Nvidia's Operating Profit Rose 19%, Net Profit 2%: Investment Gains Fell

Nvidia released its May-July results on Wednesday, August 26, about twenty minutes after the US close. Revenue came in at $96.2 billion, up 106% from a year earlier and 18% from the prior quarter. Data center alone contributed $89.0 billion, or 92.5% of the total. Wall Street's consensus was $92.2 billion; the company came in 4.3% above it.

No line in the release missed. But two profit lines separated, and the distance between them is what this quarter is actually about.

Operating profit — what is left after selling chips and paying the costs of doing so — went from $53.5 billion to $63.7 billion, a 19% gain in three months. Net profit went from $58.3 billion to $59.7 billion, a 2% gain in three months.

Same company, same quarter, two different stories. What separates them is not chip sales.

Why Two Profit Lines Diverge

Profit is not a single number on an income statement. Operating profit sits near the top: sell the product, subtract manufacturing cost, research spending and overhead, and what remains is what the business itself earned. Below that, non-operating items are added: interest income, currency effects and — decisive in Nvidia's case — changes in the value of shares the company holds in other businesses. Net profit is what falls out at the bottom.

What Nvidia has done over the past two years has made that bottom section large. The company took equity stakes in some of the same customers it sells chips to: cloud providers, AI labs, data center operators. CNBC reported in May that Nvidia's equity investments through 2026 had passed $40 billion. Some of those stakes trade publicly and some do not; accounting rules require both to be marked to current value at period end, with the difference running through the income statement.

The gain is booked whether or not a single share is sold. There is no cash in the bank; there is a line on the statement.

The company's own CFO commentary states it plainly: "Net gains from equity securities for the second quarter were $7.8 billion."

One quarter earlier, the figure at the end of that same sentence was $15.9 billion.

The $8.1 billion decline between those two numbers says nothing about Nvidia's business. It says that the shares in its portfolio rose less in one quarter than in the other. But on the net profit line, it swallows most of the increase the chip business generated.

The Arithmetic: Ten Billion Up, Eight Billion Down

Three different growth rates come out of one set of accounts: revenue up 17.9%, adjusted profit up 18.5%, net profit up 2%. The first two measure how fast the business is running. The third measures the mood of a portfolio.

The same split shows up in earnings per share. Reported diluted EPS was $2.46; adjusted EPS was $2.22. That order is normally reversed, because the adjusted figure adds back expenses such as stock-based compensation and therefore sits above the official one. At Nvidia it has been inverted for two quarters running, because the investment gains stripped out are larger than the expenses added back. For anyone learning to read a set of accounts, this is a clean demonstration of why one number is not enough.

Booked Profit and Cash Received Are Not the Same Thing

The second piece of evidence for that distinction is on the cash side. Nvidia booked $59.7 billion of net profit in the quarter. Cash generated by operations in the same quarter was $24.1 billion — 40% of the reported figure.

Part of the gap is inventory: stock rose from $25.8 billion to $31.6 billion, so $5.8 billion of cash went onto shelves. A meaningful share of the rest is the appreciation of shares that were never sold. That $7.8 billion is profit on the income statement and nothing at all in the bank account.

Cash and marketable debt securities stood at $56.6 billion at quarter end. During the quarter the company returned $26.0 billion to shareholders through buybacks and dividends, and $99.0 billion remains unused under its repurchase authorization.

The chart covers the three months before the release, not the hours after it.

Timeline

The after-hours price reaction was still forming as this was written; less than an hour had passed since the release and the call with management was under way. Going in, 24/7 Wall St reported that the options market was pricing a move of roughly 5.4% on the print — at this company's size, about $280 billion of value swinging one way or the other. There is no data yet on which way it went, and the point here is not the price but which profit line the price is being set against.

The Other Side

Both readings hold together. Choosing between them depends on how you view a supplier taking equity in its own customers — and that is not a question the accounts can settle.

What Is Left: A $279 Billion Purchase Commitment

The line with the longest life in this release is not on the income statement at all; it is in the notes. The CFO commentary reads: "Our commitments increased from $119 billion last quarter to $279 billion, primarily related to the procurement of memory."

That is a $160 billion increase in three months. It is what the company has committed to pay suppliers in future periods, and most of it is high-bandwidth memory — the memory stacks mounted alongside AI chips that determine how fast those chips can be fed. Against third-quarter revenue guidance of $108 billion, the commitment is roughly 2.6 times a quarter's sales.

Locking in capacity while memory prices are rising has a price, and it shows up in the margin guidance: a 75% gross margin this quarter, guided to 74% (plus or minus 50 basis points) for the third. On $108 billion of revenue, one point is about $1.08 billion a quarter. That is precisely where Micron, SK Hynix and Samsung collect.

Not everyone expected that. Gabelli analyst Ryuta Makino said ahead of the release: "I think NVDA has the best relationships with the leading HBM vendors and should be able to get more favorable pricing versus the market." The guided one-point decline suggests those relationships did not fully absorb the price increase.

China is effectively closed. Per the CFO commentary, shipments of Hopper data center products to China during the quarter were less than 1% of data center revenue.

Jensen Huang, in the release: "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue."

Because Nvidia is the single largest weight in the Nasdaq 100, these two charts usually point the same way. The index's past month shows how the wait for this release was priced; the release itself falls outside the window.

This article is based on Nvidia's second-quarter results release of August 26, 2026, the CFO commentary published the same day, and the prior quarter's CFO commentary filed with the SEC. Consensus figures come from analyst averages compiled by Kiplinger and 24/7 Wall St; the pre-release share move is as reported by Axios. The after-hours price reaction had not yet formed at the time of writing and is not included here. This is not investment advice.