Goldman Sachs Upgrades CoreWeave to Buy; JPMorgan Upgrades Intel Rating
nashnova research
Goldman Sachs upgraded AI cloud-compute firm CoreWeave from neutral to buy with a $106 price target — roughly 73% upside from Friday's close — while JPMorgan lifted Intel to neutral the same morning, sending both stocks higher pre-market.
CoreWeave is down over 40% — why is Goldman calling a buy now?
CoreWeave shares have fallen 43% from their post-Q2-earnings peak; the last five trading days alone saw a 24% drop, with Friday's close at $61.39.
Goldman analysts Gabriela Borges and Maura Hager raised the rating from neutral to buy but cut the price target from $139 to $106.
This means → Goldman thinks the sell-off overshot, but also concedes its earlier expectation was too high. The buy call rests on a target already marked down by nearly a quarter.
How expensive is CoreWeave's funding — and what are bulls and bears really arguing about?
The bear case centers on CoreWeave's unsecured bond yields at 13.5–15%, credit-default swaps (CDS — essentially default-risk insurance) at all-time highs, and Deutsche Bank's estimate of a roughly $35 billion funding gap over the next 12–18 months.
Goldman's rebuttal: CoreWeave funds itself mainly through DDTLs — delayed-draw term loans backed by GPUs and customer contracts. The unsecured bond market has been closed to the company for months, so quoting that market's yield overstates the actual cost of borrowing.
Goldman estimates CoreWeave's real weighted-average debt cost at roughly 800–900 basis points, with new funding possibly 100–200 bps higher but "unlikely to exceed that range." DDTL tranches 5.0 and 5.5 still trading near or above par serve as the most direct market evidence.
"Revenue per megawatt" — what does this key metric actually mean?
In plain terms = CoreWeave builds data centers, fills them with GPUs, and rents out compute power. "Annual recurring revenue per megawatt" is how much money each megawatt of electrical capacity generates in a year.
The current run-rate is about $10 million per MW, but short-term contracts are already signing at $40 million per MW. The company raised prices 25% in July and another 10% in October; Q2-vintage contracts carry contribution margins 5–10 percentage points above recent ones.
Goldman sees $13 million per MW as a reasonable baseline for new long-term contracts — this single number is the foundation of the entire buy thesis.
What does Goldman's return model actually show?
Plugging in a five-year, 200 MW contract with 80% debt financing, a 10% cost of funds, and zero GPU residual value, the model yields an unlevered IRR of roughly 15%.
CoreWeave's weighted-average cost of capital (WACC) — its blended cost of money — sits at about 11.8%, leaving a positive spread of roughly 3 percentage points.
This means → every dollar deployed earns back faster than the cost of borrowing it by 3 points. Goldman adds that if GPUs retain any residual value at lease-end, unlevered IRR could rise to 20–25%.
Goldman also notes that new capacity as a share of the installed base has peaked, so the drag from idle, revenue-free capacity should shrink — the most capital-intensive phase may already be behind.
Why did JPMorgan upgrade Intel on the same day?
JPMorgan analyst Harlan Sur raised Intel from underweight to neutral; Intel rose nearly 4% pre-market.
Sur wrote: "Although the turnaround will take years, management execution has been steady and our confidence is growing."
This reflects a shift on Wall Street from actively betting against Intel to standing aside — not a buy call, but no longer a sell call.
What do these two reports ultimately need to prove?
CoreWeave's buy thesis stands or falls on two variables: whether revenue per megawatt can climb from the current ~$10 million to Goldman's assumed $13 million, and whether asset-backed funding costs can stay below the levels implied by the unsecured bond market.
In plain terms = one question is "can the rent go up," the other is "can borrowing stay cheap." If both hold, Goldman's logic works; if either breaks, downside risk remains significant.
市场有风险，内容仅供研究参考，不构成投资建议。