Nvidia Networking Revenue Hit $14.8B—Then the Reporting Map Changed

This quarter gave investors one last clean look through Nvidia's old Data Center window. Networking supplied about $20 of every $100 of compute-plus-networking revenue, up from $13 a year earlier. Nvidia will now emphasize who bought—Hyperscale or ACIE—rather than what sold—compute or networking, breaking a product trend line just as it becomes important.

The last frame before the reporting cut

Compute still supplied about 80% of the old product split and 73% of its annual dollar growth. Networking is now large enough to affect the platform story, but the thesis holds only if that attachment stays durable and observable after the reporting change.

How networking grew from $13 → $20 of every $100

For the quarter ended 2026-04-26, Nvidia reported $60.4B of Data Center compute revenue and $14.8B of networking revenue. Year-over-year growth was 77% and 199%; sequential growth was 18% and 35%.

The prior Form 10-Q reported $34.155B of compute and $4.957B of networking. Against the current rounded figures, networking moved from 12.7% to 19.7% of the product split.

The dollar bridge filters the small-base effect. Compute added about $26.245B and networking $9.843B, so networking supplied 27.27% of the $36.088B combined increase.

The 199% rate says networking accelerated; the 27.3% contribution says it mattered. Neither reveals product-level margin.

One business, two maps

The test is whether Nvidia can keep selling the network that makes many accelerators act like one machine. The old lens showed what sold; the new lens shows who bought. Customer breadth becomes clearer while product attachment becomes harder to track. No private consensus or price target is used.

Networking growth supports the integrated-stack claim, not a separate moat. The matched-boundary comparison separates reported revenue from narrower results and future targets. The AI revenue reading guide shows why a useful reporting lens still cannot substitute for the metric it replaces. For the company-level cash lens, see whether Blackwell is funding its own ramp.

What can disappear behind the new labels

The strongest countercase is concentration. Three direct customers represented 21%, 17%, and 16% of revenue—54% combined—and were primarily tied to Compute & Networking. Because direct customers can be distributors, manufacturers, cloud providers, or integrators, this is not end-customer concentration.

A strong total cannot disprove weaker attachment if the new lens erases the product series.

The next disclosure checkpoint

Nvidia guided Q2 FY2027 revenue to $91.0B, plus or minus 2%, with no China Data Center compute revenue assumed. The total tests momentum, Hyperscale versus ACIE tests breadth, and any voluntary product bridge tests observability.

The read strengthens if Data Center growth holds, ACIE expands without Hyperscale collapsing, and product data still show networking attachment. It weakens below a comparable 15% share, after major deployment delays, or when growth cannot bridge to the old series. Networking need not exceed compute; it must remain measurable and durable.

How the two periods were matched

Results, guidance, calculations, and interpretation are separated. Networking share uses the product total; growth contribution uses the combined increase. Current figures are rounded and were rechecked 2026-07-14.

• NVIDIA Q1 FY2027 earnings release
• NVIDIA Form 10-Q for the quarter ended 2026-04-26
• NVIDIA Form 10-Q for the quarter ended 2025-04-27

Filings, calculations, captions, and EN/KO parity were reviewed. No Nvidia sponsorship is disclosed. This is general information, not individualized advice, a rating, or a price objective.