The $145 Billion Circular Bet: Inside Nvidia's GPU Financing Machine

Microsoft and Meta have committed over $122 billion to neoclouds like CoreWeave and Nebius. But the financing is circular — Nvidia invests in these companies, they buy Nvidia GPUs with debt, and those same chips back the loans. The Verge calls it 'Chipwrecked.' What happens if the GPU-backed loan bubble pops?

The AI infrastructure boom has spawned a new kind of business model — the neocloud — and with it, a financing structure that has some analysts reaching for the red pen.

CoreWeave and Nebius, the two most prominent public neoclouds, have together secured 7 GW of contracted power capacity and pulled in over $145 billion in total commitments from hyperscalers. Microsoft alone has committed roughly $60 billion across CoreWeave, Nebius, and private player Nscale. Meta has added another $62.2 billion. OpenAI and Anthropic are also in the mix, pushing total potential commitments past the $145 billion mark.

Those are staggering numbers. For perspective, CoreWeave's estimated FY2026 revenue is $12.6 billion and Nebius expects $3.4 billion — meaning their order backlogs are an order of magnitude larger than current sales.

The value proposition is real: neoclouds can deploy the latest Nvidia GPUs in weeks, compared to multi-year builds for hyperscale data centers. CoreWeave was first to market with H100, H200, and GB200 NVL72 instances. Speed matters when every quarter of AI compute delay means lost competitive ground.

But the financing that powers this speed is where things get circular.

Nvidia itself invested $2 billion into CoreWeave in January 2026 to help the debt-ridden company add 5 GW of compute capacity, as TechCrunch reported. The arrangement is elegant in a hall-of-mirrors sort of way: Nvidia invests in neoclouds, neoclouds use that capital — plus massive debt loads — to buy Nvidia GPUs, and Nvidia's financing arm backs the loans with the very same chips. Revenue flows from hyperscaler contracts through neoclouds, right back to Nvidia's top line.

As IO Fund analyst Beth Kindig detailed in a June analysis, neoclouds "do not have the same cash nor operating cash flow profiles of Big Tech. This is leading neoclouds to employ unique and circular financing structures, raising some red flags."

The Verge put it more bluntly in a piece titled Chipwrecked: Can Nvidia avoid the crash?, asking what happens if the GPU-backed loans go south.

The bears argue hyperscalers are using neoclouds to offload capex from their balance sheets — shifting AI infrastructure spend from capital expenditure to operating expense. If AI demand softens or cheaper inference models reduce the need for cutting-edge training clusters, those long-term capacity contracts could become liabilities rather than assets.

For now, the flywheel is spinning. Nvidia reports record revenue. Neoclouds book record backlogs. Hyperscalers get GPU capacity faster than they could build it themselves.

Whether the music stops gracefully or abruptly is the $145 billion question.

Sources: IO Fund, The Verge, TechCrunch