Nvidia's quarterly filing on Wednesday identified indebtedness as a risk factor, with the company cautioning that rising obligations could "adversely affect" its financial condition and cash flows.

As of July 26, Nvidia had $33.5 billion in senior notes outstanding, and a $25 billion commercial paper program.

Earlier this year, Nvidia said it could raise up to $25 billion through issuance of unsecured commercial paper notes, in what turned into the chipmakers' first bond sale since the start of the AI boom.

The quarterly filing said $15 billion of debt is due in one to five years. In its last quarterly filing, Nvidia reported $2.75 billion of debt due in one to five years.

"Maintenance of our indebtedness, contractual restrictions, and additional issuances of indebtedness could cause us to dedicate a substantial portion of our cash flows from operations towards debt service obligations and principal repayments," the company said.

-- Lora Kolodny

In an interview with CNBC's Mad Money, Huang gave some more details on the company's agreement with AWS.

He said in addition to the 2 million GPUs that Amazon plans to buy, the cloud company can purchase "millions of CPUs," Huang said, referring to central processors. Nvidia is currently ramping up sales of its first ground-up CPU called Vera, competing with Intel and Advanced Micro Devices, the traditional CPU leaders.

Huang said Amazon will also use Nvidia's technology for its robotics.

— Kif Leswing

Nvidia has made some massive investments in leading AI labs, including the purchase of a $30 billion stake in OpenAI earlier this year.

Huang said the company should have been even more aggressive ahead of expected IPOs.

"Investing in these companies are a once-in-a-generation opportunity," Huang said on the earnings call. "The only regret that I have is that I didn't invest more and sooner."

Huang suggested that OpenAI and Anthropic "will likely go public soon, and others will follow."

— Kif Leswing

Nvidia said it expects gross margin to decline and bottom out in the fourth quarter of fiscal 2027, in the range of 71% to 72%, partially due to memory prices.

"We want to be direct about this, rather than let it linger as an open question," Kress said. "Memory scarcity today is being driven in large part by the AI buildout itself."

— Kif Leswing

Amazon and Nvidia announced that Amazon Web Services will buy 2 million Nvidia GPUs and use the company's new CPU called Vera.

Nvidia said some of AWS' Vera CPUs would be "integrated with Rubin," the company's forthcoming AI chip, and some would be standalone.

The deal shows that hyperscalers continue to spend heavily despite investor fears that they may be tapping out. Kress said capex among the "top five hyperscalers" is expected to increase to $1.3 trillion next year from $800 billion in 2026.

— Kif Leswing

Nvidia said it spent $26 billion on share repurchases and dividends during the quarter, after telling investors in May that it had authorized an additional $80 billion in buybacks. It also plans to pay a 25-cent per share dividend.

The company has been stepping up its capital return in recent years as revenue and free cash flow have expanded.

— Kif Leswing

Nvidia said that its supply commitments more than doubled from $119 billion last quarter to $279 billion in the latest period, "primarily related to the procurement of memory."

"We continue to make strategic commitments across our supply, infrastructure, and partner ecosystems to capitalize on the substantial growth opportunities ahead of us," Nvidia said in the CFO commentary.

— Ari Levy

Nvidia got a green light from the U.S. government late last year to sell some older chips to China, but sales there have been slow and it's unclear to what degree the Chinese government has approved the import.

CFO Colette Kress said in written commentary that shipments of those chips made up less than 1% of the company's data center revenue.

— Kif Leswing

Huang emphasized in the earnings release how much the landscape is changing. He said AI "reached its inflection point," noting that the number of companies who need large clusters of GPUs has expanded dramatically.

"This time last year, one lab alone was driving the buildout," Huang said. "Today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online — with strong momentum across the U.S. and around the world."

— Kif Leswing

For the second straight quarter, Nvidia's gross margin was 75%, a stunning number for a hardware company, but one that's under pressure as prices for components like memory and wafers rise.

Nvidia said its gross margin in the current quarter will slip to 74%.

— Kif Leswing

While Nvidia still counts on a handful of giant internet companies — hyperscalers — for an outsized portion of its revenue, the rest of the business is picking up steam.

Nvidia's AI Clouds, industrial, and enterprise (ACIE) customers accounted for $40.3 billion in sales in the quarter, up 138% on an annual basis. Hyperscaler revenue more than doubled to $48.7 billion.

— Kif Leswing

Nvidia's net income included a $7.8 billion gain on equity investments, following a gain of $15.9 billion in the first quarter.

The chipmaker has made big investments in companies including Intel and SpaceX. 

— Kif Leswing

Nvidia reported data center sales of $89 billion, surpassing expectations of $86.33 billion, according to StreetAccount. That's an annual increase of 117%.

The company now gets 92% of its sales from its data center unit, which includes revenue from its market-leading AI chips. Nvidia's total revenue for the quarter climbed 106% from a year earlier. 

