NVIDIA Earnings Report: Three Previously Undisclosed Insights

Following the release of NVIDIA's earnings report last night, we thoroughly reviewed both the official announcement and the earnings call. Upon analysis, I identified three key points worth discussing, all of which represent information NVIDIA has not previously disclosed in its earnings reports or conference calls.

First, NVIDIA made an exception by providing guidance for the next fiscal year. Historically, NVIDIA has never issued forecasts a full year in advance. This time, however, it explicitly stated that revenue for fiscal year 2028 (corresponding to calendar year 2027) is expected to grow by approximately 70%. It further clarified that this figure reflects supply-constrained capacity, while actual customer demand points toward a doubling of revenue. Jensen Huang explained the rationale behind breaking this precedent. Given that building AI infrastructure requires locking in land, power, and supply chains two to three years in advance, he needs customers, suppliers, and shareholders to plan based on a unified set of expectations. Therefore, he preferred to provide comprehensive forward-looking guidance early.

Second, management proactively addressed the anticipated decline in gross margins. The gross margin guidance for the third quarter is 74%, expected to bottom out at 71%–72% in the fourth quarter, before recovering to 72%–73% next year, primarily due to high memory costs. Management stated verbatim: "The extreme situation in memory prices has seen increases exceeding our previous expectations, and prices will rise even further next year." Coming from the world's largest memory buyer, this statement likely carries more weight for the memory industry than for NVIDIA itself. The company also clarified its response strategy: rather than pressuring suppliers, it will initially absorb a 3–4 percentage point loss in gross margin, followed by price increases for customers, effective from the first quarter of next year.

Third, it directly addressed concerns regarding "circular financing" by disclosing exposure figures. To date, NVIDIA has invested nearly $50 billion in frontier AI labs, while also collaborating with institutions such as Apollo, BlackRock, and Blackstone to establish a third-party financing platform with a scale of $500 billion. Additionally, it has begun providing minimum revenue guarantees to emerging cloud providers in exchange for a share of rents above these guarantees. As management put it, they are "getting paid twice: once for selling hardware, and once for sharing rental income." The CFO responded, "We are aware that some may label this as circular financing, but we do not view it that way," and subsequently disclosed that laboratory demand supported by NVIDIA's balance sheet accounts for approximately one-quarter of next year's business.

Looking back at the financial report itself, let's briefly review the figures: single-quarter revenue reached $96.2 billion, doubling year-over-year, with accelerating growth for the fourth consecutive quarter; data center revenue hit $89 billion, up 117% year-over-year; gross margin stood at 75%; and $26 billion was returned to shareholders in the quarter. However, the stock price dipped slightly on the first trading day after the earnings release. The market has come to take such strong performance for granted.

During the earnings call, analysts from Morgan Stanley, Bernstein, and Goldman Sachs all focused on supply-side issues: where does the confidence for 70% growth come from, how large is the unconstrained demand really, and where are the bottlenecks? Bank of America asked about the accounting treatment of the $500 billion ecosystem commitment and the impact of OpenAI’s in-house chip development. Jensen Huang responded that self-developed chips are specialized inference hardware for single-cloud, single-service use cases, whereas NVIDIA sells a complete AI factory platform. He also hinted that two portfolio companies are nearing IPOs, remarking that his only regret was investing too little and too late. Other firms asked about the economics per gigawatt, inference market share, open-source models, and the timeline for AGI. Across questions from eight firms, not one questioned whether AI demand is real or fictitious.

Finally, let’s highlight some key aspects that are easily overlooked in the financial report:

First, GAAP EPS of $2.46 was actually higher than non-GAAP EPS of $2.22, reversing the usual trend. The difference stemmed almost entirely from unrealized gains in the investment portfolio, amounting to $7.8 billion in the quarter and $23.7 billion in the first half of the year. NVIDIA holds approximately $94 billion in equity investments; each upward revaluation of these invested AI companies flows into GAAP profits. Therefore, to assess core operational performance, one should focus on the non-GAAP figure of $2.22.

Second, accounts receivable increased by $22.3 billion in the quarter to reach $63 billion, extending days sales outstanding (DSO) to 60 days. Operating cash flow for the quarter was $24.1 billion, significantly lower than the net income of $59.7 billion. The company explained that multi-quarter orders from investment-grade major customers have lengthened payment terms.

Third, NVIDIA issued debt on a large scale for the first time, raising $24.9 billion in the quarter, which increased long-term debt from $7.5 billion to $32.4 billion. The company is simultaneously borrowing, investing in its ecosystem, and conducting share buybacks, actively leveraging its balance sheet.

Looking ahead to the next few quarters, this earnings report leaves three key metrics to monitor for validation. First is gross margin: management indicated it bottomed out in Q4 and will be restored through price increases in Q1 of next year. Whether this recovery materializes will directly test the company's pricing power with customers. Second is the gap between accounts receivable and operating cash flow; if this widens further, it suggests the company is still relaxing payment terms to secure orders. Third are the minimum revenue guarantees, currently off-balance-sheet commitments. If emerging cloud vendors fail to sustain their business, these commitments will migrate onto the balance sheet. Monitoring these three points is more useful than waiting for the next revenue record. $NVIDIA (NVDA.US)$$SK hynix (SKHY.US)$$CoreWeave (CRWV.US)$

77K Views