Nscale, a UK-based AI infrastructure company founded in 2024, is reportedly seeking $3.5 billion in pre-IPO financing as it explores a U.S. public offering as soon as this month.
According to Bloomberg, the proposed package includes $1.5 billion in convertible notes and a further $2 billion in financing from Nvidia. The discussions have not been confirmed by Nscale or Nvidia; TechCrunch reported that it contacted both companies for comment.
The timing matters because Nscale has become a conspicuous example of the new AI infrastructure business: companies securing large, long-duration compute commitments while raising enormous sums to buy, build and operate the capacity needed to fulfill them.
Financing the gap between contracts and capacity
Convertible notes are loans that may later turn into equity. For Nscale, that structure could provide capital before an IPO without immediately fixing a public-market valuation. But it also underscores a central feature of AI cloud economics: contracted demand does not eliminate the need for substantial up-front investment in GPUs, data centers, networking, power and operations.
Nvidia’s prospective $2 billion financing would be notable beyond its size. Nvidia is already an investor in Nscale, having participated in the company’s March Series B, a $1.1 billion round led by Aker. A deeper financial relationship could help Nscale fund growth, while also highlighting how closely AI infrastructure providers depend on the chipmaker’s hardware supply and broader ecosystem.
For operators evaluating compute partners, the key question is not simply whether a provider can announce capacity. It is whether it can finance and deliver that capacity on the timelines customers need, including power availability, hardware deployment, networking and service reliability.
Big contract figures require careful reading
Nscale recently signed a deal with Anthropic valued at about $45 billion, according to TechCrunch. Separately, reports said Nscale had told prospective investors it had roughly $103 billion in revenue after that deal.
That latter figure is not equivalent to recognized or current-period sales. The Information, as cited by TechCrunch, characterized it as a projection based on signed customer leases. That distinction is consequential for prospective investors and customers alike.
Long-term leases can provide an important demand signal and support financing discussions. Yet their eventual value depends on execution: customers must consume the contracted capacity, infrastructure must come online, and the provider must manage hardware costs, power procurement and debt obligations across the life of the agreements.
For founders and enterprise buyers, the lesson is to separate three metrics that are often bundled together in the AI infrastructure market: signed contract value, available capacity and realized revenue. Each says something different about a supplier’s position and risk profile.
A test of public-market appetite
Nscale raised $155 million in a Series A in December 2024, followed by the $1.1 billion Series B in March. A potential IPO so soon after those rounds would test whether public investors are ready to fund AI infrastructure companies at the scale private markets have supported.
The broader demand case is straightforward: advanced-model developers and enterprises need access to scarce, high-performance compute. The harder question is how much durable value remains with infrastructure providers after accounting for rapid hardware cycles, concentrated customers, high capital requirements and reliance on a small number of semiconductor and cloud partners.
What to watch next is whether the reported financing closes, what terms are attached to the convertible notes and Nvidia commitment, and whether Nscale provides clearer detail on the duration and conditions of its contracted revenue. Those disclosures would offer a more useful view of the company’s financial footing than headline contract totals alone.
