The grid has never been more powerful. The balance sheet has never been more levered.
With a $100B contracted backlog and 130% revenue growth, CoreWeave is the definitive infrastructure layer for the generative AI boom. Yet the stock has whipsawed as the market grapples with its staggering debt load (~$21B+ carrying value) and extreme capital intensity. Is this a generational hyper-growth tech platform, or an aggressively over-levered hardware lessor?
Gray line = CoreWeave's actual price into today (IPO Mar ’25 → $173 peak Jul ’25 → $64 low Dec ’25 → $96.58 now); colored paths = synthesized scenario midpoints forward, probability-weighted (base 50% · bull 25% · bear 25%). Linear scale, mid-year marks. Wall Street 12-month consensus ≈ $132.
Re-weight the capacity outcomes
Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear (debt default) and bull (hyperscaler dominance) 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.
Four analyst lenses, four answers
The same financials support wildly different conclusions depending on which framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target.
The Hyper-Scaler
130% LTM revenue growth with a $100B contracted backlog is unprecedented. CoreWeave is executing massive $6B deals (Jane Street, Anthropic) because they offer custom AI-native orchestration the legacy hyperscalers can't match. As compute capability per megawatt scales, margins expand. The debt is just fuel for a monopoly in the making.
The Debt Skeptic
This is an aggressively levered hardware lessor masquerading as a high-margin software stock. They burned $4.7B in Q1 alone to rack servers that will depreciate rapidly. A 7.4x debt-to-equity ratio utilizing delayed draw term loans means if GPU rental rates dip even 15%, the debt service swallows the equity whole. The downside floor is essentially zero.
The Capacity Glut
Nvidia’s supply constraints are loosening, and hyperscalers (AWS, Azure, GCP) are deploying their own custom silicon while building massive new datacenters. When supply catches up to demand, compute becomes a commodity. Take-or-pay contracts are only as good as the counterparty's willingness to pay; if AI ROI falters, the "backlog" evaporates.
The Power Broker
The real moat isn't the GPU—it's the power grid. CoreWeave holds over 1 gigawatt of active power and 3.5 gigawatts of contracted power in an era where utilities take 5+ years to approve new substations. They own the scarcest physical resource in the AI value chain. The multiple compresses as growth slows, but the asset value is highly defensible.
Wall Street 12-month price targets
What the sell-side expects over the next year. Note the massive spread: the highest target ($250) is almost 4x the lowest target ($67), reflecting the highly polarized fundamental debate.
Select sell-side 12-month targets out of the ~23 active coverage desks. The dispersion is extremely wide: bearish analysts focus on dilution and leverage risks (targeting significantly below today's price), while bulls project hyper-growth executing against the $100B backlog. Firms, ratings, and targets illustrative.
Where the grid leads
Synthesized scenario midpoints (mid-year). Returns shown vs. today's $96.58. Note the asymmetry: the bull case implies 5x upside, while the bear case contemplates near-total equity wipeout due to leverage.
1 Year
Mid-20272 Years
Mid-20283 Years
Mid-20295 Years
Mid-2031▸ Bull case — show the assumptions & math
▸ Base case — show the assumptions & math
▸ Bear case — show the assumptions & math
Revenue, capex, free cash flow & debt ($B)
The tension of the entire stock exists right here: surging revenue vs. astronomical capex and debt requirements. To win the AI infrastructure war, CoreWeave is spending tens of billions more than it brings in.
The extreme capital intensity of the AI era. CoreWeave leverages delayed draw-term loan facilities and prepayments to fund massive upfront purchases of Nvidia GPUs and networking gear (Capex). This pushes Free Cash Flow deeply negative, ballooning Total Debt to over $30B by 2026. The base case relies on FCF inflecting positive by late 2027 as initial datacenter deployments transition to steady-state high-margin revenue. Figures illustrative; debt is total gross facilities.
EPS ladder underpinning the targets ($)
Accounting net income currently reflects massive depreciation and interest expenses, completely hiding the operating leverage. Here is the estimated EPS ladder the targets are actually built on.
CoreWeave is deeply unprofitable on a GAAP basis today due to the astronomical depreciation of server racks and interest on $21B+ of debt. The target prices assume that as datacenters are completed and begin amortizing over long-term take-or-pay contracts, net EPS inflects violently upwards. The base case targets a ~$16 EPS out-year.
Hyper-scale metrics
LTM metrics vs prior year — read these against the bearish fears of leverage.
