NVIDIA (NVDA) says one direct customer represented 16% of total revenue in Q2 and that 44% of H1 total revenue was from three direct customers

• We will be shipping both Blackwell and Rubin systems in the future, and currently experiencing certain supply constraints.

Customer concentration disclosures of this kind are a recurring feature of large-cap semiconductor filings, and in past episodes of hyperscaler-driven demand cycles the pattern has been for a small number of direct buyers, typically the major cloud platforms, to account for a large share of revenue at the peak of a buildout. The analytical distinction worth drawing is between direct customers and end demand: direct customers are often distributors, server original design manufacturers, or integrators, so disclosed concentration can overstate true end-user concentration even where the underlying buying ultimately traces back to the same handful of hyperscalers. That gap matters because it determines whether concentration risk sits in credit terms and inventory at intermediaries or in final capex budgets. Historically, rising concentration at a dominant supplier has cut both ways, signalling strong order books while leaving the equity sensitive to any single buyer's capex guidance, which is why hyperscaler earnings calls tend to trade as a read-through for the supplier. The supply constraint language alongside it fits the established sequence of constrained leading-edge capacity during an accelerator ramp, where allocation rather than demand sets the print. Follow-ons are the filing's full segment detail and any shift in the identity or share of the named counterparties over coming quarters.