J.P. Morgan reiterates "Overweight" rating on CoreWeave (CRWV.US): Market's debt concerns are "overblown"; project IRR exceeds cost of capital

According to APP Zhihui Finance, JPMorgan reiterated its "Overweight" rating on CoreWeave (CRWV.US) in a research report dated October 16, led by chief software industry analyst Samik Chatterjee.

CoreWeave dropped roughly 20% last week. The catalyst for this decline was concern over rising bond yields, compounded by a forecast released this week by Gartner that questioned whether AI revenues would be sufficient to support healthy returns for NeoCloud (the new cloud service provider). Together, these factors amplified investor concerns over overcapacity and the sustainability of debt financing.

JPMorgan’s stance is clear: it acknowledges that CoreWeave’s valuation multiples depend not only on its own fundamentals but also on investors' overall appetite for its data center debt. However, the firm believes current concerns are “overblown” relative to the business fundamentals.

Five Reasons: Why Concerns are Exaggerated

The report cited five arguments:

First, data center demand remains robust, and in recent months has actually strengthened rather than cooled, with virtually no data supporting “long-term demand uncertainty.”

Second, bringing capacity online is much harder than extrapolating linearly from announced contracted capacity—while it may appear that industry-announced planned capacity can be directly added together as operational capacity, there is a huge execution gap between “contracted” and “energized and online.” CoreWeave’s edge in this step has been repeatedly highlighted in client research.

Third, project-level economics remain healthy; the company is able to simultaneously generate healthy profit margins and cash flow returns, despite initial concentrated investment putting temporary pressure on profit margins and cash flows.

Fourth, in addition to the sound economics of typical 5-year contracts, there is an extra monetization lever—GPUs can be re-contracted after the 5-year term.

Fifth, project-level returns (unleveraged IRR) exceed the company’s cost of capital, and any rise in cost of capital can likely be embedded into pricing for new projects.

The report also hedges: it acknowledges that recent concerns in the debt market have raised capital costs, potentially compressing the scope for positive returns. However, it argues that the favorable backdrop of computing power demand empowers the company with pricing power, allowing it to pass costs on to clients and thereby limit changes to the underlying revenue/profit model.

Capacity Is “Presold,” Not Built Speculatively in Anticipation of Forecast Demand

This is the key line of the report: CoreWeave’s capacity is presold under long-term “take-or-pay” contracts, rather than being built first to match projected demand later.

The Gartner forecast reveals an industry-wide gap: by 2030, AI data center operators will need a cumulative $8.5 trillion in revenue just to cover announced capital expenditures, financing costs, and achieve a 12% ROIC; however, for the same period, Gartner forecasts that enterprise and consumer AI revenues attributable to AI data centers will be only $3.6 trillion at most. JPMorgan points out that this gap signals the risk of “building capacity ahead of demand,” not the “contract-first, online-later” model like CoreWeave’s.

Moreover, contrary to Gartner’s concerns, computing power demand has been accelerating, not slowing, in recent months.

Even if investors remain concerned about demand durability and whether uncontracted capacity can be rented out, the report offers a buffer: in a typical data center’s capital expenditure, over two-thirds is spent on technology and short-lifecycle assets, rather than the facility shell itself—meaning the window to cut capex is relatively short, thus limiting long-term cash outflow risk. In other words, even if demand softens, CoreWeave can quickly curb capex, as it is primarily purchasing chips and other short-lifecycle assets, not locking up capital in the facility shell upfront.

Execution Leadership: Nearly 500 MW Added in a Single Quarter, Operational Capacity Now at 1.5 GW

Turning planned capacity into operational capacity requires strong execution. In just the second quarter of fiscal 2026, CoreWeave added nearly 500 megawatts (MW), bringing operational power to 1.5 gigawatts (GW). As of August 11, contracted power stood at 4.2 GW, and the company’s goal is at least 8 GW by 2030.

Contract Economics: After Paying Down Asset-Level Debt, Parent Still Receives About 16% of Total Contract Value as Cash Flow

According to the company’s indicative contract cash flow: for a 5-year contract, capital expenditure is roughly 51% of the total contract value, about 80% of which is financed by asset-level debt; once asset-level debt is fully repaid, the parent company receives about 16% of the contract’s total value as residual cash flow—even if the capacity is not renewed, this payout is enough to cover around five years of data center obligations.

JPMorgan sees these indicative economics as conservative because the actual working life of GPUs now exceeds the six-year depreciation schedule and contract period: V100s are in use for roughly 8 years, T4s for around 7 years, and contracts for the A100s launched in 2020 already run into 2029. This means that once asset-level debt is cleared, CoreWeave only bears leasing and electricity costs to continue generating cash flow beyond the contract period—“renewal monetization” thus becomes an option independent of the standard 5-year contract.

Project Return: Unlevered IRR “in the Mid-Teens,” Above the 9%–11% Blended Cost of Capital

Based on the report’s indicative estimates, the unlevered IRR for a typical 5-year contract is “in the mid-teens” percent, while the blended cost of capital is 9%–11% (assuming most capex is financed by a combination of parent and asset-backed debt). If the contract is renewed after the first five years, IRR rises further.

The report does not shy away from the risk of rising capital costs but expects the pricing of incremental computing power to absorb higher funding costs, thereby keeping company returns structurally intact.

Burning Cash Not Due to Weak Economics, but Rapid Ramp-Up

The recent capital needs are mainly due to a steep capacity ramp: operational capacity is set to rise from 1.85 GW at year-end to 8 GW by 2030, equivalent to a fourfold increase over four years or roughly 1.5 GW added per year—for comparison, only 500 MW will be added in 2025 and 1 GW in 2026.

Using the report’s simplified economics: once a project is at full capacity, each GW delivers annualized revenues of $1 to $1.5 billion and contract-level cash EBITDA margins around 75%. During ramp years, capex per GW exceeds $3.5 billion. The core issue is timing mismatch—revenues are only gradually released after full capacity, while capex is concentrated upfront in the ramp year. The company thus needs the “base of stable operating projects” to be 4 to 5 times as large as “capacity ramped in the current year” for cash flow to turn positive after covering incremental capex.

Currently, the ratio of operational power at the start of the year to new power added during the year is about 1:1—CoreWeave enters 2026 with about 0.85 GW in operation and expects to add about 1 GW that year. By 2027, this ratio is still expected to be around 1.2 times. According to the report, it won't reach 4.5 times until 2030.

Thus, JPMorgan concludes that the current cash burn is not a result of poor project economics, but the relatively high capacity ramp versus operational capacity each year. The free cash flow timing issue will be largely resolved as CoreWeave scales up.

From another angle, this selloff has clearly had a credit impact. According to domestic media, CoreWeave bond yields have surpassed 15%, and higher capital costs directly compress the safety margin of the aforementioned “residual cash flow”—which is precisely the focus of the bull-bear divide.

It should be noted that JPMorgan’s disclosures state: in the past 12 months, JPMorgan has provided investment banking services to CoreWeave for compensation and expects to seek further compensation in the next three months; it also holds more than 1% of CoreWeave’s common stock and acts as market maker for related securities. Investors should take this report as a reference factor in their investment decisions.