The British AI cloud and data-center Nscale filed for a U.S. IPO on Sept. 18 and plans to list on the New York Stock Exchange under NSCL, according to its SEC registration statement. The filing does not yet specify the number of shares or an IPO price range.
Nscale’s current operations are still small compared with the size of its future commitments. Revenue reached $140.6 million in the first half of 2026, up 1,252% year over year, while its net loss widened to approximately $1.02 billion.
$103B in Contracts Changes the Valuation Question
The reason investors may look past current revenue is Nscale’s enormous contract book.
By the end of August, the company had built more than $103 billion of active and contracted total contract value, with around 461,000 GPUs active or contracted. Major customers include Microsoft and Anthropic.
That makes Nscale less a bet on what it earns today and more a bet on whether it can actually build enough data-center capacity to convert those commitments into revenue over time.
We previously reported the company’s IPO filing and $1 billion six-month loss, while the broader AI infrastructure financing boom shows why rapidly growing contract books do not eliminate the need for huge amounts of outside capital.
| Metric | Amount |
|---|---|
| Contracted revenue / TCV | $103B+ |
| H1 2026 revenue | $140.6M |
| H1 2026 net loss | $1.02B |
| Recent convertible financing | $3.1B |
| Nvidia portion | $1B |
Nvidia Is Now Part Supplier, Investor and Financier
Nscale recently secured approximately $3.1 billion in convertible financing, including $1 billion from Nvidia.
That relationship is important because Nvidia is not simply selling GPUs into the AI buildout. It is increasingly helping finance companies that buy and deploy its hardware.
Coinpaper’s deeper look at who is financing the $3.6 trillion AI data-center boom showed how Nvidia, private-credit funds, banks and bond investors are becoming part of the same capital ecosystem funding new compute capacity.
For Nscale, the financing is essential. Building enough capacity to service more than $100 billion of contracted business requires data centers, power, networking and enormous GPU deployments before much of that revenue arrives.
