- Nscale is seeking up to $3.5B pre-IPO, split between $1.5B in notes and $2B from Nvidia.
- Third Point will lead the convertible notes; Goldman Sachs is advising on the deal.
- Nscale’s contracted deals now total around $103B, up from $51B a month earlier.
Nscale is in talks to raise up to $3.5 billion ahead of a planned US listing, according to a Bloomberg report citing people familiar with the matter. The London-based AI cloud provider is looking to sell up to $1.5 billion in convertible notes to a group of investors, while separately seeking about $2 billion in financing from Nvidia.
Daniel Loeb’s Third Point is expected to lead the convertible note investment, and Goldman Sachs is working on the fundraising.
The notes are reportedly priced at a double-digit discount to Nscale’s eventual IPO price, but the discount applies only up to a $30 billion valuation — above that, the conversion price remains unchanged.
If Nscale prices its IPO well above $30 billion, noteholders convert into more shares than IPO buyers receive for the same amount of money, making a high valuation more dilutive to existing shareholders than it would otherwise be.
A backlog that’s nearly doubled in a month
Nscale is briefing investors that its total contracts are worth around $103 billion, up from the roughly $51 billion figure it was citing when it first flagged a possible September listing.
Much of that jump traces to a six-year, $45 billion compute deal with Anthropic covering capacity at Nscale’s West Virginia campus, a deal that Microsoft and Google both reportedly passed on.
Nscale has told investors the backlog could eventually translate into around $18.1 billion in annual revenue and $13.6 billion in adjusted EBITDA (earnings before interest, taxes, depreciation, and amortization), though it stresses those figures are illustrative rather than formal guidance.
For context, actual revenue was around $33 million for all of 2025, climbing past $100 million in the second quarter of 2026, the gap between backlog and cash collected that convertible-note buyers and IPO bankers will be pricing into this deal.
This round would land on top of an already breathless run. Nscale closed a $2 billion Series C in March at a $14.6 billion valuation, closed roughly $3 billion in investment-grade debt at the end of August to fund GPU campuses in Ward County, Texas, and Madison County, North Carolina, and agreed to supply Figure with at least $3.5 billion of AI cloud compute, while taking an undisclosed equity stake in the humanoid robotics startup.
Nvidia is financing both ends of its own supply chain
Nvidia’s potential $2 billion check adds another layer to an already tangled relationship. The chipmaker is both a supplier and an investor across Nscale’s business: Nscale has contracted for around 194,000 of Nvidia’s next-generation Vera Rubin GPUs, and most of its currently active chips run on Nvidia’s Blackwell series.
Nvidia was also an investor in Figure before Nscale took its own stake in the robotics firm days earlier, meaning Nvidia now sits on both sides of a contract for its own hardware.
It’s the same circular pattern drawing scrutiny elsewhere in AI infrastructure: an investor funding the customers who buy its chips, then funding the cloud provider that sells compute to those same customers.
Nscale founder Josh Payne, whose LinkedIn traces a path from Australian coal mines to running a company chasing a possible $30 billion valuation, framed the Figure tie-up differently: “We’re proud to be enabling the future of AI robotics together.”
Nscale has also been buying its way up the stack rather than just renting out raw GPUs. In July, it agreed to acquire distributed AI software company Anyscale for around $1.65 billion, adding roughly 200 employees and customers including Coinbase, Runway, and Bedrock Robotics, part of what Nscale has framed as a bid to build a full-stack AI hyperscaler spanning power, GPUs, data centres, and software.
Whether that bet pays off now depends on a listing that bankers, noteholders, and Nvidia itself all have a stake in pricing correctly.
