J.P. Morgan Reiterates "Overweight" Rating on CoreWeave (CRWV.US): Market Debt Concerns "Overdone," Project IRR Exceeds Cost of Capital

J.P. Morgan Reiterates "Overweight" Rating on CoreWeave (CRWV.US): Market Debt Concerns "Overdone," Project IRR Exceeds Cost of Capital In an October 16 research report, J.P. Morgan reiterated its "Overweight" rating on CoreWeave (CRWV.US), with Samik Chatterjee, Chief Analyst for the Software sector, as the lead signatory. CoreWeave has fallen about 20% over the past week. The trigger for this decline was concerns sparked by rising bond yields, compounded by a forecast published by Gartner this week that questioned whether AI revenue would be sufficient to support healthy returns for NeoCloud (new-type cloud service providers). Together, the two amplified investor worries about overcapacity and the sustainability of debt financing. J.P. Morgan's position is clear: it acknowledges that CoreWeave's valuation multiple depends not only on its own fundamentals but is also constrained by investors' overall appetite for its data center debt; however, it believes current concerns are "overdone" relative to the business fundamentals. Five Reasons: Why the Concerns Are Overstated The report offers five arguments: First, data center demand is robust and has strengthened rather than cooled in recent months, with almost no data points to support "long-term demand doubts." Second, actually bringing capacity online is far harder than linearly extrapolating from announced contracted capacityit may seem possible to simply sum up the industry's announced planned capacity into actual capacity, but there is a huge execution gap between "signed" and "powered on and online," and CoreWeave's lead in this link has been repeatedly mentioned in its customer surveys. Third, project-level economics remain solid, and the company can simultaneously generate healthy margins and cash flow returns, although concentrated upfront investment will bring temporary pressure on margins and cash flow. Fourth, in addition to the solid economics of a typical 5-year contract, there is an additional monetization leverGPUs can be re-contracted beyond the 5-year contract term. Fifth, project-level returns (unlevered IRR) exceed the company's cost of capital, and increases in the cost of capital are likely to be embedded in the pricing of new projects. The report also leaves room: it acknowledges that recent debt market concerns have pushed up the cost of capital, which may compress the space for generating positive returns; but it believes the favorable backdrop of computing demand gives the company pricing power to pass costs on to customers, thereby limiting changes to the fundamentals of its revenue/profit model. Capacity Is "Pre-Sold," Not Built First Based on Forecast Demand and Then Waiting for Customers This is the most critical sentence in the entire piece: CoreWeave's capacity is pre-sold under long-term "take-or-pay" contracts, rather than being built first and then matched to estimated demand. What Gartner's forecast reveals is an industry-level gap: AI data center owners need to cumulatively generate $8.5 trillion in revenue by 2030 to cover announced capital expenditures, financing costs, and achieve a 12% return on invested capital (ROIC); yet the institution forecasts that enterprise and consumer AI revenue attributable to AI data centers over the same period will be at most $3.6 trillion. J.P. Morgan therefore points out that this gap points to the risk of "building capacity ahead of demand," not the risk of CoreWeave's "sign first, then bring online" model. Moreover, contrary to Gartner's concerns, computing demand in recent months has been accelerating rather than decelerating. Even if investors remain worried about demand durability and whether uncontracted capacity can be leased out, the report offers a buffer: in a typical data center's capital expenditure, more than two-thirds goes to technology and short-cycle assets rather than the building shell itselfmeaning the window for compressing capital expenditure is short, thereby limiting the risk of long-term cash outflows. In other words, even if demand does weaken, CoreWeave can relatively quickly rein in capital expenditure because it is mainly buying short-cycle assets such as chips, rather than sinking money into the building shell all at once. Execution Lead: Nearly 500 MW Added in a Single Quarter, Operating Power Already at 1.5 GW Converting planned capacity into operating capacity requires extremely strong execution. CoreWeave added nearly 500 megawatts (MW) in the second quarter of fiscal 2026 alone, bringing operating power to 1.5 gigawatts (GW); contracted power as of August 11 was 4.2 GW, while the company targets at least 8 GW by 2030. Contract Economics: After Repaying Asset-Level Debt, the Parent Company Still Receives Residual Cash Flow of About 16% of Total Contract Value According to the company's illustrative contract cash flow: for a 5-year contract, capital expenditure is about 51% of total contract value, of which about 80% is financed by asset-level debt; after the asset-level debt is fully repaid, the parent company can obtain residual cash flow of about 16% of total contract valueeven if the capacity is not renewed, this money is enough to cover about 5 years of data center obligations. J.P. Morgan believes this illustrative economics is conservative, reasoning that the actual useful life of GPUs is exceeding the 6-year depreciation period and contract term: V100s on the market have been in use for about 8 years, T4s about 7 years, and for the A100 specification launched in 2020, contracts have already been signed through 2029. This means that once asset-level debt is repaid, CoreWeave only needs to bear leasing and power costs to continue generating cash flow beyond the contract term"renewal monetization" thus becomes an additional option independent of the 5-year contract. Project Returns: Unlevered IRR in the "Mid-Teens," Above the 9%11% Blended Cost of Capital According to the illustrative calculations in the report, the unlevered IRR of a typical 5-year contract is in the "mid-teens" range, while the blended cost of capital is 9%11% (assuming the vast majority of capital expenditure is jointly financed by parent company debt and asset-backed debt). If the contract is renewed after the first 5-year term, the IRR would rise further. The report does not shy away from the risk of rising capital costs, but expects pricing for new computing capacity to embed higher funding costs, thereby keeping the company's return fundamentals unchanged. Burning Cash Is Not Poor Economics, but Too-Fast Ramp-Up Recent high funding needs stem mainly from a steep capacity ramp: operating capacity must rise from 1.85 GW at the end of this year to 8 GW in 2030, equivalent to quadrupling in 4 years and adding about 1.5 GW per yearby comparison, only 500 MW was added in 2025, and 1 GW is expected to be added in 2026. According to the simplified economics in the report: once a project reaches full capacity, annualized revenue per GW is $10 billion to $15 billion, corresponding to a contract-level cash EBITDA margin of about 75%; in ramp-up years, capital expenditure per GW is more than $3.5 billion. The crux is a timing mismatchrevenue is gradually released only after the project reaches full capacity, while capital expenditure is loaded upfront in the ramp-up year. On this basis, the company needs the "installed base of steadily operating projects" to reach 4 to 5 times the "capacity ramping up in the current year" for cash flow to turn positive after covering the capital expenditure of new capacity. The current situation is: the ratio of operating power at the beginning of the year to newly added power for the year is currently about 1:1CoreWeave entered 2026 with about 0.85 GW of operating power, and is expected to add about 1 GW for the full year; by 2027 this ratio is expected to remain around 1.2 times. The report calculates that it will not reach about 4.5 times until 2030. Therefore, J.P. Morgan's conclusion is: current cash burn is not the result of poor project economics, but of annual capacity ramp-up being too high relative to existing operating capacity; the timing issue of free cash flow generation will be largely resolved as CoreWeave scales up. From the other side of the market, this selloff does have a real credit-side feel. According to a citation by domestic media, CoreWeave bond yields have breached 15%, and rising capital costs will directly compress the safety margin of the aforementioned "residual cash flow"this is precisely the focus of the current bull-bear divergence. It should be noted that the report's disclosure page shows: J.P. Morgan has provided investment banking services to CoreWeave over the past 12 months and received compensation, and expects it may still seek related compensation over the next three months; it also holds a beneficial interest of more than 1% of CoreWeave's common stock and acts as a market maker for the relevant securities. Investors should treat this report as one reference factor in investment decisions.