Nvidia (NASDAQ: NVDA) is paying a price for its ambitious AI ecosystem financing strategy. Less than two months after unveiling its "AI Compute Partnership" initiative, the world's most valuable company has quietly suspended portions of the program, while the first-time disclosure of a staggering $530.5 billion in off-balance-sheet commitments and guarantees has credit market investors on edge.
According to reports from August 27, insiders revealed that Nvidia paused certain agreements under the revenue-sharing plan last week, partly due to internal employee concerns that the model could trigger antitrust scrutiny and questions about the chip giant's level of operational control over its customers.
Meanwhile, Morgan Stanley initiated credit coverage on Nvidia with a neutral rating, warning that the company is transforming its balance sheet into a strategic AI financing vehicle — dubbed "Balance-Sheet-as-a-Service" — thereby introducing new tail risks. Analysts believe these two developments occurring simultaneously have sharply intensified Wall Street's unease over Nvidia's "circular financing" model. Nvidia's credit default swap (CDS) quotes have climbed near historical highs, settling at 83 basis points — more than double the historical average.
Morgan Stanley credit analyst Lindsay Taylor stated bluntly that Nvidia's over $500 billion in partnerships remain at the letter-of-intent stage, with limited disclosure on revenue-sharing and credit support frameworks, making the tail risk "too early-stage, too opaque, and too massive."
Behind the Halt: Antitrust Concerns and Control Disputes
Nvidia launched the "AI Compute Partnership" in July, designed to solve financing challenges for small and mid-sized cloud service providers. Building AI cloud infrastructure requires billions in upfront spending on Nvidia GPUs and data centers, yet these providers often need financing before securing sufficient customer contracts. Nvidia's solution: commit to leasing GPU capacity from these providers when they cannot find other customers, thereby guaranteeing revenue streams that help secure construction loans.
The plan also creates dual profit paths for Nvidia: chip sales plus a share of revenue generated from customers' leased-out compute. Insiders noted that both parties set a baseline hourly rate covering the provider's costs — including chip depreciation, data center operations, and staffing — above which Nvidia captures 50% of excess revenue. Initial participants included Sharon AI and Firmus Technologies.
However, friction emerged within weeks. Nvidia required cloud providers to lease chips only to customers approved by Nvidia, favoring distribution across multiple small AI firms rather than concentrating on a single large tenant. Some providers strongly resisted, viewing customer selection autonomy as a fundamental business right. Nvidia employees subsequently expressed antitrust concerns to existing and prospective customers about the sensitivity of the chip giant's intervention in customer operational decisions.
Nvidia formally suspended portions of the plan last week, less than two months after its announcement. A spokesperson stated the July-launched business model "remains valid and continues to evolve due to strong demand," without denying the pause. Insiders indicated the program may be redesigned or folded into other initiatives.
Off-Balance-Sheet Commitments First Revealed: $530.5 Billion Stuns the Market
In the same week as the halt news, Nvidia disclosed in its latest quarterly filing the specific scale of off-balance-sheet commitments — a figure that caught the market's attention. Commitments span multiple layers: purchase commitments for storage capacity from suppliers jumped from $119 billion to $279 billion within three months; land, power, and data center capacity support commitments to emerging cloud providers like Firmus and Iren reached $56 billion; and loan guarantees to other data center developers totaled $108.5 billion. Combined, Nvidia's disclosed off-balance-sheet forward commitments and guarantees amount to a staggering $530.5 billion, extending beyond 2032.
CFO Colette Kress addressed the circular financing criticism directly on the earnings call: "We recognize the scale of this support, and we know some will call it circular financing. We see it differently. We are going through a major computing platform transition, one of the most important technologies in human history, and these are once-in-a-generation opportunities." She also projected this new revenue stream could generate billions in gains over the medium to long term.
Yet the financials reveal another layer of concern: second-quarter net income reached $56 billion, while operating cash flow was just $24 billion — a $36 billion gap. This discrepancy stems primarily from a surge in accounts receivable, which ballooned to $63.1 billion by period-end from $38.5 billion at the end of January. Nvidia acknowledged in its filing that "financing arrangements with certain investment-grade customers, including extended payment terms under large multi-quarter agreements, will continue to impact the timing of operating cash flow." Notably, just five companies — speculated to be Microsoft, Amazon, Alphabet, Meta, and Oracle — account for 70% of that receivables balance, underscoring significant concentration risk.
Morgan Stanley Dissects Tail Risk: Three Exposures Broken Down
Morgan Stanley's inaugural credit report on Nvidia constructs an analytical framework beyond traditional leverage metrics, categorizing Nvidia's credit exposure into three buckets and projecting total credit exposure of approximately $200 billion by end-2028, with roughly $170 billion from adjusted items and contingent obligations.
