Why Neoclouds Exist: Nvidia Demand Needed a New Home

Why do neoclouds like Nebius, CoreWeave and IREN exist at all?
Ben Pouladian’s answer is that the hyperscalers created the opening themselves.
Google, Amazon and Microsoft decided Nvidia GPUs were too expensive to rely on forever, so they began investing heavily in their own custom AI chips.
But custom silicon is not a perfect substitute.
As Dylan Patel has argued, switching accelerators often requires redesigning models, software and infrastructure around an entirely different stack.
That leaves a large group of customers that still wants pure Nvidia compute.
Neoclouds stepped into that gap.
Their disadvantage is financing. Google can borrow at extremely low rates because of its balance sheet. A neocloud cannot.
So firms like Goldman Sachs, Brookfield and other infrastructure lenders finance the GPUs and data centers directly, allowing neoclouds to scale without needing hyperscaler-level balance sheets.
The result is a new layer in the AI stack:
Hyperscalers chase custom silicon.
Neoclouds absorb the demand that still wants Nvidia.
And Wall Street provides the capital that makes the model possible.