Nscale is in active discussions with potential backers to secure roughly $3.5 billion in financing before pursuing an initial public offering, a person familiar with the matter said. The package under negotiation would split into two pieces: as much as $1.5 billion in convertible notes sold to a syndicate of investors and about $2 billion in equity or similar funding from Nvidia, the source added.
Goldman Sachs is advising Nscale on the fundraising. The convertible note tranche is expected to be led by New York-based hedge fund Third Point, according to the source.
Those convertible notes are being offered at a double-digit discount to Nscale's eventual IPO price. The discount mechanism would be adjusted up to a valuation threshold of $30 billion - beyond that point the conversion price would no longer be altered, the source said.
Representatives of Goldman Sachs, Third Point and Nvidia did not immediately respond to requests for comment, and Nscale declined to comment, the source said.
Deliberations over the identity of participating investors and the ultimate size of each investment remain ongoing and could change as talks progress, the source cautioned.
Earlier reporting indicated that Anthropic had agreed to a six-year, $45 billion contract to lease AI computing capacity from Nscale's West Virginia data center campus. That multi-year commitment was cited in discussions about Nscale's revenue prospects and capacity demand.
Nscale was valued at $14.6 billion in March after completing a $2 billion Series C funding round. The company, founded in 2024, operates its own data centers, its own graphics processing units and its own software stack to provide large-scale, GPU-powered AI compute.
Key players are positioned as follows - Nscale is courting institutional capital and strategic investment ahead of an IPO; Goldman Sachs is advising; Third Point is expected to lead the convertible notes; and Nvidia is the target for a sizeable investment allocation.
The discussions reflect investor appetite for companies that control end-to-end AI compute infrastructure, combining data centers, GPUs and software to support large AI workloads. However, the final structure and participation remain subject to negotiation and change.