— Kif Leswing

Nvidia said it sees sales of $108 billion in the current quarter, plus or minus 2%. Analysts were expecting guidance of $104.2 billion.

Nvidia said its outlook includes no data center sales from China.

— Kif Leswing

Nvidia has seen its stock retreat the day after reporting results in each of the previous four quarters — despite meeting or beating estimates for earnings per share, revenue and forward guidance. 

Over the past year, Nvidia has beaten estimates for all the top-line metrics – with the lone exception of the second quarter of last year, when it simply met guidance forecasts – and yet has still been punished by investors during the subsequent trading day, according to Bespoke Data.

Nvidia shares were down about 1% in the last hour of regular trading, dropping for the eighth time in nine days.

--Tobias Burns

One number that will be closely watched on Wednesday's report is non-hyperscaler sales.

Last quarter, half of Nvidia's sales went to a handful of internet giants. That was revealed as Nvidia started to report its data center revenue split into two buckets: hyperscalers, and AI clouds, industrial, and enterprise customers.

In the second-quarter report, investors will be looking to see if all other Nvidia data center chip buyers are growing revenue faster than its hyperscaler customers.

"The easiest go-to-market, of course, is the hyperscaler, because there are only five or six of them," Huang said on the May earnings call. "The rest of them, the rest of the industry, represents 250,000 companies around the world."

— Kif Leswing

The relationship between Huang and Elon Musk has been tightening of late.

Nvidia recently disclosed a $21 billion stake in SpaceX, stemming from its earlier $10 billion investment in xAI ahead of a merger between the two Musk companies. Those shares are now worth about $17 billion.

SpaceX's record IPO took place in June, and earlier this month the company reported earnings for the first time as a public company. On the call, Musk endorsed Nvidia's Vera Rubin platform, and said SpaceX would build its AI data centers exclusively using Nvidia chips.

On Monday, Musk and Nvidia said the companies were also making progress on developing orbital data center technology, known as the SpaceX Starmind AI satellite. In a post on X, which is owned by SpaceX, Musk said the companies had "designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028."

The companies haven't said how many chips may be needed by SpaceX to fulfill its goals, but Musk said on the call he expects to get a "significant" allocation of Vera Rubin chips next year.

— Lora Kolodny

Nvidia's customers are spending heavily on their AI buildouts, but it's not clear how long that can last.

Amazon and Alphabet both reported negative free cash flow for the second quarter, as did Elon Musk's public companies, Tesla and SpaceX. Meta, meanwhile, saw its cash generation dwindle by about 90%.

The hyperscalers are collectively spending hundreds of billions of dollars this year on AI infrastructure, and Goldman Sachs expect that number to reach $1.2 trillion in 2027. Those companies are raising historic amounts of capital, both equity and debt, while also partnering with financial firms to push some risk off their balance sheets.

They all say they're just trying to keep up with demand, but volatility has picked up in the markets this year partly due to concern that the industry is taking on too much risk.

— Ari Levy

Investors aren't ready for a slowdown.

Analysts polled by LSEG are expecting Nvidia to guide to earnings of $2.38 per share on $104.2 billion in revenue. The sales figure that would represent a jump of 83% from $57 billion a year earlier.

Some analysts expect even more from the chipmaker. The highest estimate for third-quarter revenue among analysts polled by LSEG is $112.2 billion.

— Kif Leswing

Memory prices around the world are soaring due to a shortage, partially caused by Nvidia, which needs high-performance HBM memory for its chips, and large quantities of lower-specification DRAM for its complete systems.

The price for DRAM for servers rose 64% in the second half of last year, and Trendforce expects it to jump 260% in 2026, according to an estimate published this month.

In the most recent quarter, Nvidia said higher device prices due to memory hurt demand for some consumer GPUs, but the company said it spent $145 billion in the first quarter to secure supply of the parts it needs.

"While we are not immune to supply challenges, we remain confident in our ability to support the growth opportunity ahead," CFO Colette Kress said.

— Kif Leswing

Nvidia is expected to report a healthy 75% gross margin, the same place landed in the first quarter.

But Nvidia has spent heavily ahead to build inventory for the wide release of its Vera Rubin AI systems, and lots of parts like memory have seen prices skyrocket.

The company could also share some information about widely reported price hikes for its AI chips.

— Kif Leswing

Huang announced a program earlier this month with six leading financial firms who have pledge up to $500 billion in financing from investors. Their thesis is that chips have become an investable asset.

Nvidia's prodigious cash flow and strong balance sheet is allowing it to offer residual value backstops to certain projects, and investors have wondered what that may mean for the chipmaker's finances.

Morgan Stanley initiated credit coverage of Nvidia on Monday, and said there's some strength to the company's claim that financing pricey AI chips makes sense. But the firm wrote that the long-term risk is still hard to quantify.

"Our model and relative value are supportive, but we think the tail remains too early-stage, opaque, and sizable to step in," Morgan Stanley credit analyst Lindsay Tyler wrote in the note.

— Kif Leswing