Nvidia Splits Its Earnings Reports to Isolate Hyperscaler Pressure

*Nvidia will now break out revenue from the largest cloud providers separately from all other customers, drawing a line between segments where its chips face commoditization and segments where it still controls the full stack.*

The change appears in Nvidia’s upcoming earnings presentation. It reflects a deliberate effort to show investors two distinct businesses rather than a single, undifferentiated growth story.

Previously, Nvidia reported data-center revenue as one line item. That figure mixed sales to hyperscalers such as Microsoft, Google, Amazon, and Meta with sales to smaller cloud operators, enterprises, and research institutions. The new format will separate the two groups.

The company’s stated rationale is that hyperscaler purchases now dominate volume and carry different margin dynamics. In that segment Nvidia competes primarily on price and performance per watt against alternative accelerators and custom silicon. In the remaining segment the company supplies not only chips but also networking, software, and reference architectures, giving it greater pricing power.

No numerical breakdown accompanied the announcement. Nvidia said the revised categories will begin with the fiscal first-quarter report due in late May 2026.

Why the distinction matters

The split lets Nvidia present two narratives at once. One shows the scale of hyperscale demand while acknowledging that those buyers can and do design their own silicon. The other shows a long tail of customers who continue to purchase the complete Nvidia platform and therefore accept higher average selling prices.

Analysts have noted for months that gross margins on the largest orders have already begun to compress. By isolating those orders, Nvidia can demonstrate whether the compression is confined to the hyperscale bucket or whether it is spreading.

The move also reduces the risk that future quarters will be read as simple “beats or misses” against a single growth rate. Instead, observers will track two separate trajectories: one driven by volume and price competition, the other driven by platform lock-in.

Nvidia has not indicated whether it will provide historical recast numbers for the new categories. Without that data, quarter-over-quarter and year-over-year comparisons will be harder to calculate for at least one reporting cycle.

---

Sources:

• Stratechery

{
  "excerpt": "Nvidia will split its data-center revenue reporting to separate hyperscaler sales from all other customers.",
  "suggestedSection": "business",
  "suggestedTags": ["nvidia", "earnings", "ai-infrastructure"],
  "imagePrompt": "An abstract arrangement of layered translucent circuit boards and stacked server racks fading into soft shadow, with cool blue and warm amber light accents highlighting the separation between two vertical columns of hardware. muted color palette, cinematic lighting, 16:9"
}