First: lease liabilities and guarantee exposures. Nvidia has disclosed $32.4 billion in undischarged lease obligations with terms of 3 to 20 years, expected to commence between fiscal Q2 2027 and fiscal 2033. Morgan Stanley estimates these commitments could ultimately add approximately $20.8 billion in lease liabilities. Additionally, Nvidia disclosed $3.5 billion in maximum exposure under partner facility lease guarantees and residual value guarantees (RVG) covering approximately 4.25 gigawatts of IT load at Ohio's PORTS-Pike technology park. S&P projects debt adjustments will rise from $4.2 billion in 2028 to roughly $37.7 billion by 2031-2032 as related buildings come online.
Second: residual value support under the $500 billion partnership program. In August, Nvidia announced collaborations with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize over $500 billion in third-party capital for AI infrastructure. In certain projects, Nvidia may provide up to 25% residual value support. Morgan Stanley built a multi-period, scenario-driven model extending this framework across 15 financing tranches of approximately $50 billion each, estimating peak contingent obligations near $90 billion, occurring shortly after end-2028. The bank noted the $500 billion program remains at letter-of-intent stage with high uncertainty around scale, pacing, and guarantee structures, and future disclosure may be limited as exposure moves through special purpose vehicles (SPVs) and private structures.
Third: revenue sharing and unsold compute backstop arrangements. Morgan Stanley views this category as potentially the most consequential and hardest to track. Modeling 5 gigawatts of supported compute, the bank estimates contingent obligations of approximately $81 billion (approximately $64 billion after tax) by end-2028. The most direct precedent is Nvidia's $6.3 billion commitment to CoreWeave — purchasing remaining cloud compute capacity if CoreWeave cannot find other customers. S&P has already incorporated such unsold capacity backstops into debt adjustments, treating them similarly to lease guarantees.
Credit Market Divergence: Solid Fundamentals but Underpriced Tail Risk
Morgan Stanley's analysis reveals an inherent tension: fundamental data remains robust, yet tail risk opacity prevents a buy recommendation. On fundamentals, even incorporating all $170 billion in adjustments and contingent obligations, Nvidia's balance sheet remains exceptionally strong. Morgan Stanley projects net cash — after deducting balance sheet debt — of approximately $520 billion by fiscal 2029/calendar 2028, or roughly $350 billion on a fully adjusted basis. All-in leverage sits at only about 0.4x, with free cash flow after shareholder returns exceeding 100% of all-in debt. Even if EBITDA and free cash flow stagnate at fiscal 2027 levels, these metrics would remain at reasonable levels of approximately 0.7x and 15% in fiscal 2029.
However, credit market pricing has shifted. Nvidia's 5-year CDS quotes around 84-89 basis points, roughly 25 basis points wider than same-rated Alphabet and Amazon. Nvidia's 2036 bonds trade only 10-20 basis points tighter than lower-rated IBM and AT&T paper. Nick Ferres, Chief Investment Officer at Vantage Point Asset Management, noted that how one interprets Nvidia's results depends on whether one is an equity or credit investor — for the latter, the additional risk layer from circular financing arrangements remains ever-present.
Morgan Stanley advises investors to stay on the sidelines for both cash bonds and CDS, suggesting they would only consider entry if CDS breaks above 100 basis points and cash bond spreads approach AT&T levels. The bank also warned that CDS may underperform cash bonds due to lender hedging demand and thematic positioning.
Nvidia's Defense: Is It Truly Different This Time?
Facing comparisons to fiber-optic giants of the internet bubble era who "lent money to customers to buy their own products," Nvidia management is fighting back. Colette Kress cited hyperscaler capital expenditure forecasts in defense: projected capex for the top five hyperscalers approaching $800 billion in 2026 and $1.3 trillion in 2027. CEO Jensen Huang argued for growth sustainability from the demand side, emphasizing AI's shift from human-driven to AI-agent-driven: "An agent requires 15 to 100 times more compute than human usage, and the scale of compute required is extraordinary."
He also unusually pre-announced fiscal 2028 full-year guidance, projecting revenue growth of 70% — far exceeding the market's 45% consensus — which finally drove a stock rebound after five consecutive quarters of "beating earnings yet declining." However, financial blog ZeroHedge noted that "this time is different" has historically been the four most dangerous words in finance. Whether Nvidia's circular financing logic holds ultimately depends on whether AI demand materializes along management's projected trajectory.
Morgan Stanley's conclusion perhaps best captures current credit market sentiment: Nvidia's fundamentals are impeccable, but until over $1 trillion in GPU and XPU-related ecosystem financing is deployed through innovative structures, whether this balance sheet is a moat or a liability remains to be tested by time.
Comments