Some investors are pushing back on the roughly $30bn valuation sought by British AI cloud provider Nscale for its planned New York listing, Bloomberg reported on Sunday. A range of $16bn to $19bn would be more likely to attract buyers, investors told the outlet. The dispute matters because Nscale plans to raise up to $3bn and its recent financing is tied to the $30bn level.
Contract base, listing terms and site delays
Nscale, which rents computing capacity for training and running AI models, filed for its IPO in September. It plans to list on the New York Stock Exchange under the ticker NSCL. Its last private funding round valued the company at about $15bn, while The Telegraph has put the possible listing value at up to $35bn. No final decision has been made and formal marketing of the shares has not yet started.
The company reported $103.4bn in signed and active contracts at the end of August, according to its IPO filing. Only a small share of that volume was already running. Microsoft and Anthropic account for about 85% of the total, Bloomberg has reported. That structure leaves future revenue dependent on two large customers moving from signed commitments to operating workloads.
Valuation is also linked to pre-IPO financing raised in September. Nscale raised $3.36bn in convertible notes led by Third Point, with $1bn from Nvidia still to arrive. The notes convert into shares at a discount, with the conversion price capped at a $30bn valuation. A lower IPO valuation would therefore directly affect terms for existing backers.
What the dispute means for buyers of AI capacity
Investor concerns centre on execution delays and weaker demand for AI infrastructure listings, according to Bloomberg. The flagship site in Norway is running about six months behind schedule. Microsoft has also withdrawn from a deal to be the main tenant at the planned data centre in Loughton, Essex. Grid operators do not expect to supply power to that site until the 2030s.
For companies that buy AI computing power, the episode points to a difference between contracted capacity and available capacity. Large customers can secure long-term commitments, while smaller buyers may face waiting times if flagship sites slip. A reduced valuation could make the listing easier to complete, but it would not by itself add operating data centres. Procurement plans should therefore separate announced contracts from commissioned power and hardware.
Several points still need verification before the shares are marketed. The timing of the Norway site reaching full operation remains unclear, as does replacement demand for Loughton after Microsoft exited weeks before the planned listing. Australian operator Firmus dropped its IPO earlier this month after weak demand, which shows the issue is not limited to one company. Buyers should ask vendors about power connection dates, tenant commitments and the share of revenue already running.
The marker to follow is the valuation range in the formal offer documents and whether marketing begins at $30bn or closer to $16bn to $19bn. A second signal is an update on Norway commissioning and a new anchor tenant for Loughton. If both move forward, the business case for contracted AI clouds will look stronger.
