Post-IPO Guidance Hit-Rate
4 of 4
Quarters Beat Since IPO
Key Snapshot
$5.1B
FY25 Revenue (+700% YoY from 2024)
$12–13B
FY26 Revenue Guide (+140%)
$66.8B
Q4 2025 Revenue Backlog
~250K+
GPUs Deployed (YE25)
$21B
Total Debt Outstanding (Q3 2025)
~4.0x
Net Debt / TTM EBITDA
~$1.2B
TTM Interest Expense
1.7 GW
Power Capacity Target YE26
Quarterly Guidance vs. Actuals — 2025
| Quarter |
Metric |
Initial Guide |
Actual |
Beat |
Verdict |
Q1 2025
First post-IPO print |
Revenue | $970–$990M | $981.6M | In-range | Met |
| Adj Op Income | ~$145M mid | $162.6M | +12% | Exceeded |
| Q2 2025 |
Revenue | $1.06–$1.10B | $1.213B | +12% | Exceeded |
| Adj EBITDA | $670–710M | $753M | +8% | Exceeded |
| Q3 2025 |
Revenue | $1.26–$1.30B | $1.365B | +5% | Exceeded |
| Adj EBITDA | $780–820M | $838M | +5% | Exceeded |
| Q4 2025 |
Revenue | $1.48–$1.52B | $1.572B | +4% | Exceeded |
| Adj EBITDA | ~$900M mid | ~$960M | +7% | Exceeded |
| FY2025 Full |
Revenue | $4.9B (initial May) | $5.13B | +4.7% | Exceeded |
| Capex | $20–23B (initial) | ~$10.3B (lowered Q3) | Cut >50% | Capex Shift |
| Adj EBITDA | ~$2.5B implied | ~$2.7B | +8% | Exceeded |
| FY2026 Guide |
Revenue | $12.0–13.0B | — | +140% YoY | Pending |
| Adj EBITDA | $900M–$1.1B | — | Tight margin | Pending |
| Capex | $30–35B | — | 3x FY25 actual | Pending |
Note: FY2026 adj EBITDA guide is notably lower than FY25's run-rate in absolute terms on a margin basis because interest expense from the debt tower is reclassifying through the P&L and newer contracts lease at slightly thinner gross margins.
The Debt Tower — $21B and Counting
💰 Why the debt matters
CoreWeave invented a new financing primitive: GPU-backed debt with customer contracts as collateral. Each facility is secured by specific racks of Hopper/Blackwell GPUs plus the associated take-or-pay contracts from Microsoft, OpenAI, or Meta. The structure funded a fleet that grew from 32K GPUs at IPO to ~250K+ by year-end 2025 — but it also means a single major customer loss could ripple through covenant tests. At ~$21B of debt against ~$2.7B of FY25 adj EBITDA (net leverage ~4.0x), CRWV now sits in the "HY borrower that looks investment grade" zone that cycles of tighter credit have historically punished.
Major Debt Facilities
$8.5B
DDTL 4.0 — "Landmark" Facility
Closed late 2025. First GPU-backed financing to earn investment-grade ratings on the senior tranches (~$5B IG). Blended cost ~5.9% fixed / SOFR+225 bps floating. Maturity Mar 2032. Secured by GPU fleet + specific customer contract assignments.
$7.5B
Blackstone / Magnetar Facility (May 2024)
Co-led by Blackstone Credit and Magnetar Capital. Supports equipment acquisition. Subordinated tranches with higher coupons. Still the original "GPU-backed leverage" template that institutionalized the model.
$2.25B
Senior Notes (2025)
Upsized due to demand. Publicly placed. Helps diversify the financing stack beyond bilateral private-credit facilities. S&P-rated.
$3.0B
Convertible Notes due 2032
Convertible into CRWV equity — reduces near-term interest drag but could dilute shareholders if conversion triggers are met as stock rises.
Interest Expense Trajectory
$104M
Q3 2024
Quarterly Interest Expense
$311M
Q3 2025 · +3x YoY
Quarterly Interest Expense
~$1.2B
TTM Q4 2025
Annualized Interest Expense
~23%
Q4 2025
Interest as % of Revenue
Covenant watch: Debt Service Coverage Ratio (DSCR) covenant of 1.15x kicks in Oct 31, 2027. Until then, unlimited "equity cure" is permitted (management can inject equity to cure test failures). This is structurally favorable — but also flags that lenders know FY26–FY27 is the cash-flow stress window.
