Bitget App
Trade smarter
MarketsTradeFuturesEarnAISquareMore
AI "New Cloud" operator Nscale (NSCL.US) races to US IPO, aiming to raise up to $3 billion; Anthropic spends $45 billion to secure computing power

AI "New Cloud" operator Nscale (NSCL.US) races to US IPO, aiming to raise up to $3 billion; Anthropic spends $45 billion to secure computing power

智通财经智通财经2026/09/18 23:36
Show original
By:智通财经

Artificial intelligence data center developer Nscale has submitted an initial public offering application to the U.S. Securities and Exchange Commission, planning to list on the New York Stock Exchange with the stock symbol "NSCL."

According to Zhitong Finance APP, AI data center developer Nscale has filed for an initial public offering (IPO) with the U.S. Securities and Exchange Commission (SEC), planning to list on the New York Stock Exchange under the ticker symbol “NSCL.” According to previous reports, the company is seeking to raise up to approximately $3 billion through this IPO, with proceeds to be used for data center equipment investments to meet the continuously growing computing power demand of the AI industry.

Nscale is headquartered in London and is one of the “new cloud” operators rapidly emerging amid the recent AI infrastructure boom. The company primarily leases computing resources to enterprises developing AI models and applications. Its partners include Nvidia (NVDA.US) and Microsoft (MSFT.US), and it has signed large-scale computing power agreements with AI firms such as Anthropic.

However, Nscale’s rapid expansion comes with significant losses. According to documents submitted by the company to the SEC on Friday, for the six months ending June 30, Nscale achieved revenue of $140.6 million, a more than 12-fold increase from $10.4 million in the same period last year. During the same period, net losses expanded from $368.9 million to $1.02 billion.

Nscale was spun out of a cryptocurrency mining business at the beginning of 2024 and quickly transformed into an AI data center operator. Unlike traditional hyperscale cloud providers such as Amazon AWS, Microsoft Azure, and Google Cloud, Nscale belongs to the “new cloud” category that has risen rapidly in recent years. It primarily builds infrastructure around GPUs and other AI computing resources and then leases this computing power to large tech companies and AI model developers.

In March of this year, Nscale completed a Series C funding round led by Norwegian industrial investment group Aker and investment firm 8090 Industries, valuing the company at around $14.6 billion. Nvidia and Nokia also participated in this funding round. Nscale's intent to raise up to $3 billion in its IPO suggests that the company may further tap into public capital markets to fund its rapidly expanding data center and GPU infrastructure.

One of Nscale’s core assets for fast expansion is its massive power and data center resources. As of August 2026, the company’s data centers own or control more than 10 gigawatts (GW) of power capacity, with contracts totaling approximately $103 billion in value.

The company is building so-called “hyperscale AI hubs” in Norway, Portugal, Texas, and West Virginia, each with power capacity of over 200 megawatts (MW). The data center in Narvik, Norway, has become a key project of Nscale’s collaboration with Microsoft. In April this year, the company agreed to supply an additional 30,000+ Nvidia chips on top of the existing computing power lease agreement with Microsoft, further expanding its AI computing capabilities.

Nscale’s business model requires significant upfront investment in GPUs, data center construction, and securing abundant power resources, while long-term computing power lease agreements with major tech and AI companies help increase predictability of future revenue.

One of the company’s most notable customer agreements is with Anthropic. In August 2023, Anthropic agreed to spend $45 billion to lease AI cloud computing capability from Nscale’s flagship project in West Virginia, the Monarch Compute Campus. The Monarch campus covers about 2,250 acres. According to filings Nscale submitted, this project is expected to have around 2GW of total power capacity online by the first half of 2028 and is planned to further expand to about 8GW by 2031.

If these plans proceed smoothly, Monarch will become one of Nscale's key infrastructure bases for future computing power expansion. Meanwhile, in July, Nscale also agreed to acquire San Francisco-based AI software startup Anyscale for $1.65 billion, aiming to improve clients’ usage efficiency of AI computing resources via enhanced software capabilities. This marks the company’s extension from pure data center and GPU power provision towards the AI infrastructure software layer.

Nscale’s close relationship with Nvidia is a crucial foundation for the company’s rapid expansion of AI computing supply, but Nscale also lists this dependency as a risk factor in its IPO filings. The company states that it relies on Nvidia for GPU supply. In its Series B funding round in 2025, Nscale issued warrants to Nvidia, making the AI chip giant a shareholder with over 5% equity. Should Nvidia exercise those warrants in the future, it will acquire non-voting shares in Nscale.

Their capital ties have continued to deepen recently.

According to the IPO filing, as part of a subscription agreement, on September 15 Nscale agreed to offer investors $2.1 billion in unsecured convertible loan notes, and up to $1 billion in unsecured convertible loan notes or non-voting shares specifically to Nvidia, with the Nvidia-related transaction expected to be completed on November 16.

For Nscale, forging close capital and supply chain relationships with Nvidia helps secure the most critical GPU resources for AI infrastructure construction; on the other hand, any changes in GPU supply, procurement conditions, or cooperation between the two parties could impact the company’s expansion plans.

Despite rapid revenue growth, Nscale is still in a phase of large-scale capital investment. For the six months ended this June, the company’s revenue surged from $10.4 million last year to $140.6 million, but net losses grew from $368.9 million to $1.02 billion. This simultaneous rapid increase in revenue and losses reflects the asset-heavy expansion model typical of AI data centers. GPU procurement, electricity infrastructure, land acquisition, data center construction, and cooling systems all require massive upfront capital, while income realization occurs gradually as computing power comes online and client contracts are executed.

For Nscale as it prepares for the public market, what investors may focus on is not only its contract portfolio exceeding $100 billion, but also how fast these contracts are converted to revenue and cash flow, and how much capital the company must invest to fulfill related orders. Nscale’s board is also notable, including former Meta Platforms (META.US) COO Sheryl Sandberg and former Meta head of global affairs and former UK Deputy Prime Minister Nick Clegg.

The company’s IPO filing shows that Aker, Sandton Capital Partners, and a trust entity linked to Nscale founder and CEO Josh Payne all hold at least 5% of the company’s shares.

Nscale’s decision to go public coincides with the AI infrastructure theme becoming a major driver in the global IPO market. Data shows that IPO fundraising so far this year has reached about $161.4 billion, the highest since 2021, with companies offering AI infrastructure investment exposure being a key driving force.

This year, SpaceX raised a record $86.2 billion through its IPO, with the sale of AI computing power to companies like Anthropic being a key part of its listing story; Korean memory chipmaker SK Hynix’s $26.5 billion listing also highlighted data center demand. Brookfield-backed data center firm Csquare raised $1.2 billion through an IPO in July.

Nscale’s IPO is led by Goldman Sachs (GS.US), JPMorgan Chase (JPM.US), and Morgan Stanley (MS.US), with 19 other banks participating as underwriters.

0
0

Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.