A $104 billion backlog wired to a 16%-yield debt grid — does the megawatt outrun the maturity wall?
CoreWeave has compounded revenue 112%+ YoY for six straight quarters, locked in $104B of contracted RPO (+$25B signed early in Q3 ’26), and scaled past 1.5 GW of active power. Yet the stock just plunged to a 52-week low of $55.60 after its 2031/2032 bond yields spiked to 15–16%, 5-year CDS crossed 1,000 bps, and Bernstein warned of an $85B funding gap. Five analyst lenses, three scenarios, four time horizons.
Gray line = CoreWeave's actual trading path from its March 28, 2025 IPO at $40 → $33.52 April ’25 low → $183.58 June ’25 peak → $141.79 October ’25 52-week high → $104.52 post-Q2 ’26 earnings pop (Aug 11, 2026) → $55.60 close on Oct 22, 2026 amid bond-yield stress. Colored paths = synthesized scenario midpoints forward, probability-weighted (base 50% · bull 25% · bear 25%). Wall Street 12-month consensus ≈ $132 (range $36–$250 across 36–41 analysts).
Re-weight the scenarios
In a highly leveraged neocloud, the distribution of outcomes is wide and sensitive to credit conditions. Drag the sliders to assign your own probabilities to the bear case (refinancing crunch / contract roll-off) and bull case (backlog monetization + cost-of-capital compression); the blended target, chart trajectory, and scenario cards all recalculate live.
Five analyst lenses, five answers
CoreWeave produces one of the widest dispersion cones on Wall Street (12-month targets span $36 to $250). Below are five independent, self-contained institutional lenses reasoned from first principles before synthesis.
The Hyper-Scale Compounder
Six consecutive quarters of 112%+ YoY revenue growth, a $104.2B backlog (+$25B added early Q3 ’26), and FY26 revenue guided to $12.8B midpoint exiting at $19B ARR. Nearly 100% of 2027 capacity is already contracted, while high-margin software layers (Managed Inference >$250M exit ARR; Storage/CPU/Networking >$400M ARR) expand unit economics.
- Demand visibility: $129B+ inclusive of early Q3 signings locks in multi-year hypergrowth.
- Operating leverage: Q2 G&A rose just $4M YoY ($178M vs. $174M) while revenue doubled.
- Nvidia alignment: Priority Blackwell/Rubin allocation + $6.3B residual capacity backstop through 2032.
The Four-Clocks Mismatch
Read the 10-Q footnotes: the $103.7B backlog has a weighted average life of just 31.6 months (2.6 years), yet GPUs are depreciated over 6 years, data centers are leased for 12 years, and $35.1B of debt carries 7–15% coupons ($15B maturing by YE2028). An active megawatt returns ~7.2% unlevered while CRWV’s 2031/2032 bonds trade at 15–16% yields.
- Funding hole: $35–39B FY26 CapEx vs. $679M Q2 OCF (negative ex-$790M prepayments); ~$76–85B external capital still needed.
- Interest coverage 0.20x: Q2 interest ($640M) was 5× Adjusted Operating Income ($128M).
- Renewal cliff: When 3-year contracts expire, 3-year-old GPUs command steep price discounts.
The Self-Amortizing SPV Engine
The credit panic misreads how CoreWeave’s Delayed Draw Term Loans (DDTL 4.0 at $8.5B, DDTL 5.5 at $2.6B) work. Debt is raised inside asset-specific SPVs backed by investment-grade take-or-pay contracts (Microsoft, Meta, OpenAI) that fully amortize principal over the contract term, leaving post-payback GPU cash flows to equity.
- Contracted payback: Take-or-pay cash flows cover capex + debt service inside the initial 3-year window.
- Customer diversification: Top-3 concentration fell to 72% from 71% for a single customer a year ago.
- Valuation reset: At $55.60, EV is down to ~5.9x FY26E sales and ~10.0x FY26E Adj. EBITDA.
The Energized-Power Tollbooth
Raw GPUs commoditize over time, but energized high-density liquid-cooled megawatts cannot be conjured overnight. With 1.5 GW active (>1.85 GW by YE26) and 3.1 GW contracted, plus proprietary Slurm/SUNKs orchestration that delivers 20%+ higher model-training MFU than legacy hyperscalers, CoreWeave owns the scarce physical bottleneck.