Growth is staggering across the board, but note the bottom two bars (clay). The $100B backlog is expanding faster than revenue itself as enterprises lock in 5-10 year AI capacity agreements. However, Total Debt is compounding nearly 2x faster than revenue to physically finance those future data centers.
Bull vs. Bear
The entire valuation argument compresses into one disagreement: can CoreWeave outrun its debt obligations with generational AI demand?
▲ THE BULL CASE
- Generational backlog. The $100B backlog secures years of future revenue. Counterparties include fortress balance sheets like Jane Street ($6B agreement) and Microsoft.
- Scarcity value in power. The gating factor for AI isn't silicon, it's electricity. With >3.5 GW of contracted power, CoreWeave bypasses the multi-year utility queue holding back competitors.
- Best-in-class orchestration. They aren't just racking servers; their Kubernetes-native infrastructure trains deep models significantly faster (e.g., DeepSeek-V3 record times) than AWS or Azure.
- Phenomenal unit economics. Gross margins sit at ~65%. As electricity and baseline depreciation per unit of compute decrease, profitability will explode.
- Valuation reset already happened. Stock is down ~44% from its mid-2025 peak due to macro and debt panic. The underlying earnings power is unaffected.
▼ THE BEAR CASE
- Lethal leverage. A ~7.4x debt-to-equity ratio utilizing expensive delayed draw term loans and convertible notes leaves zero margin for error in execution.
- Pricing compression. GPU rental prices are destined to fall as hyperscalers flood the market with new capacity and deploy custom silicon (e.g. Google TPU, Amazon Trainium).
- Cap-ex treadmill. To maintain growth, CoreWeave must perpetually replace obsolete GPUs every 18-24 months, ensuring free cash flow remains chronically negative.
- Counterparty risk. In an "AI Winter" or spending pause, startups will default on their take-or-pay contracts, leaving CoreWeave holding empty datacenters and immense fixed debt.
- Insider selling. Persistent insider sales alongside massive new debt issuance signals caution from those closest to the hardware supply chain.
Risk map — likelihood × impact
Where each risk sits, not just how big it is. The hot upper-right corner is the one that breaks the equity thesis.
- Supply chain bottlenecks
- Hyperscaler custom silicon
- GPU rental rate compression
- Power delivery delays
- Debt covenant breach / refi crisis
- AI capital spending pause
- Draconian AI regulation
GPU rental rate compression
As Nvidia supply catches up to demand and hyperscalers aggressively rack servers, the per-hour price of compute plunges, crushing CoreWeave's operating margins.
Debt covenant breach
A slight miss in revenue or delayed datacenter deployment triggers a restructuring of the $21B+ debt load, wiping out equity holders.
AI spending pause
Enterprises realize generative AI ROI is further out than expected, halting training runs and causing startups to default on CoreWeave's backlog contracts.
Hyperscaler custom silicon
AWS, Azure, and Google successfully migrate massive workloads off expensive Nvidia hardware onto cheaper proprietary ASICs, bypassing CoreWeave's value proposition.
Power delivery delays
Utility companies fail to upgrade substations in time for CoreWeave's new datacenter builds, stranding billions of dollars in idle GPU capital.
Supply chain bottlenecks
Temporary shortages of advanced networking cables, cooling systems, or high-bandwidth memory slow capacity deployment by a quarter or two.
The jargon, decoded
Hover the dotted terms in the metrics, or scan the desk's working definitions here.
- Neocloud
- A new generation of specialized cloud providers (like CoreWeave or Lambda) purpose-built solely for AI and high-performance computing, bypassing legacy enterprise IT bloat.
- Take-or-pay contracts
- Agreements where customers (like OpenAI or Jane Street) must pay for reserved GPU capacity whether they use it or not. The foundation of the $100B backlog.
- Adjusted EBITDA
- Operating profit before interest, tax, depreciation and amortization. Crucial for CoreWeave because it ignores the massive depreciation of servers and the crushing interest expense.
- Free cash flow
- Cash generated from operations minus capital expenditures. For CoreWeave, this is deeply negative because building AI datacenters costs billions upfront.
- Delayed draw term loan
- A debt facility that lets CoreWeave borrow chunks of money gradually as they build out specific datacenters, rather than taking it all at once.
- Gigawatt (GW)
- A billion watts of power. Standard measure for modern AI datacenters. CoreWeave has 3.5 GW contracted — enough to power millions of homes.
- Exit multiple
- The P/E or EV/Revenue ratio assumed at the end of the forecast period. Multiply it by projected metrics to get a target price.
- Prob-weighted
- Each scenario's price × its probability, summed into a single expected value across bear, base and bull cases.