AI cloud company Nscale is reportedly preparing for a major US initial public offering that could raise as much as $3 billion. The London-based company could list its shares as soon as September 2026, according to people familiar with the matter. Nscale has hired Goldman Sachs and JPMorgan Chase to work on the possible deal.
The IPO plan is not final yet. The company could change the size of the offer or delay the listing. Still, the report shows how strong investor interest has become in companies that provide the computers, data centers, chips, and power needed for artificial intelligence.
Nscale is not a normal cloud company. Its main business is to build and operate large data centers for AI workloads. It then provides customers with access to the computing power inside those sites. That puts the company close to the heart of the fast-growing AI market.
A Company Built Around AI Infrastructure
The rise of AI has created a huge need for computing power. AI companies need thousands of advanced chips to train and run their models. Those chips need large data centers, strong power supplies, cooling systems, and software.
Nscale aims to provide much of this infrastructure.
The company has data center projects in places such as Norway and West Virginia. It currently has about 831 megawatts of active and contracted power. It also plans to add about 10 gigawatts of power for its AI computing centers.
The difference is enormous. Nscale is not just trying to add a few new data centers. Its plan calls for a much larger network that can support a huge amount of AI computing.
The company has also expanded its chip base. As of the second quarter of 2026, Nscale had about 25,000 active chips and about 289,000 active and contracted chips. Most of its active chips are from Nvidia’s Blackwell family. It has also contracted for about 194,000 Nvidia Vera Rubin GPUs.
$51 Billion in Contracted Revenue
One of the biggest numbers in the Nscale IPO story is its reported $51 billion in total contracted revenue.
Nscale has shared this figure with potential investors before the IPO, according to people familiar with the matter. The number represents the value of contracts that customers have made with the company. It does not mean Nscale has already received $51 billion in cash or recorded that amount as current revenue.
That difference matters.
Nscale’s actual revenue is much smaller today. Company executives reportedly told potential investors that revenue rose to more than $100 million in the second quarter of 2026. That compares with about $37 million in the first quarter and roughly $33 million for all of 2025.
The sharp rise shows that Nscale’s business is growing at a very fast rate. At the same time, investors will want to know how much of the $51 billion in contracts will turn into real revenue, when that revenue will arrive, and what costs the company must pay before it can earn profits.
Microsoft Is a Major Customer
Microsoft is an important part of Nscale’s growth story.
Nscale’s large West Virginia project has Microsoft as its anchor tenant. Microsoft has commissioned 1.35 gigawatts of capacity from Nscale for that site. Nscale plans to provide this computing power with Nvidia’s Vera Rubin NVL72 systems. Each system has 72 Rubin GPUs.
Nscale also plans to host 300,000 previous-generation Blackwell Ultra chips for Microsoft at four other sites.
The infrastructure contract for those GPUs is reportedly worth about $14 billion. This gives investors a clear example of where Nscale’s large contract figure comes from.
The Microsoft relationship also matters for another reason. A major customer can provide long-term demand and help support the large cost of building AI data centers. But it can also create customer concentration risk if a small number of companies account for a large share of future business.
The West Virginia Project
Nscale’s West Virginia project is one of its most ambitious plans.
The company has a 2,250-acre data center campus there. Nscale says the site could theoretically support more than 8 gigawatts of computing capacity. The campus will also have a dedicated electrical grid and on-site power infrastructure.
Power has become one of the biggest challenges for the AI industry. Modern AI data centers can require enormous amounts of electricity. As companies add more GPUs, they also need more power to run and cool those machines.
This makes access to electricity almost as important as access to chips.
Nscale’s plan to add 10 gigawatts to its existing 831 megawatts therefore shows the scale of its ambition. It wants to secure the power first and then use that capacity to support more AI computing.
Nscale Is Also Building Software
Nscale’s business is not limited to physical data centers.
The company also offers cloud services that help customers use AI computing resources. Its platform supports managed versions of Kubernetes and Slurm, two tools that help companies manage computing workloads.
Nscale also has a prompt engineering tool that helps developers improve the quality of responses from AI models.
In July, the company agreed to buy AI software startup Anyscale for $1.65 billion. Anyscale is known for its commercial version of Ray, an open-source tool that helps companies manage and optimize AI computing clusters.
The deal shows that Nscale wants to offer more than raw computing power. It wants to give customers software that helps them make better use of that power.
Why Investors May Pay Close Attention
The possible IPO comes at a time when investors have shown strong interest in AI infrastructure.
Building AI systems requires huge amounts of money. Companies need Nvidia chips, data centers, electricity, networking equipment, cooling systems, and other hardware. Nscale is part of this wider push to build the infrastructure required for the next phase of AI.
That gives the company a strong growth story.
Its revenue has risen quickly. Its contract value is large. Its planned power capacity is expanding. It has major technology partners and customers. It also has plans to add software to its cloud business.
But the same story carries major risks.
AI data centers cost billions of dollars to build. Nscale must spend large sums before it can earn revenue from many of its projects. The company must also secure enough power, equipment, financing, and customers to make its plans work.
The $51 billion contract figure also needs careful attention. Contracted revenue is a sign of future demand, but it is not the same as money already earned.
A Test for the AI Infrastructure Market
Nscale’s IPO could become an important test for the value investors place on AI infrastructure companies.
The company could seek as much as $3 billion from the public market. The exact size, valuation, and timing remain uncertain. Goldman Sachs and JPMorgan are working with Nscale on the possible deal, while the company has not confirmed the final IPO details.
If Nscale completes the listing at a strong valuation, it could show that public investors still have a strong appetite for AI infrastructure. It could also encourage more private companies in the sector to consider public markets.
If investors demand a lower valuation, however, it could show that the market wants more proof that huge AI contracts can turn into lasting profits.
What Comes Next for Nscale
For now, Nscale’s story is about scale.
The company has reported $51 billion in contracted revenue, more than $100 million in second-quarter 2026 revenue, 831 megawatts of active and contracted power, and plans for another 10 gigawatts. It also has major Nvidia chip commitments and a large Microsoft relationship.
The possible US IPO could give investors a closer look at the numbers behind that growth.
Yet the final test will not be the size of the contracts alone. Investors will want to see how quickly Nscale can turn those contracts into revenue and, later, profit.
For the wider AI industry, the potential IPO is another sign that the AI boom has moved far beyond model companies. The next major battle is for the physical infrastructure that makes AI possible. Nscale wants to be one of the companies at the center of that market.
