Nvidia is transitioning from the world's leading AI chip supplier into the credit cornerstone of the entire AI supply chain. Its latest earnings report not only showcased remarkable growth momentum again but also revealed a massive $530.5 billion off-balance-sheet commitment system—financing guarantees, lease endorsements, and revenue-sharing agreements are systematically transforming Nvidia's balance sheet into a funding engine for AI infrastructure.

The earnings figures themselves are indisputable: for the fiscal second quarter ending July 26, 2026, NVIDIA reported revenue of $96.2 billion, up 106% year-over-year, with data center revenue reaching $89 billion, both exceeding analyst expectations. The third-quarter revenue guidance of $108 billion signals the potential start of the $100-billion-per-quarter era. During the earnings call, CEO Jensen Huang unusually provided a 70% revenue growth guidance for fiscal 2028, far exceeding the market consensus of approximately 45%, which single-handedly reversed the post-earnings after-hours decline.

However, upon delving into the earnings details, credit market investors have a distinctly different focus compared to equity investors.

Nvidia disclosed total off-balance-sheet forward commitments and guarantees of up to $530.5 billion, extending beyond 2032. Meanwhile, a gap of over $36 billion has emerged between net income and operating cash flow, with accounts receivable increasing by more than $22 billion during the quarter. Morgan Stanley has initiated credit coverage on NVIDIA, assigning a neutral rating and advising investors to temporarily hold off in both the cash bond and credit default swap (CDS) markets. Nvidia's 5-year CDS is quoted at approximately 84 to 89 basis points, near historical highs and about 25 basis points wider than similarly rated Google and Amazon.

Earnings Performance: Data Center Engine Remains Powerful

Nvidia's second-quarter data center revenue reached $89 billion, up 117% year-over-year, accounting for over 90% of total revenue and exceeding analyst expectations of approximately $85.8 billion. Hyperscale customer revenue hit $48.7 billion, significantly surpassing the anticipated $43.6 billion, indicating that major cloud providers' capital expenditure expansion momentum remains robust.

On a non-GAAP basis, adjusted earnings per share for the second quarter were $2.22, up 120% year-over-year and above the market expectation of $2.09. Non-GAAP operating margin was approximately 66%, with gross margin holding at 75%. CFO Colette Kress stated during the earnings call that the five major hyperscale cloud providers are expected to invest close to $800 billion in capital expenditures in 2026, rising to $1.3 trillion by 2027.

Third-quarter revenue guidance of $108 billion represents approximately 90% year-over-year growth, higher than the analyst consensus of around $105 billion. Notably, this guidance explicitly excludes contributions from data center computing business in China, making the guidance more conservative and implying that potential recovery revenue from the Chinese market cannot yet be counted as certain growth. Third-quarter non-GAAP gross margin guidance of 74% represents a decline of about 1 percentage point from the second quarter, raising concerns among some investors about a potential peak in gross margins.

The Vera Rubin platform has entered full-scale production, with racks being deployed at partners including CoreWeave, Google Cloud, and Microsoft Azure. Huang emphasized in the earnings announcement: "Now, compute is revenue." He also pointed out that AI's shift from being driven by human instructions to being driven by AI agents will increase compute demand by 15 to 100 times, creating a snowball effect for Nvidia's products.

Off-Balance-Sheet $530.5 Billion: How Nvidia Became the 'AI Central Bank'

Beyond the impressive earnings figures, NVIDIA disclosed for the first time in this quarter's report the full scale of its off-balance-sheet commitments, which has put credit market investors on high alert.

According to calculations, Nvidia's current off-balance-sheet commitments span multiple layers: memory capacity purchase commitments to suppliers surged from $119 billion three months ago to $279 billion; support commitments for land, electricity, and data center shells provided to emerging cloud service providers such as Firmus and Iren amount to approximately $56 billion; and loan guarantees extended to other data center developers total roughly $108.5 billion. Combined, these commitments reach as high as $530.5 billion, extending beyond 2032.

Nvidia has not concealed these arrangements. The company explicitly stated in its filing: "Financing arrangements with certain investment-grade customers, including extended payment terms under large multi-quarter agreements, will continue to affect the timing of operating cash flows." Meanwhile, Kress directly addressed the market's "circular financing" concerns during the call: "We are fully aware of the scale of this support, and we understand that some may call it circular financing. But we see it differently. We are experiencing a major computing platform shift—the birth of one of the most important technologies in human history—and these companies are once-in-a-generation enterprises."

This defense has reminded some market participants of the "circular financing" model during the internet bubble era, when fiber-optic companies provided loans to customers to purchase more cable. Nvidia's response is that this time is different—demand will only continue to grow.

Cash Flow Gap and Receivables: Near-Term Financial Risk Signals

In addition to off-balance-sheet commitments, Nvidia's current financial statements also reveal noteworthy signals.

