Nscale filed a US IPO registration statement on September 18, revealing that contracts worth up to $44.6 billion with Anthropic still await required financing. Nscale sells data-center capacity and software to teams building, adapting and running AI models. It combines GPU systems with power, cooling, networking and storage so customers can use the computing capacity without assembling the physical infrastructure themselves.
Anthropic is buying dedicated, single-tenant infrastructure rather than an ordinary shared-cloud account. A disclosed services order covers Nvidia Vera Rubin rack-scale systems and the associated computing, networking and data-center facilities. Anthropic receives logical control of the systems, including the operating system and software stack, while Nscale provides and manages the underlying facilities.
The contract makes delivery the dividing line between a large commitment and billable service. Anthropic pays by GPU-hour only after accepting a capacity tranche. Late delivery can trigger discounts and eventually allow Anthropic to terminate the affected tranche, while either party can terminate the order if qualifying equipment and facility financing isn’t secured by a contractually defined deadline. Investing.com reported from the filing that Nscale had not yet secured that financing.
Nscale also markets on-demand GPU and CPU computing, virtual machines, networking and storage, alongside managed inference endpoints and software for fine-tuning models. That menu serves a different workflow: developers can provision infrastructure or use a managed endpoint instead of committing to a dedicated single-tenant deployment.
The filing puts Nscale’s current operations beside those commitments. The company reported $140.6 million of first-half revenue and a $1.02 billion net loss, while Axios reported more than $103 billion of total contract value and Investing.com reported that only $2.6 billion was active at the end of August. The Anthropic portion still has to move through financing, delivery and customer acceptance before it produces service fees.