Nvidia’s synthetic hyperscaler: software, finance, and no owned data centers
Nvidia appears to be building a “synthetic hyperscaler” by pairing its GPU hardware and software stack with large external financing to fund AI data centers run by others. Its DSX OS, Mission Control, Omniverse, and Dynamo cover the operational layer, while Apollo-linked lenders (BlackRock, Goldman Sachs, Brookfield, KKR, and peers) mobilize more than $500 billion for capacity. Neoclouds like CoreWeave, Nebius, Lambda, Crusoe, and Nscale supply the compute, but lack Nvidia’s balance sheet, shifting risk to outside investors. The model lets Nvidia monetize its reference architecture without owning data centers, yet it hinges on ongoing demand and heavy cash burn, raising questions about its moat versus risk.
