Morgan Stanley Initiates Credit Coverage on NVIDIA; Contingent Liabilities Could Reach $200 Billion
Nashnova编辑部
Morgan Stanley has initiated credit coverage on Nvidia with a neutral rating and a hold recommendation; its core finding is that off-balance-sheet contingent liabilities may reach $170 billion, putting the AI chip leader's true credit exposure far beyond what traditional financials show.
Why is Morgan Stanley saying "don't jump in yet"?
Morgan Stanley initiated Nvidia credit coverage on August 24, concluding: fundamentals are solid, but too much remains opaque.
Analysts Lindsay A Tyler and Nishant Satyam estimate that including all contingent obligations, Nvidia's full-scope credit exposure could reach roughly $200 billion by end-2028.
About $170 billion of that sits off the balance sheet — This means → looking only at reported debt drastically understates Nvidia's actual credit risk.
The bank recommends staying on the sideline in both cash bonds and CDS — credit default swaps, essentially insurance contracts against bond default.
How did Nvidia's balance sheet become an "AI financing tool"?
In plain terms = Nvidia no longer just sells chips for cash. It deploys its own cash and credit as leverage, helping partners unlock far larger pools of outside capital.
Three mechanisms are at work: up to 25% residual-value support under a $500 billion-plus partnership framework; guarantees for a roughly 4.25 GW shell facility in Ohio; and revenue-sharing plus credit support for emerging cloud providers.
Morgan Stanley projects Nvidia's net cash position will reach about $520 billion by fiscal year-end 2028; even after subtracting the $170 billion in contingent items, net cash remains roughly $350 billion, with full-scope leverage at about 0.4×.
This reflects a deeper shift: in the AI compute race, a chip company's balance sheet has itself become a product.
Where does the $170 billion in off-balance-sheet exposure come from?
Category one — roughly $38 billion: lease liabilities, partner lease guarantees, and CoreWeave capacity arrangements. Nvidia has disclosed $32 billion in not-yet-commenced lease obligations spanning three to twenty years.
Category two — roughly $68 billion: residual-value support — commitments to cover the gap if chip assets financed by partners depreciate below loan balances — tied to the $500 billion-plus partnership framework. This means → if those chip assets lose value, Nvidia absorbs the shortfall.
Category three — roughly $64 billion: revenue-sharing and credit-support exposure targeting customers beyond the hyperscalers. The clearest precedent is Nvidia's $6.3 billion CoreWeave residual cloud-capacity purchase commitment announced in September 2025.
How do S&P and Moody's see this — and why do they disagree?
S&P has laid out a clear methodology: chip residual-value support can be treated as a debt adjustment based on the gap between amortized loan balances and depreciated chip values, provided the ultimate beneficiary is sub-investment-grade.
S&P expects Ohio-campus-related debt adjustments to rise from $4.2 billion in 2028 to roughly $37.7 billion by 2031.
Moody's flags about $125 billion in potential support commitments as a credit pressure, but has not published a specific adjustment methodology.
In plain terms = both agencies acknowledge the issue, but one has handed investors a ruler while the other has only pointed a direction — there is still no unified yardstick for measuring these off-balance-sheet risks.
Are current spreads cheap or expensive?
Nvidia's 5-year CDS is quoted at roughly 84–89 basis points, about 25 bps wider than similarly high-rated Google and Amazon.
Its 2036 bond spread is only about 10–20 bps tighter than lower-rated IBM and AT&T at comparable tenors — This means → the market is effectively pricing a Aa1/AA-rated issuer at BBB-level spreads.
Morgan Stanley sees current pricing as not expensive, but not cheap enough to warrant immediate entry.
The bank's trigger: CDS above 100 bps and cash-bond spreads reaching AT&T levels — only then would the risk premium be sufficiently attractive.
What to watch next?
The $500 billion-plus partnership framework remains at the letter-of-intent stage — not yet converted into binding contracts.
The revenue-sharing and credit-support framework has limited disclosure; support scale, pace, and ultimate rating-agency treatment are all unresolved.
This reflects the central tension in the Nvidia credit story: cash flow and fundamentals are exceptionally strong, but transparency on off-balance-sheet commitments is not keeping pace with their growth.
In plain terms = the ultimate scale and transparency of contingent liabilities is the single verification point that will determine whether Nvidia's credit spreads can reprice.
Content is for reference only, not financial advice.