Customer Concentration — The Other Big Risk
CoreWeave's S-1 flagged customer concentration as the #1 risk. Pre-IPO, Microsoft was ~85% of revenue. Today it's still ~62–67% of FY25 revenue — but the backlog tells a very different story. Three mega-contracts have rewritten the mix:
Revenue Mix — FY25
Microsoft · ~65%
Customer B · ~15%
Others · ~12%
On-demand · ~8%
Revenue Backlog — $66.8B (Diversified)
OpenAI · $22.4B (35%)
Meta · $35.2B → 2026+ (40%)
Microsoft · ~$12B (18%)
Others · ~$4B (7%)
The backlog mix vs. reported-revenue mix gap is the most important chart for CRWV: it shows that as contracts ramp through 2026–2028, Microsoft's share falls below 30% of revenue, while Meta becomes the largest customer and OpenAI a close second. Diversification is real — but happens on a 3-year lag.
Mega-Contracts Signed
$11.9B
OpenAI (Mar 2025) · 5-year
OpenAI's first major non-Azure compute deal. Announced same week as CRWV's S-1 pricing. OpenAI received ~$350M in CRWV equity as part of the deal.
+$10B
OpenAI Expansion (late 2025)
Top-up to the March deal, bringing total committed OpenAI spend to ~$22.4B — the single largest contract in the backlog.
$14.2B
Meta (Sep 2025)
First major hyperscaler-hosted AI contract for Meta's Llama training + inference. Multi-year term.
+$21B
Meta Expansion (Apr 9, 2026)
Meta commits an additional $21B, bringing its total CRWV spend to ~$35.2B — making Meta CoreWeave's largest single contracted customer.
Bull vs Bear
🐂 Bull Case
1. NVIDIA Preferred Allocation
CRWV gets GPU allocation ahead of hyperscalers on new SKUs (H200, B200, GB200, GB300). Structural: NVIDIA holds equity + wants a non-hyperscaler distribution channel. That's ~6–12 month earlier access to monetize at premium pricing.
2. $66.8B Backlog = 13x FY25 Revenue
Take-or-pay contracted revenue visibility through 2030. Very few growth companies have 13x revenue visibility on the books.
3. Debt Structure Is Unusually Sophisticated
Investment-grade ratings on senior tranches despite sub-IG corporate-level leverage. Lenders have direct claim on specific GPUs and contract cash flows. It works because the underlying assets + cash flows are real.
4. Execution Speed Wins the AI Race
Hyperscalers take 24–36 months to add a GW of AI-capable power. CRWV has done it in 12. In a demand-constrained market, time-to-power = revenue.
🐻 Bear Case
1. $21B Debt vs ~$5B Revenue = Fragile
~4.0x net leverage. A 30% drop in AI demand (hyperscaler pause, OpenAI fundraising hiccup, export controls) cascades through covenant tests. Refi cycles every 3–5 years put terminal value at risk if credit spreads widen.
2. OpenAI Counterparty Risk
OpenAI is projected to lose $14B in 2026 and relies on Microsoft capital to continue operating. ~35% of CRWV's backlog is OpenAI. Any funding hiccup at OpenAI ripples directly into CRWV's revenue visibility.
3. H100 Fleet Depreciation
Current depreciation schedule assumes ~6-year useful life. If Vera Rubin / B300 makes H100 economically obsolete 2–3 years sooner (likely), accelerated impairments flow to P&L and EBITDA covenants compress.
4. Hyperscaler In-Sourcing
Microsoft explicitly listed as a competitor in the S-1. MSFT has the capital to build equivalent Fairwater capacity in-house. As their AI workload stabilizes, CRWV could be downgraded from "partner" to "overflow supplier" at contract renewal.
5. Interest Consuming 20%+ of Revenue
TTM interest expense ~$1.2B against ~$5B revenue. FY26 interest expense likely $2B+ as new capacity draws debt. Free cash flow is negative through at least FY26 and arguably FY27.
Management on the Leverage Question
"Scaling is expensive. If you want to be a leader in AI infrastructure, you need to be willing to invest ahead of demand — and we have structured our financing so that each dollar of debt is backed by a dollar of committed customer revenue."
— Michael Intrator, CEO · CNBC · Apr 2026
"Customer concentration has been the headline risk. At IPO, Microsoft was 85%. By the end of 2025, they represented less than 30% of our remaining backlog. We've diversified faster than we thought possible — OpenAI, Meta, and others."
— Michael Intrator · Fortune interview · Dec 2025
"We raised over $18 billion in 2025 across debt and equity, working with more than 200 financial partners. Cost of debt came down more than 300 basis points over the year. The market has learned to underwrite AI infrastructure."
— Nitin Agrawal, CFO · Q4 2025 earnings call
"This is a violent change in supply. Anybody who thinks AI capex is a bubble doesn't understand what demand looks like at scale — we sell out every new data center before we break ground."
— Michael Intrator · Fortune · Dec 2025
Financial Trajectory
$1.92B
FY24 · +744%
Revenue
$5.13B
FY25 · +167%
Revenue
$12–13B
FY26 Guide · +140%
Revenue
~$2.7B
FY25
Adj EBITDA (~53% margin)
$30–35B
FY26 guide
Capex (3x FY25)
$66.8B
Q4 2025
Revenue Backlog