Nvidia's networking business is a fifth of the data centre, and Nvidia no longer tells you

Nvidia reported $89.02 billion of Data Center revenue for the quarter ended July 26th, and nowhere in the 10-Q it filed on August 26th does it say how much of that was networking. The Next Platform put the figure at $17.45 billion, or 19.6% of the data centre.

That is a model, and its author says so.

A year ago it was not a model. Here is the same filing, twelve months earlier, in the management discussion:

Data Center compute revenue was $33.8 billion, up 50% from a year ago. Sequentially, compute revenue declined 1%, driven by a $4.0 billion reduction in H20 sales. Networking revenue was $7.3 billion, up 98% from a year ago and up 46% sequentially, driven by the growth of NVLink compute fabric for GB200 and GB300 systems, the ramp of XDR InfiniBand products, and adoption of Ethernet for AI solutions at cloud service providers and consumer internet companies.

That paragraph is gone.

In its place the current 10-Q splits Data Center by customer — Hyperscale at $48.71 billion, AI Clouds, Industrial and Enterprise at $40.31 billion — and adds a new Edge Computing line at $7.20 billion. A footnote explains the change. "In the first quarter of fiscal year 2027, we changed our presentation of revenue by market platform, and the comparable periods were recast." Recast, not restored. Nobody moved the product split. It left.

So where does 19.6% come from? From Collette Kress saying on a call that networking was about $15 billion and compute about $60 billion in the April quarter, and from an analyst growing those forward. We ran the only comparison the filings support. Networking was $7.3 billion against $41.10 billion of Data Center revenue in the July 2025 quarter, which is 17.8%. If the modelled $17.45 billion is right, the mix has moved roughly two points in a year while the base has more than doubled.

Two of those three bars are not audited.

Does that matter to you? Only if you are pricing Broadcom or Arista. Networking is the line with named competitors who report quarterly, and it is now the line Nvidia describes least.

Our read is that the reclassification is defensible and convenient at once. Something genuinely new arrives with the customer cut — the neoclouds, sovereigns and enterprises in the ACIE bucket grew 138% and are now nearly as large as the hyperscalers. And something leaves: the single number a competitor's investor could benchmark against, in the quarter Nvidia's own supply commitments went from $119 billion to $279 billion. Companies rarely reorganise a disclosure in a direction that costs them.

There is a good defence. Segment presentation follows how the chief operating decision maker actually looks at the business, and the accounting standard is written to make that so. Nvidia's Compute & Networking reportable segment still exists and still reports $88.30 billion. It just contains both things and separates neither.

We would expect the compute-versus-networking split to stay out of the Q3 filing in late November, and to keep appearing only as round numbers on earnings calls, where nobody signs anything. If Nvidia restores it, we will have read the motive wrong and will say so.

One networking detail the filing does keep. Nvidia's networking supply chain, including its Israel operations, employs about 6,200 people — roughly one employee per $2.8 million of modelled quarterly revenue. Take the model away and you cannot compute even that.