Nvidia made $58 billion in a quarter: Here’s what is really driving profits

Nvidia (NVDA) has now beaten Wall Street in something like 18 of its last 20 quarters, and for three reports running, the stock has been sold off on reaction anyway. The first quarter of fiscal year 2027 was never going to break the streak of beats: Revenue landed near $81.6 billion, up 85% YoY, and net income hit a record $58.3 billion, an absurd 211% increase YoY. 

However, the number that should make investors pause is not whether Nvidia beat; it is where the profit came from. Net income for the quarter actually exceeded operating income, a quirk that does not happen when a company is simply selling more product. Something else is doing the heavy lifting behind the doors of Nvidia, and it is worth understanding before the next blowout earnings headline rolls past.

The profit that wasn’t chip sales

Operating income, the money Nvidia made selling graphics processors and networking gear, came in around $53.3 billion, a 147% YoY gain. Yet, net income came in even higher, at $58.3 billion. The gap is almost a single line: Other income of roughly $15.9 billion, compared to a small loss in Q1 last year. Nearly all of that was unrealized gains on Nvidia’s investment portfolio, about $13.4 billion on publicly-held stocks and another $2.6 billion on private holdings. These are paper marks, not cash from operations, and they can reverse just as quickly in a down quarter. On a non-GAAP basis, which strips out non-operational gains, earnings per share (EPS) came in near $1.84 versus a consensus of $1.76. It’s still a beat, but not nearly as theatrical a one as the 211% net income figure implies. The generally accepted accounting principles (GAAP) profit, the one swelling Nvidia’s headline figures, is being flattered by Nvidia’s own bets on the very AI boom that is bolstering Nvidia into its phonebook-number valuation.

Funding the customers that fund the boom

Those bets are the second story. Through the quarter, Nvidia plowed around $18.6 billion into private companies and infrastructure funds, up from a comparative rounding error of $649 million a year earlier. The filing is candid that some of these are AI model makers that may end up buying or using Nvidia products in the cloud. It also flags that one AI company drove a meaningful slice of revenue by purchasing cloud services from Nvidia's own customers. Read that chain slowly: Nvidia invests in AI firms, the rising value of those firms shows up as a gain in Nvidia's net income, and those same firms loop back as demand for Nvidia chips. Total investment commitments now stand at nearly $27 billion. 

None of this is illegal or even unusual for a company holding this much cash, but it muddies the question every AI skeptic keeps asking: How much of this demand is organic, and how much is Nvidia quietly financing its own order book?

Three customers, over half the revenue

Concentration continues to creep in the wrong direction. Three direct customers accounted for 21%, 17%, and 16% of total revenue this quarter; put another way, just over half of Nvidia’s entire business income came from just three buyers. A year ago, it was two customers at 16% and 14%. On the balance sheet it is starker, with three names making up close to two-thirds of receivables. When a handful of hyperscalers can dictate the trajectory of the most important stock in the market, any single capital spending pause stops being a company problem and becomes a systemic one.

China is now a rounding error

The geographic story has quietly inverted. Revenue from customers headquartered outside the US fell to 22% of the total, down from 42% a year ago. China-based revenue roughly halved to around $4.6 billion, and Nvidia shipped no datacenter products into China at all this quarter, against $4.6 billion a year earlier. The newer H200 licensing channel has generated no revenue yet, and any chips that do ship face a 25% tariff on the way back through a mandatory US inspection. 

The US now supplies close to $63.8 billion of the quarter's revenue, nearly 78% of the total. Nvidia has, for now, been effectively foreclosed from the Chinese data center market, and the company says as much in plain language. The bull case treats that as free optionality if policy thaws. The bear case notes that the world's second-largest economy has been swapped out for sovereign and US hyperscaler demand that may prove every bit as cyclical.

The $80 billion tell

Then there is what management chose to do with the cash. Nvidia’s board approved a fresh $80 billion buyback authorization on top of $38.5 billion still unspent, and lifted its quarterly dividend from a token $0.01 to $0.25 per share, a 25-fold raise. For comparison, one of the largest dividend increases ever (from a non-penny stock) was when the US mortgage black hole called Federal National Mortgage Association, or as it's colloquially known, Fannie Mae, raised its quarterly dividend 600%, from $0.05 to $0.35 per share. This was just months before the subprime mortgage crisis nearly wiped Fannie Mae out of existence.

Buybacks and a sudden dividend hike of that scale usually signal a company generating more cash than it can productively reinvest. For a firm still pointing to $119 billion in supply commitments and a relentless cadence of new architectures, that is either reassuring discipline or a quiet admission that the highest-return reinvestment days are starting to mature. Either way, $58.3 billion of quarterly net income buys a great deal of room to maneuver.

So Nvidia beat again, comfortably, and the operating business remains a genuine cash machine running at nearly 75% gross margin. However, the record headline profit leans on investment gains that can vanish at a moment’s notice, and the demand picture is increasingly entangled with Nvidia's own capital, while a shrinking circle of customers holds the whole thing up.

Nvidia veterans know better than most that a beat alone will not stop the stock from fading, but the harder question outlasts the print: When the most valuable company on earth starts booking the AI boom as investment income, is that a sign of total dominance, or the first hint that selling chips is no longer the whole story?