- Time-to-power moat: 3.1 GW contracted grid capacity shields CRWV from 4-year utility interconnection queues.
- Cluster reliability: Purpose-built InfiniBand/NVLink fabric beats generic clouds on multi-node uptime.
- Switching friction: $400M+ ARR in attached storage/networking anchors enterprise workloads.
The Credit-Spread Reflexivity Trade
Following June 2026 Nasdaq-100 inclusion, CRWV became the liquid proxy for the entire AI-credit cycle. Hotter September 2026 US CPI pushed Treasury yields up, blowing CRWV’s 5-year CDS past 1,000 bps and knocking the stock 38% below its 200-day SMA ($92.11). In reflexive credit-equity loops, bond Stabilization precedes equity mean-reversion.
- Extreme oversold technicals: RSI washed out after -30% October drawdown on 3.7× average volume.
- Macro rate sensitivity: Every 100 bps move in SOFR/high-yield spreads swings annual interest ~$350M+.
- Catalyst ahead: Mid-November Q3 ’26 print ($3.45–$3.60B revenue guide) tests liquidity runway.
Wall Street 12-month price targets
How sell-side desks are positioned as of October 23, 2026 following the October credit selloff. Bars are sorted low to high; the dashed line marks today's $55.60 close — only Bernstein's freshly cut $51 target sits below the current tape.
October 2026 saw fierce two-way Street action: on Oct 15, Mizuho cut its target to $88; on Oct 19, Goldman Sachs upgraded CRWV from Neutral to Buy (with a $106 target, trimmed from $139) while Cantor Fitzgerald reiterated Overweight at $176; on Oct 22, BofA reiterated Buy at $140 while Bernstein slashed its target from $74 to $51 (Underperform), triggering Thursday's -10.1% slide to $55.60. Full consensus across 36–41 brokers sits near $132 (+137%).
Where the grid leads
Synthesized scenario midpoints (mid-year) vs. today's $55.60 close. These are illustrative valuation frameworks grounded in EV/EBITDA and Non-GAAP EPS multiples, not guaranteed forecasts — five-year outcomes depend on whether contract renewals validate 6-year GPU economic lives.
1 Year
Mid-20272 Years
Mid-20283 Years
Mid-20295 Years
Mid-2031▸ Bull case (25% default weight) — show the assumptions & math
▸ Base case (50% default weight) — show the assumptions & math
▸ Bear case (25% default weight) — show the assumptions & math
Revenue, capex, free cash flow & debt ($B)
The entire CoreWeave debate lives in the violent divergence between these four bars: revenue is exploding ($0.23B → $12.8B), but CapEx ($37B midpoint in 2026E) is running 3× revenue, creating a massive negative FCF trough bridged by a $51.4B+ debt-and-lease stack.
Sources: FY2025 Form 10-K (Rev $229M in ’23, $1.92B in ’24, $5.10B in ’25; FY25 OCF $3.06B vs. CapEx $10.30B = −$7.24B FCF; YE25 debt $29.8B) and Q2 2026 Form 10-Q / earnings release (FY26 revenue guide $12.4–$13.2B, midpoint $12.8B; FY26 CapEx guide $35–$39B, midpoint $37.0B; Q2 ’26 principal debt $35.1B + $16.3B lease liabilities = $51.4B). Note how FCF (olive bar extending below $0) deepens as CapEx front-runs backlog monetization.
EPS path underpinning the targets ($)
Why does a company with 59% Adjusted EBITDA margins post negative EPS today? Because front-loaded 6-year GPU depreciation and $2.5B+ of annualized interest expense overwhelm early-stage cluster revenue. As 1.85+ GW of active power reaches full utilization and initial SPV debt amortizes, Non-GAAP EPS inflects sharply positive from 2027E onward.
In Q1 ’26 and Q2 ’26, CoreWeave reported GAAP diluted EPS of $(1.40) and $(1.14) (Non-GAAP $(1.03) in Q2 ’26); full-year 2026 GAAP consensus is $(4.15) with Non-GAAP near $(3.10). As FY26 Exit ARR ($18.5–$19.5B) flows through in 2027 and DDTL principal amortizes down, base-case Non-GAAP EPS turns positive in 2027E (+$0.85) and scales to ~$8.65 by 2031E. At a 22× exit P/E, $8.65 supports the $190 Base Case 5-year target.