Second-quarter GAAP net income was $56 billion, while operating cash flow was $24 billion, creating a gap of approximately $36 billion. A year ago, the absolute size of this gap was much smaller. The primary driver of the widening gap is the rapid expansion of accounts receivable: as of the end of the second quarter, accounts receivable reached $63.1 billion, up significantly from $38.5 billion at the end of January, increasing by more than $22 billion during the quarter.

Nvidia disclosed in its filing that just five companies (widely believed to be Microsoft, Amazon, Alphabet, Meta, and Oracle) account for 70% of the accounts receivable balance. While concentration among the highest-credit-quality hyperscale cloud providers reduces default risk, it also highlights the customer concentration risk of Nvidia's business.

Nick Ferres, Chief Investment Officer at Vantage Point Asset Management, believes that how one interprets Nvidia's earnings depends on the investor's perspective: for equity investors, the certainty that the AI capital expenditure boom will continue through 2028 has been reinforced; but for credit investors, the additional layer of risk brought by circular financing arrangements remains an ever-present concern.

In the second quarter, Nvidia's free cash flow was $21.3 billion, down notably from $48.6 billion in the first quarter. The company returned approximately $26 billion to shareholders during the quarter, including roughly $19.7 billion in buybacks, with approximately $99 billion remaining in buyback authorization at quarter-end.

Morgan Stanley Credit Coverage: Solid Fundamentals, but Premature for Tail Risks

Morgan Stanley initiated credit coverage on NVIDIA on August 24, 2026, with a neutral rating, advising investors to remain on the sidelines in both the cash bond and CDS markets. Analyst Lindsay Taylor and strategist Nishant Satyam noted in their report that Nvidia's credit fundamentals remain solid, but "tail risks are still in their early stages, opaque, and substantial in scale—it is too early to get involved now."

Morgan Stanley categorized Nvidia's contingent obligations into three main categories, totaling approximately $170 billion in adjustments. When combined with reported balance sheet debt, total credit exposure could reach approximately $200 billion by the end of 2028.

The first category, approximately $38 billion, encompasses lease liabilities (including about $32 billion in commitments not yet commenced), partner lease guarantees, residual value guarantees (RVG) for the Ohio PORTS-Pike project, and CoreWeave capacity arrangements.

The second category, approximately $68 billion, relates to residual value support (RVS) under the framework of partnerships exceeding $500 billion. Nvidia has indicated it can provide up to 25% RVS in specific projects, and Morgan Stanley estimates the after-tax contingent obligation peak from RVS could reach approximately $90 billion by the end of 2028.

The third category, approximately $64 billion, stems from the "revenue sharing and credit support model" proposed by NVIDIA in July 2026. Morgan Stanley modeled 5 gigawatts of supported capacity and estimated after-tax contingent obligations of approximately $64 billion by the end of 2028.

However, even including all adjustments, Morgan Stanley's model shows NVIDIA remains in a deeply net cash position: approximately $350 billion in net cash by FY29/CY28, with total leverage of approximately 0.4 times on a fully loaded basis, and free cash flow after shareholder returns exceeding 100% of total fully loaded debt. Even in a stress scenario where growth plateaus, leverage would still be around 0.7 times, with free cash flow debt service coverage of approximately 15%.

Rating Agency Positions: Divergent Approaches to Contingent Obligation Debt Treatment

S&P and Moody's have both commented on Nvidia's contingent obligations, but their approaches differ, adding complexity to credit analysis.

S&P stated in its August 18, 2026 rating commentary on NVIDIA that chip RVS could form debt adjustments based on the gap between amortized loan balances and depreciated chip values, provided the ultimate beneficiary is a non-investment-grade entity, with adjustments processed on an after-tax basis at the 21% U.S. corporate tax rate. S&P also indicated that debt adjustments related to the PORTS-Pike RVG would rise from approximately $4.2 billion in fiscal 2028-2029 to approximately $37.7 billion in fiscal 2031-2032, before gradually declining as lease payments progress. Third-party lease guarantees and support for "counterparties failing to purchase cloud capacity" could also be included in debt-like adjustments.

Moody's, in its commentary on the same day, noted that approximately $125 billion in potential support commitments could pressure the credit profile but did not publish a specific debt adjustment methodology, listing "maintaining exceptional financial strength after guarantee exposure" as a clear prerequisite for a further upgrade.

From a relative value perspective, NVIDIA's 5-year CDS trades at approximately 84 to 89 basis points, about 25 basis points wider than similarly highly rated Google and Amazon. Nvidia's 2036 maturity bonds trade only about 10 to 20 basis points tighter than same-maturity bonds from lower-rated IBM and AT&T. Morgan Stanley believes current spread pricing is closer to BBB-level rather than AA-level; absolute levels are not expensive, but given that over $500 billion in partnerships remain at the letter-of-intent stage and information transparency may remain limited as financing moves through SPVs and private structures, the firm suggests that intervention would become more attractive if CDS exceeds 100 basis points or cash bonds reach AT&T levels.