London-based AI cloud company Nscale announced on September 25 that it had secured $3.36 billion in pre-IPO convertible financing, one of the largest private infrastructure raises of the year, as the self-described ‘neocloud’ operator races to build out power plants, liquid-cooled data centers and GPU clusters ahead of a planned New York Stock Exchange listing. The round was led by hedge fund Third Point and included Nvidia, Apollo-managed funds, Citadel, Hudson Bay Capital, the Abu Dhabi Investment Council and 8090 Industries, among more than a dozen institutional investors, underscoring how much capital is still flowing into companies that build the physical backbone for AI workloads even as questions mount about the sustainability of that spending.
A Record Pre-IPO War Chest
The $3.36 billion raise arrives just a week after Nscale filed its Form S-1 with the Securities and Exchange Commission on September 18, proposing to list on the NYSE under the ticker NSCL. Goldman Sachs is serving as placement agent on the convertible note offering, which company founder and chief executive Josh Payne described as putting Nscale in position to ‘accelerate our data center buildouts globally.’ The size of the raise reflects both the capital intensity of modern AI infrastructure and investor appetite for exposure to the buildout without waiting for a public offering to price.
Inside the Deal Structure
Unlike a traditional equity round, the financing is structured as convertible loan notes that automatically convert into ordinary shares once Nscale completes its IPO. Of the total, $2.36 billion closed immediately, while a further $1 billion commitment from Nvidia is expected to fund in mid-November. Nvidia’s stake will convert into non-voting shares, a structure increasingly common among chipmakers that want financial exposure to their own customers without the governance entanglements of a board seat. The layered tranches give Nscale committed capital it can deploy now while leaving room for the Nvidia commitment to land closer to the IPO itself.
From Power Plants to GPU Clusters
Nscale says the proceeds will go toward expanding behind-the-meter power generation, liquid-cooled data center capacity and large-scale GPU clusters, the three bottlenecks that have repeatedly slowed AI infrastructure expansion industrywide over the past two years. The company describes itself as a full-stack AI cloud platform spanning everything from power generation through physical data centers to the GPU compute layer itself, a vertically integrated model it argues lets it move faster than competitors that must negotiate separately for electricity, real estate and chips. Nscale has disclosed more than $103 billion in total contracted value across its customer agreements, a figure that gives some sense of the scale of compute capacity it has already committed to deliver.
Why Nvidia Keeps Writing the Checks
Nvidia’s participation fits a now-familiar pattern in which the chipmaker invests directly in the cloud operators and AI companies that purchase its GPUs, a strategy supporters say accelerates the buildout of infrastructure Nvidia’s own hardware depends on. Industry analysts tracking the AI infrastructure sector note that this kind of circular investment, where a supplier backs its own customers, has become increasingly common across the so-called neocloud segment, with Nvidia stakes now appearing in multiple GPU-focused cloud providers racing to secure long-term compute contracts with major AI labs. Proponents argue the arrangement is simply rational capital allocation given how tightly matched GPU supply is to demand right now.
Skeptics Question the Neocloud Boom
Not everyone is convinced the capital-intensive model is built to last. Some financial analysts have raised concerns about vendor-financed infrastructure spending, warning that cross-investments between chipmakers and the cloud operators that buy their products can obscure how much of reported demand is organic versus circular. Critics of the broader AI infrastructure financing wave have also pointed to the heavy use of debt-like convertible instruments, rather than straightforward equity, as a sign that investors want downside protection in case AI compute demand or public market appetite for AI infrastructure stocks cools before IPOs can price favorably. Those concerns sit alongside separate industry estimates, including one from IDC presented at October’s OCP Global Summit, that global data center infrastructure spending tied to AI could climb roughly 2.7 times by 2029, a trajectory that only works if revenue from AI services keeps pace.
What Comes Next
Nscale’s planned IPO will be an early test of how public markets price a pure-play AI infrastructure company at a moment when enthusiasm for the sector remains high but increasingly scrutinized. If the listing proceeds as outlined, the convertible notes from this round will convert at a price tied to the IPO valuation, meaning investors like Nvidia and Third Point have a direct financial stake in seeing the offering succeed. For now, Nscale’s rapid run from a $2 billion Series C earlier this year to a $3.36 billion pre-IPO raise illustrates just how much money is still chasing AI data center capacity, even as the industry debates whether that capacity will be matched by durable demand in the years ahead.
Photo: PIRO4D / PIXABAY via Pixabay