Hypergrowth top line vs. hypergrowth interest
Q2 FY26 year-over-year growth across core financials and frontier capacity lines — read these Triple-Digit rates against a stock down 61% from its 52-week high.
Notice the terracotta bar in the middle: while Revenue (+112%) and Backlog (+246%) are compounding at triple-digit rates, Net Interest Expense (+140% YoY to $640M in Q2 ’26) is growing even faster than revenue. Closing that jaws-of-leverage gap as 1.85+ GW of power turns on in H2 ’26 is the single operational test for the November Q3 print.
Bull vs. Bear
The entire valuation argument compresses into one structural question: is CoreWeave a self-amortizing infrastructure tollbooth backed by the world's richest tech giants, or a duration-mismatched balance sheet borrowing at 15% to buy depreciating silicon?
▲ THE BULL CASE
- $104.2B contracted backlog (+246% YoY) + $25B in Q3 signings. Very few companies in history have had 8× their current-year revenue ($12.8B midpoint) legally contracted in multi-year take-or-pay RPOs with investment-grade hyperscalers and frontier AI labs.
- 2027 capacity is nearly 100% sold out. Per Cantor Fitzgerald and management commentary, virtually all capacity coming online through 2027 is already spoken for, insulating near-term top-line growth from macro jitter.
- Nvidia’s $6.3B residual backstop & strategic halo. Nvidia owns a major equity stake (investing another $2B in Q1 ’26) and committed up to $6.3B through April 2032 to purchase qualifying residual unsold capacity. On Oct 15, CEO Jensen Huang publicly stated CoreWeave’s strength is “improving every day.”
- Energized power is the ultimate moat. Scaling active power to 1.5 GW (>1.85 GW by YE26) with 3.1 GW contracted gives CRWV multi-year physical scarcity value while competitors wait in 4-year grid interconnection queues.
- High-margin software & attachment layers are scaling. Storage, CPU, and networking exceeded $400M ARR in Q2 ’26, and Managed Inference (launched Q1 ’26) hit $100M ARR on its way to $250M+ by year-end — lifting gross margin +40 bps QoQ to 65.9%.
- Extreme valuation compression creates asymmetric upside. Down 61% from its 52-week high to ~5.9× FY26E EV/Sales and ~10.0× EV/EBITDA, with Blackstone COO Jon Gray noting on Oct 22 that the “underlying business is actually quite strong.”
▼ THE BEAR CASE
- The “Four Clocks” duration mismatch (31.6 months vs. 6 & 12 years). Per Q2 ’26 10-Q disclosures, the $103.7B backlog has a weighted average life of just 31.6 months (41% inside 24 months, 39% in months 25–48), whereas GPUs depreciate over 6 years and data center leases run 12 years.
- Bond yields at 15–16% & 5-yr CDS > 1,000 bps. In mid-October 2026, CRWV’s 2031 and 2032 notes traded at 15–16% yields and 5-year CDS priced a >60% 5-year default probability. Credit markets are actively questioning whether unit ROIC covers marginal borrowing costs.
- $76B–$85B estimated funding gap & $15B maturity wall by YE2028. FY26 CapEx guidance of $35–$39B is nearly 3× revenue, and Bernstein estimates CRWV needs ~$85B of external capital to build out its committed backlog.
- Interest expense ($640M/qtr) is 5× Adjusted Operating Income ($128M). Net interest surged +140% YoY in Q2 ’26. Even with 59% Adjusted EBITDA margins, D&A and interest consume 100%+ of GAAP operating profit.
- Operating cash flow relies on customer prepayments. Q2 ’26 OCF of +$679M included $790M of deferred revenue inflows; excluding upfront prepayments, core operating cash flow was slightly negative.
- Customer concentration & custom silicon substitution. Top 3 customers still represent 72% of revenue, and hyperscalers (Microsoft, Meta, Google, AWS) are aggressively deploying custom ASICs to reduce reliance on third-party GPU neoclouds when initial contracts expire.
