Nvidia Reports $96.2 Billion Quarter, Guides to 70% Revenue Growth for Fiscal 2028 as Memory Costs Squeeze Margins

Nvidia posted fiscal second-quarter revenue of $96.22 billion on Wednesday, beating the roughly $92.2-92.4 billion Wall Street expected, according to figures reported by ZeroHedge and confirmed across outlets including Reuters and CNBC. Adjusted earnings came in at $2.22 a share versus the $2.09-2.10 analysts projected.

The stock initially dipped in after-hours trading on margin worries — down 1.8% according to Euronews, or "over 1%" per Reuters, and 2% per Business Insider — then reversed to close up more than 5% once CFO Colette Kress delivered fiscal 2028 guidance during the earnings call, according to ZeroHedge, Euronews and Reuters. By premarket trading Thursday, shares were up 7.2%, CNBC reported, dragging chip peers Micron, Marvell, Arm, Intel and AMD higher along with it.

The number that moved the stock: Kress told analysts Nvidia expects revenue to grow roughly 70% in fiscal 2028, the period running from February 2027 to January 2028. Reuters reported that analysts had, on average, projected just 44% growth for that same period heading into the print. Euronews and FXStreet pegged the consensus at 45%. Fiscal 2028 hasn't started yet.

"AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue, and demand is accelerating," Huang said in a statement carried by Euronews and ZeroHedge.

Kress went further on the call, telling analysts Nvidia is "seeing demand acceleration even at our scale" and that customer forecasts point to growth doubling next year, according to ZeroHedge and Reuters. Huang added that demand is "much greater than 70%," per CNBC and Yahoo Finance, but the company is supply-constrained. It can currently fill only around 70% of what customers want, per FXStreet and Yahoo Finance.

The Margin Problem Is Real

Not everything in the report was good news, and Nvidia said so itself. Gross margin, which came in at 75% for the quarter, is expected to fall to 74% in the third quarter and then to 71-72% in the fourth quarter, according to thelec, Business Insider, Reuters and FXStreet, which cited Nvidia's own Aug. 26 earnings call. The company blamed rising memory-chip prices, saying the magnitude of the increases "exceeded its previous expectations" and that prices are headed higher still next year.

Nvidia is in talks with Samsung Electronics, SK hynix and Micron Technology to expand memory capacity, thelec reported. Margin recovery to 72-73% isn't expected until fiscal 2028.

Free cash flow told a similar story. FXStreet reported free cash flow of $21.3 billion, down 56% quarter-on-quarter, though still up 59% from a year earlier. Business Insider quoted Kress calling memory costs "extreme" and saying the company was "resetting expectations today."

The Concentration Question

Nvidia's growth is still tied to a small number of giant customers, and those customers are actively building alternatives to Nvidia's chips. Euronews and thelec reported that Amazon, Google and Microsoft — Nvidia's biggest buyers — along with Anthropic and OpenAI, are each designing their own AI chips (ASICs) to reduce dependence on the company. OpenAI said this week its in-house Jalapeno processor outperformed Nvidia's current lineup in testing, though Euronews noted that comparison excluded Nvidia's newer Vera Rubin chips, which have begun shipping to customers.

Custom silicon from hyperscalers is a real competitive threat to Nvidia's near-monopoly on advanced AI chips, and CNBC reported that analysts flagged exactly this risk on Thursday. Nvidia's own response, relayed by thelec, is that these custom ASICs are narrower inference chips built for single clouds, while Nvidia's platform spans training and inference across every customer. Huang, per Business Insider, put it more bluntly on the call: "We're building something very different... Nvidia is a platform, an entire AI factory platform that spans the entire AI lifecycle that you can use in any cloud." That's a distinction, not a denial.

Nvidia did point to diversification: its AI Clouds, Industrial and Enterprise (ACIE) segment brought in $40.3 billion, up 138% year-over-year, according to CNBC and Yahoo Finance. That's real growth outside the core hyperscaler relationship. But data center revenue overall — $89 billion — still made up more than 92% of total revenue, per FXStreet, so "diversification" here means diversifying within AI infrastructure, not away from it.

Circular Financing, Again

Kress also pushed back directly on accusations that Nvidia's investment deals with customers amount to circular financing. Nvidia is funding some of the companies that then buy its chips. "We recognize the scale of this support, and we know some will call this circular financing. We see it differently," she said, according to Business Insider.

Thelec reported Nvidia telling analysts that OpenAI "will grow into the largest technology company in history," which the company offered as its answer to concerns about recovering the money.

Nvidia's outlook assumes zero China data-center compute revenue, according to Yahoo Finance and Euronews, reflecting ongoing geopolitical restrictions. China sales, including gaming and non-AI hardware, totaled $7.88 billion for the period, nearly double a year earlier, per Euronews, though H200 sales to Chinese customers made up less than 1% of data center revenue. Whether that changes, and whether the 70% guidance survives contact with fiscal 2028 once it actually begins in February 2027, are the two questions that will define whether Wednesday's rally holds.