Nvidia Broke Out $14.8 Billion of Networking Revenue in Its April Quarter. The August Release Publishes No Split at All.

by Nikolai Petrov · August 27, 2026

Technology Reporter, Frontier Tech Wire

Nvidia reported its second fiscal quarter of 2027 after Wednesday's close, and the number that moved the AI complex was the October-quarter guide: revenue of $108.0 billion, plus or minus 2 percent, according to the company's earnings release. Sister coverage across this network has the earnings arithmetic. What follows is a narrower point about disclosure, and it matters most to the small and mid-cap suppliers that sit downstream of Nvidia's networking business rather than its accelerator business.

The reporting framework itself is not new this quarter, and it is worth being precise about what changed and when. In the quarter ended April 26, 2026, Nvidia's release said the company was, in its words, "transitioning to a new reporting framework that better reflects its current and future growth drivers." That April release already presented the business as Data Center and Edge Computing, and it already carried no separate Gaming, Professional Visualization or Automotive figures — the three lines the comparable release a year earlier had broken out at $4.3 billion, $601 million and $586 million respectively. What the April release also carried, and the year-earlier one did not, was a split of Data Center revenue into compute and networking: Data Center of $75.2 billion, comprising $60.4 billion of compute and $14.8 billion of networking, alongside $6.4 billion of Edge Computing, on total revenue of $81.6 billion. It also described two Data Center sub-markets, Hyperscale and what the company labelled ACIE. All of those figures appeared in that release's narrative text, not in its financial statements.

The release published Wednesday for the quarter ended July 26, 2026 keeps the framework and drops the split. Its highlights give Data Center revenue of $89.0 billion, up 18 percent from the previous quarter and up 117 percent from a year ago, and Edge Computing revenue of $7.2 billion, up 13 percent sequentially and up 27 percent year over year. Neither is a reported segment line in the accounting sense: both appear in the release's narrative, while the condensed consolidated statements of income carry a single undifferentiated revenue line of $96,221 million for the quarter, with no segment detail beneath it. There is no compute-versus-networking breakdown anywhere in the document, and no Hyperscale-versus-ACIE breakdown either. So this quarter's change is a subtraction within an existing framework, not the arrival of a new one.

The distinction did not disappear — only the number did

The compute category is still live in Nvidia's own language. The single China reference in the outlook section reads, in the company's words, that Nvidia "is not assuming any Data Center compute revenue from China in its outlook." That sentence draws a line between compute and everything else in the Data Center segment. The company is evidently still tracking the distinction internally. It simply is not publishing where the line falls this quarter.

For context on scale: at the April quarter's disclosed figures, networking was $14.8 billion of $75.2 billion in Data Center revenue, or a little under 20 percent of the segment. Applying that ratio forward is exactly the kind of interpolation this desk does not do, and readers should treat any figure that purports to be Nvidia's July-quarter networking revenue as an estimate by whoever produced it, not a disclosure.

One caveat, stated plainly, because this article is an argument from absence and those fail if the number is merely somewhere the writer did not look. This analysis is based on the earnings press releases as published on Nvidia's newsroom and investor relations sites, read in full including the financial statement tables and the non-GAAP reconciliations. Segment footnotes in the company's quarterly report filed with the Securities and Exchange Commission are a separate document and were not reviewed for this piece; they may present the segment detail differently, and a reader who needs the split should look there before concluding it does not exist. The claim here is scoped to the press releases and no further: the release for fiscal Q2 2027 does not carry the compute-versus-networking split that the fiscal Q1 2027 release did.

Why a missing line item is a supply-chain story

Optical transceiver makers, cable assembly houses, switch silicon suppliers and connector specialists have spent two years being valued partly against a demand signal they do not control and cannot see directly. Nvidia's networking line was one of the few quarterly, company-published anchors available. It was crude — it aggregated switching, network interface cards and interconnect across product generations — but it was a published number from the buyer's side of the ledger rather than an estimate from the sell side.

With that anchor absent for at least this quarter, the read-through path gets longer. Investors and analysts now have to work from the suppliers' own reported results, from the total Data Center figure, and from qualitative commentary. That is a real change in information quality for the smaller names in the chain, several of which have very concentrated customer bases and correspondingly volatile quarterly results.

What the guide does and does not say

The October-quarter guide of $108.0 billion implies roughly $11.8 billion of sequential revenue growth on the $96.221 billion reported, or about 12 percent. Nvidia guided gross margin to 74.0 percent plus or minus 50 basis points on both a GAAP and a non-GAAP basis, a step down of about 100 basis points from the 75.0 percent it reported for the July quarter on both measures. GAAP operating expenses are guided to approximately $9.2 billion and non-GAAP operating expenses to approximately $9.0 billion, with a tax rate of 16 to 18 percent on both bases.

Two reported figures are worth flagging for anyone reading headline numbers. GAAP net income of $59.688 billion and GAAP diluted earnings per share of $2.46 both exceeded the non-GAAP figures of $53.954 billion and $2.22 — the reverse of the usual direction, in which non-GAAP excludes charges and comes in higher. The release does explain it, in two places, and the explanation is worth stating rather than leaving the inversion to look like someone's arithmetic error. First, the reconciliation on the face of the release bridges GAAP net income of $59.688 billion to non-GAAP net income of $53.954 billion, and the largest single reconciling item is a subtraction of $7.771 billion for net gains from equity securities — an investment gain the company removes because it is not operating income. Total pre-tax non-GAAP adjustments come to negative $7.251 billion, against a $1.517 billion income tax impact; $59.688 billion less $7.251 billion plus $1.517 billion is $53.954 billion, which reconciles exactly. Second, the release states that "beginning in the first quarter of fiscal 2027, NVIDIA's non-GAAP financial measures no longer exclude stock-based compensation expense," and that the historical non-GAAP information presented has been updated to include it. Subtracting an investment gain while no longer adding back stock compensation is enough to put non-GAAP below GAAP. The inversion is a definitional consequence of the company's own stated policy, not a mistake, and it means non-GAAP figures for fiscal 2027 are not comparable to non-GAAP figures the company published before that change without restating them.

On China, the outlook assumes nothing. That is a guidance assumption, not a forecast of policy, and it is the same posture the company took a year earlier, when the fiscal Q2 2026 release stated there were no H20 sales to China-based customers in that quarter and that no H20 shipments to China were assumed in the following quarter's outlook.

For the frontier-tech names further down the chain — the pre-revenue photonics startups, the quantum hardware companies buying cryogenic and control electronics, the optics suppliers with single-digit-percentage share of a hyperscale program — none of this changes their own cash runway, dilution risk or order books. It changes how much of the demand picture they can be measured against using a document anyone can read. That is a smaller thing than a guidance number and a more durable one.

This article contains no investment recommendation and no view on the value of any security mentioned.