Risk map — likelihood × impact (3–5 yr horizon)
Unlike asset-light platforms, CoreWeave's primary risks sit directly in the upper-right “Likely × High Impact” and “Possible × High Impact” credit and duration quadrants. Beneath the grid, each card explains exactly “what breaks.”
- Insider lockup / float supply
- Power & transformer delays
- Gross margin D&A drag
- Refinancing & spread shock
- Dilutive equity issuance
- Competitive neocloud price war
- Contract renewal cliff (Month 32)
- Top-3 customer churn / ASICs
- AI capex freeze / counterparty default
Refinancing & spread shock
What breaks: High-yield bond spreads stay at 14–16% into the $15B 2028 maturity wall, making new DDTL debt facilities negative-NPV on marginal megawatts.
Dilutive equity issuance
What breaks: To bridge the ~$35–39B annual CapEx bill and appease credit agencies (BB+/Ba2), CRWV issues billions in stock or convertibles near $55, permanently capping per-share upside.
Contract renewal cliff (Month 32)
What breaks: When 31.6-month average contracts expire, customers refuse to re-lease 3-year-old Hopper/Blackwell GPUs at rates needed to cover 6-year depreciation and 12-year building leases.
Top-3 customer churn / ASICs
What breaks: Microsoft, Meta, or OpenAI (top 3 = 72% of revenue) shift post-contract training and inference to internal custom silicon or owned data centers.
Power & transformer delays
What breaks: Utility interconnection or liquid-cooling supply-chain bottlenecks delay the 1.85 GW YE26 power target, pushing revenue recognition right while debt interest ticks daily.
Gross margin D&A drag
What breaks: Pricier Blackwell and Rubin GPU cohorts accelerate depreciation faster than software attachment ($400M ARR) can offset, compressing gross margins below 62%.
AI capex freeze / counterparty default
What breaks: A macro recession or frontier-model monetization bust causes a major AI lab counterparty to restructure its take-or-pay RPO commitment — piercing the SPV credit thesis overnight.
Competitive neocloud price war
What breaks: Well-funded rivals (Nebius, Lambda, Crusoe, IREN) undercut spot and reserved GPU rental rates, squeezing uncontracted renewal margins.
Insider lockup / float supply
What breaks: Continued insider selling ($2.7B / 24.1M shares sold over a recent 90-day window) weighs on technical momentum during market drawdowns.
The neocloud & credit jargon, decoded
Hover the dotted terms throughout the report, or scan the desk's working definitions here.
- RPO (Revenue Backlog)
- Remaining Performance Obligations — legally binding, multi-year take-or-pay customer contracts not yet recognized as revenue ($103.7B–$104.2B as of Q2 ’26).
- DDTL (Delayed Draw Term Loan)
- Asset-backed project credit facilities (e.g., $8.5B DDTL 4.0, $2.6B DDTL 5.5 at SOFR+5.5%) drawn as GPUs and data centers are delivered against signed contracts.
- 5-Year CDS (Credit Default Swap)
- Tradable credit insurance on CoreWeave's debt. Crossing 1,000 bps in Oct 2026 signaled credit traders pricing >60% cumulative default odds over five years.
- Four-Clocks Mismatch
- The structural duration gap between 31.6-month average customer contracts, 5-year DDTL debt maturities, 6-year GPU depreciation schedules, and 12-year facility leases.
- Active vs. Contracted Power (GW)
- Active power (~1.5 GW in Q2 ’26, >1.85 GW YE26E) is energized and generating revenue today; contracted power (~3.1 GW) includes utility capacity reserved for future buildouts.
- Nvidia Residual Backstop
- A September 2025 commercial agreement (up to $6.3B through April 2032) under which Nvidia purchases qualifying unsold compute capacity — a utilization floor, not a debt guarantee.
- Adjusted Operating Income
- Operating profit before stock-based compensation and one-off items ($128M in Q2 ’26), but after GPU depreciation — the cleanest gauge of pre-interest unit profitability.
- Megawatt Unit Economics
- Each MW requires ~$22.5M of GPU/tech gear + ~$4.0M fit-out ($26.5M/MW total) and generates ~$8.24M/MW-yr of revenue; returns hinge on post-Year-3 GPU residual value.