CRWV: Backlog of $66B Silences the Bears
$66.8 billion. That is CoreWeave's contracted revenue backlog as of April 2026 — nearly 13 times its full-year 2025 revenue of $5.13 billion. While the market obsesses over CoreWeave's negative net income and 4.5x debt-to-equity, it is missing the point entirely. This is not a broken company. This is a capital-intensive infrastructure business executing one of the fastest revenue ramps in enterprise tech history, with its future revenue already locked in by the biggest names in AI.
The contrarian case is simple: buy the backlog, not the P&L. CoreWeave's -36x PE and negative FCF are artifacts of front-loaded depreciation ($821M in Q4 alone) and $387.6M in quarterly interest expense — the direct cost of financing the GPU infrastructure that generates 67.6% gross margins. Gross profit of $1.06 billion in a single quarter from a company that didn't exist as a public entity two years ago. The bears see debt; the data shows a toll booth on the AI compute highway with nine of the top ten AI model providers already paying.
The April 10-11 catalyst pair — a multi-year deal to power Anthropic's Claude models and a $21B expanded agreement with Meta through 2032 — is not a short-term pop. It is structural validation. At $102 and 19.8x trailing sales, CRWV is priced for execution, not perfection. Analysts project a path to $9.5B in revenue by 2028 with positive EPS of $0.68. Investors who wait for profitability will buy it at a far higher price.
Valuation: Expensive Until You Price the Backlog
At first glance, CoreWeave's multiples look like a cautionary tale. P/S of 19.8x, EV/Sales of 26.9x, EV/EBITDA of 57.1x, P/B of 9.3x — every traditional screen flags it as overvalued. The consensus view is that these multiples bake in flawless execution across a volatile AI cycle. That consensus is wrong, for one reason: the backlog.
A $66.8 billion backlog against $5.13 billion in 2025 revenue means CoreWeave has already sold its next 13 years of current revenue. That is not a projection — it is contracted. The company does not need to win new customers to grow; it needs to build and operate the infrastructure it has already been paid to deliver. This reframes the valuation entirely. You are not buying a speculative AI name hoping clients show up. You are buying a utility with a preloaded order book.
Compare to AWS or Azure at their hypergrowth inflection points, where EV/Sales routinely ran 20-30x on the strength of long-term enterprise contracts. CoreWeave's $21B Meta deal alone through 2032 is a seven-year annuity. The bears pricing this at a discount to hyperscalers are ignoring that CoreWeave's customer concentration risk cuts both ways: it also means revenue certainty at a scale most SaaS companies never achieve.
Earnings Performance: Margins Are the Story
CoreWeave grew quarterly revenue from $981.6M in Q1 2025 to $1,571.9M in Q4 2025 — 60% growth in nine months. That is not a rounding error. It is a business doubling while simultaneously expanding gross margins to 67.6%.
The Q4 2025 gross profit of $1.06 billion is the number that matters. At scale, an AI infrastructure provider with 67%+ gross margins is a fundamentally different business than the low-margin colocation or cloud reseller models it superficially resembles. These margins reflect CoreWeave's direct GPU ownership model — no middleman markup, no hyperscaler premium layer.
The headline losses are real but mechanistic. Q4 operating income of -$89.6M and net income of -$451.7M are overwhelmed by $821M in D&A — non-cash charges against the GPU fleet — and $387.6M in interest expense on the debt that financed those GPUs. Strip those out and Q4 EBITDA was $741.8M with a 47.2% margin. The company is generating cash operationally ($3.58/share in operating cash flow) and reinvesting every dollar into capacity. The losses are a balance sheet choice, not a business failure.
Financial Health: Debt-Fueled by Design
Debt-to-equity of 4.5x and negative FCF of -$5.75/share are the two numbers the bears lead with. Both deserve context.
The debt load exists because CoreWeave built $15B+ in GPU infrastructure ahead of demand — a bet that paid off spectacularly. The Anthropic and Meta deals announced this week are direct proof that the supply-side build was rational. The interest expense ($387.6M in Q4 alone, annualizing over $1.5B) is the carrying cost of that bet. The question is not whether the debt is high — it is — but whether the contracted revenue justifies it. $66.8B in backlog against roughly $10B in gross debt is a 6:1 coverage ratio. That is not a credit crisis; it is leverage working as intended.
