London-based Nscale has greatly expanded its AI data center business in recent years, and is currently building data centers in Norway and West Virginia in the US. As global technology companies' demand for AI computing power grows rapidly, the company hopes to raise more capital through the capital market to support its huge infrastructure expansion plans.

People familiar with the matter have previously revealed that Nscale has indicated to potential investors that before going public, the total contract revenue of the company was about 51 billion US dollars, providing important business support for its IPO.

However, discussions on the IPO are still ongoing. The specific financing scale and listing time may still change, and there is a possibility that the IPO will be postponed. Nscale declined to comment.

According to the plan currently being discussed, NScale's current US IPO may raise up to about 3 billion US dollars. If successfully completed, it will be one of the largest listed financing projects in the AI infrastructure field in the near future.

Nscale's board lineup also received a lot of attention from the market, including Cheryl Sandberg, former chief operating officer of Meta Platforms (META.US), and Nick Clegg, former president of global affairs at Meta and former British Deputy Prime Minister.

nScale's business model mainly revolves around building large-scale AI data centers, purchasing a large number of AI chips and other infrastructure equipment, and then leasing computing power to customers who need to train and run artificial intelligence models.

With the rapid development of generative AI, global technology companies' demand for GPUs, data centers, and power resources has increased dramatically, and AI infrastructure companies such as nScale are investing tens of billions of dollars to expand computing capabilities.

People familiar with the matter revealed earlier that the company plans to add about 10 gigawatts (GW) of electricity capacity to its AI computing center. Prior to that, Nscale already had approximately 831 megawatts (MW) of operating or contracted power capacity.

As a reference, 1 gigawatt of electricity can meet the electricity needs of approximately 750,000 American households at any given time. This means that if Nscale finally achieves its goal of adding 10 gigawatts, the scale of its AI infrastructure will expand significantly.

The huge demand for electricity also reflects the current capital-intensity of AI infrastructure construction. Companies not only need to buy a large number of AI chips, but must also invest huge sums of money to build data centers and obtain sufficient electricity supply and related infrastructure resources.

In addition to expanding the size of data centers, nScale is also expanding into the field of AI software and computing power management.

In July of this year, the company agreed to acquire software startup Anyscale for $1.65 billion, hoping to use its technology to help customers use artificial intelligence computing resources more efficiently.

By combining data center infrastructure with software platforms, Nscale hopes to not only provide customers with computing power, but also improve the efficiency of the use of AI computing power, thereby expanding its business coverage in the AI infrastructure industry chain.

As Nscale is preparing to go public, AI infrastructure companies are setting off a wave of capital market financing.

In May of this year, the data center acquisition platform Blackstone Digital Infrastructure Trust raised $2 billion through an IPO; Csquare, which is supported by Brookfield, went public and raised $1.21 billion in July.

Meanwhile, the data center operator Switch has secretly submitted a listing application and is likely to launch an IPO in November of this year as soon as possible. Earlier news showed that the valuation sought by the company in a round of financing could be close to 50 billion US dollars, including debt.

Behind this series of financing activities is the rapid expansion of capital demand for global AI infrastructure construction. As large-scale AI models require more and more computing resources, data center operators need to invest huge sums of money to buy chips, build computer rooms, and obtain sufficient electricity supply, prompting more and more enterprises to turn to stock and debt market financing.