Nvidia's Growth Engine Faces Its Hardest Question Yet: Who Ultimately Pays for the AI Buildout?

Nvidia's record Q2 revenue and strong forecast couldn't lift shares, as investors worry about its $200B credit exposure and paused financing deals.

Investors got everything they wanted from Nvidia's latest earnings — record revenue, a blockbuster forecast, and a demand pipeline that seems to know no bounds. The stock fell 4.1% anyway.

That disconnect has become the defining feature of Nvidia's market narrative. The company posted $96.2 billion in second-quarter revenue for fiscal 2027, up 106% year over year, with the data center business surging 117% to $89 billion. It was the 13th consecutive record quarter. Management guided to $108 billion in third-quarter sales, plus or minus 2%, and signaled roughly 70% growth for fiscal 2028 — far ahead of the roughly 45% the market had been pricing in. None of it was enough to keep the shares from sliding to €187.78 on Friday, leaving the stock 7.3% below its 52-week high of €202.50.

The explanation for that paradox lies less in the numbers themselves than in the machinery Nvidia has built to generate them.

A Financing Model Under Scrutiny

The company has effectively become a banker to its own boom, using revenue-sharing agreements and direct financing commitments to help customers afford the chips they crave. The scale of that exposure is staggering. Nvidia has pledged up to $105 billion to back OpenAI's 20-year data center lease in Ohio. Morgan Stanley estimates the company's total credit exposure could approach $200 billion by the end of 2028, with roughly $170 billion of that sitting off-balance-sheet.

Against that backdrop, news that Nvidia has paused some of its revenue-sharing agreements under its AI cloud financing initiative has taken on outsized significance. The company confirmed it had stepped back from the program last week, even as a spokesperson insisted the core compute capacity business remains unchanged and continues to evolve.

The market is now asking a question that growth metrics alone cannot answer: Can Nvidia sustain this level of customer financing without compromising its own balance sheet? The pause in certain deals suggests the company itself may be reassessing the risk profile of its financial engineering — a development that could weigh more heavily on the stock in coming weeks than any revenue print.

The Bull Case: An Ecosystem Without Peer

For optimists, the operational fundamentals remain compelling. Supply commitments have ballooned from $119 billion to $279 billion in just three months — a clear signal that customers are still lining up. Amazon has ordered an additional 2 million GPUs for 2027 and 2028, and Nvidia's financing platform, backed by partners including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, is designed to mobilize more than $500 billion for AI infrastructure.

CFO Colette Kress has said that AI labs co-financed by Nvidia could contribute roughly a quarter of fiscal 2027 revenue. The planned $12.9 billion acquisition of Hugging Face would further cement the company's position across the AI stack, while the Vera Rubin architecture — which pairs Nvidia's own CPU with its GPUs — promises to accelerate data traffic significantly.

If the 70% growth target for 2028 proves realistic, the current valuation, with a price-to-earnings ratio below its five-year average, could look cheap in hindsight. The stock still trades 34% above its 52-week low and above all major moving averages, suggesting the medium-term uptrend remains intact despite the recent pullback.

The Bear Case: Margins, Competition and Credit

The risks, however, are mounting on multiple fronts. Most immediately, rising memory costs are squeezing profitability. Kress has guided that gross margins will fall to 71-72% over the next two quarters, down from 75%, and servers equipped with Nvidia AI chips could become more than 15% more expensive for systems shipping early next year.

Competition is also emerging from an unexpected direction: Nvidia's own customers. Google, Amazon, Microsoft, Meta and OpenAI are all developing custom AI chips, with OpenAI's "Jalapeno" processor reportedly outperforming Nvidia's offerings in tests. The political landscape adds another layer of uncertainty, with the US government considering expanded semiconductor tariffs that could eliminate existing exemptions for data centers — though no such measure has been finalized.

Then there is the credit question. The off-balance-sheet financing structure that fuels growth in good times could become a serious liability in a downturn. Mirae Asset Securities notes that the analytical focus has shifted from raw GPU demand to metrics like customer AI returns, margin stability and free cash flow. If customers such as Meta see AI-driven advertising clicks rise without corresponding returns on their massive data center investments, order cuts could follow — even with order books currently full.

The stock's elevated annualized 30-day volatility of 45% reflects precisely this uncertainty, suggesting the market is already pricing in a wide range of outcomes.

What Happens Next

The near-term path depends on whether Nvidia can deliver on its third-quarter guidance of $108 billion while managing the margin compression from memory costs. The Vera Rubin rollout this fall will serve as a key test of the company's ability to maintain its technological edge, while the trajectory of memory prices will shape server pricing into early 2027.

The broader question — whether the AI buildout can translate into sustainable customer profitability — will ultimately determine whether Nvidia's financing commitments look like prudent investment or overreach. For now, the growth story remains intact, but the market's reaction to this quarter suggests that investors are no longer willing to take the financial architecture behind that growth on faith alone.