Negative FCF is a deliberate capex posture, not operational weakness. CoreWeave is in the same position Amazon was in its early AWS years — consuming cash to build infrastructure that generates monopoly-like returns once utilization scales. Operating cash flow of $3.58/share confirms the underlying economics are sound. As GPU capex normalizes relative to revenue, FCF turns positive. Analyst consensus has that happening well before 2028.
The SBC-to-revenue ratio of ~10% warrants monitoring. At $500M+ annualized, it is dilutive. But with 525.7M shares outstanding and a $53.6B market cap, the dilution math is manageable if the revenue ramp continues.
Growth & Competitive Position: The Moat Nobody Talks About
Nine of the top ten AI model providers run on CoreWeave. Read that again. This is not a vendor relationship — it is an ecosystem lock-in. Switching AI infrastructure mid-training-run is not a software toggle. It requires re-engineering workflows, migrating petabytes of model weights, requalifying hardware, and accepting service disruption during the most competitive period in AI history. The switching cost is existential for the customer.
The Anthropic deal is particularly significant beyond its dollar value. Anthropic — the company making Claude, currently one of the leading frontier models — has chosen CoreWeave over AWS, Google Cloud, and Azure for a multi-year compute contract. Anthropic's own investors include Amazon ($4B+) and Google ($2B+). The fact that Anthropic is buying compute from CoreWeave despite hyperscaler pressure from its own cap table is the strongest possible third-party validation of CoreWeave's technical and cost advantage.
The $21B Meta extension through 2032 adds another dimension: durability. Meta's AI buildout under Zuckerberg is a multi-decade infrastructure project. Locking in CoreWeave as a primary compute partner through 2032 is not a tactical decision — it is strategic infrastructure alignment. CoreWeave's backlog is not a coincidence of timing. It is the result of being the only at-scale, GPU-native cloud provider that can match hyperscaler reliability without hyperscaler pricing.
Forward Outlook: The $9.5B Question
Analyst consensus projects CoreWeave revenue reaching approximately $9.5B in 2028 with positive EPS of $0.68. The path there is straightforward: the backlog executes, margins hold, and interest expense shrinks as a percentage of revenue. The May 13, 2026 earnings call is the next major data point — Q1 2026 results will test whether the Q4 momentum continues and provide guidance on new contract velocity.
The bear case is not about execution — even skeptics concede the backlog is real. The bear case is concentration risk (Microsoft historically has been ~60% of revenue), AI capex cycle reversal (if hyperscalers pull back GPU orders), and Nvidia dependency (CoreWeave's model collapses if GPU supply normalizes to commodity pricing). These are genuine risks. They are also priced in at $102 — a stock that traded at $187 earlier in its 52-week range.
At current levels, CRWV trades at roughly 5.6x 2028 consensus revenue of $9.5B. For a business with 67%+ gross margins, a $66.8B backlog, and locked-in relationships with Anthropic, Meta, and eight other top-ten AI providers, that multiple is not demanding — it is reasonable. The stock surged 11% on the latest deal announcements because the market is slowly waking up to what the backlog implies. Investors waiting at the sidelines for a profitable quarter are making the same mistake as those who waited for Amazon to be profitable before buying AWS exposure.
Conclusion
CoreWeave is the most contrarian buy in large-cap tech right now — not because it is cheap by traditional metrics, but because the traditional metrics are measuring the wrong thing. The P&L looks like a money-losing infrastructure startup. The backlog looks like a regulated utility with seven-year contracted revenue from the most cash-rich companies on earth.
The math is not complicated: $66.8B backlog, 67.6% gross margins, 47.2% EBITDA margins, and nine of the top ten AI model providers already on platform. The debt is real and the losses are real, but they are the cost of being the first mover in GPU cloud infrastructure — a position that is increasingly difficult to replicate. Own CRWV before the May 13 earnings call validates the Q1 trajectory. The bears will still be waiting for profitability when the stock is back near its 52-week high.
Sources & References
www.coreweave.com
www.coreweave.com
ir.coreweave.com
Disclaimer: This content is for informational purposes only and does not constitute financial advice. Consult qualified professionals before making investment decisions.