportedly building "Meta Compute" to resell its own capacity, i.e. the customer becoming a competitor. Five of six frameworks are bearish, three at high conviction, several warning of genuine permanent-loss risk if the refinancing or demand assumptions slip. The counterweight is real — the stock is already halved, it's far cheaper than peer Nebius on an enterprise-value basis, and the backlog is take-or-pay — which is why conviction is MEDIUM and one lens reads it as merely fairly priced. But on the weight of a broken tape, a stretched balance sheet, and a fresh competitive threat, the synthesis is SELL.

Key levels

Key resistance starts at ~$98-100 (round-number, near the broken trend that has flipped to overhead supply), then a ~$120 mid-range swing, up to a ~$150-160 distribution zone that would require a full trend reversal to reach. Key support sits at ~$80-82 (a near-term round-number shelf), then ~$67 (near the 52-week low around $64 and a mid-range air pocket), with a deeper IPO-era reference near ~$45. The analyst consensus target spans an enormous USD 45 to 170 (some models run higher still), averaging around USD 140 — well above the price, but those targets largely predate both the Meta Compute news and the neocloud de-rating, so treat the gap as stale rather than as margin of safety; the panel's own value estimates ($55-117) sit far lower.

What legendary investors think

We ran CoreWeave past a panel of six legendary investors' frameworks.

Where they agree — and where they clash

Common ground

• The top line and backlog are genuinely strong: revenue +112%, a $99.4B contracted backlog, reaffirmed $12-13B FY2026 guidance.
• The balance sheet is the central risk: ~$25-30B of debt, ~25% of revenue consumed by interest, a 0.31 current ratio, and a $4.2B maturity due in 2026 against ~$2-3B of cash.
• The economics are deteriorating even as revenue grows: adjusted operating margin fell from 17% to ~1%, the GAAP loss widened, and free cash flow ran about -$4.7B in the quarter.
• The tape is broken — a Stage-4 decline ~53% off the high — and the Meta Compute report is a genuine, not cosmetic, threat to the neocloud demand thesis.

The real debate

• Priced-in or more to fall? Bulls (and the valuation-on-EV read) note the stock is already halved and cheap versus Nebius (~5-6x EV/revenue vs ~60-70x sales); the levered-residual and quality lenses see fair value still below the price with real refinancing risk.
• Backlog durability. The $99B is take-or-pay and hard to cancel near-term; but Meta building competing capacity signals hyperscalers may backward-integrate rather than expand future orders.
• Can it self-fund? Bulls trust backlog-to-cash conversion; bears see a business that needs continuous capital-market access to service debt and fund $31-35B of 2026 capex — fragile if credit tightens.

The question it comes down to: Is CoreWeave an early leader in AI infrastructure whose $99B backlog and post-crash valuation offer asymmetric upside — or a highly levered, negative-cash-flow GPU landlord with a refinancing wall, a customer becoming a competitor, and fair value still below the price, where "cheap" is a value trap until the balance sheet is de-risked?

The numbers

Metric	Value
Price / Market cap	USD 86.48 / ~47B
EV / fwd revenue	~5–6x (EV ~$70–72B incl. ~$25–27B net debt) — vs Nebius ~60–70x P/S
Revenue (Q1 2026)	USD 2.08B (+112% YoY); FY2026 guide $12–13B; exit ARR $18–19B
Margins	Gross ~68% (down from 78%); adj. operating ~1% (from 17%); GAAP net loss -$740M
Interest / FCF	Interest ~$536M/qtr (~25% of revenue); free cash flow ~-$4.7B/qtr
Debt / liquidity	Debt ~$25–30B; current ratio 0.31; $4.2B maturity due 2026 vs ~$2–3B cash
Backlog / concentration	~$99.4B contracted (Microsoft ~67% of FY2025 revenue; OpenAI ~$22B; Meta ~$21–35B)

Figures as of Q1 2026 (quarter ended March 31, reported May 7) / July 1, 2026; sourced from CoreWeave IR/SEC filings, CNBC, Motley Fool, 24/7 Wall St, MarketBeat, Bloomberg via news coverage. Total debt figures vary by source ($25-30B depending on capital-lease inclusion) — flagged, not resolved. The debt inflates enterprise value relative to market cap, so EV/revenue (~5-6x) is the more honest cross-comparison than the 8x trailing P/S — and the reason CRWV screens "cheap" versus Nebius. The Meta Compute report (July 1) is a Bloomberg story with no disclosed pricing or launch — an emerging risk, not yet a confirmed fact pattern; existing take-or-pay contracts are near-term unaffected. Analyst consensus ($135-140 average) predates the news.

The bottom line

CoreWeave is the panel's clearest case of a real growth story colliding with a stretched balance sheet and a broken tape. The bull facts are genuine: revenue up 112%, a $99.4B take-or-pay backlog anchored by OpenAI, Microsoft, Meta, and Anthropic, and — after a ~53% crash — a valuation that, on enterprise value, is a fraction of peer Nebius's. That's enough for the innovation lens to underwrite a multi-year $350-450 case and for the fundamental read to call it "priced-in" rather than expensive. But five of six frameworks, three at high conviction, see a business that loses more money the faster it grows: ~$25-30B of debt, roughly a quarter of revenue consumed by interest, adjusted operating margin collapsing from 17% to 1%, ~$4.7B of quarterly cash burn, and a $4.2B maturity due in 2026 against ~$2-3B of cash — a duration mismatch (long-lived debt against fast-depreciating GPUs whose rental rates have already fallen 50-70%) that the value, GARP, and forensic lenses flag as genuine permanent-loss risk, not just volatility. Layer on ~67% Microsoft revenue concentration, Nvidia sitting on both sides as shareholder and supplier, $3.4B of post-lockup insider selling with zero buys, a Stage-4 chart the macro lens won't touch, and a largest customer now building the competing product, and the balance of risks skews down. What would change the call is a clean refinancing of the 2026 maturity at reasonable rates, margins recovering toward the guided long-term target, and evidence Meta Compute is additive rather than substitutive. What keeps it a sell is that, until the balance sheet is de-risked, "cheap" is a levered bet on continuous capital-market access into a tightening Fed and a broken tape — and the same backlog would be far safer to own once the financing and demand questions are answered. The stock is already halved and the backlog is real, which is why conviction is Medium rather than high — but the direction of least resistance is still lower.