London AI startup Nscale seeks $2.7B in funding to build global data centers in partnership with ByteDance

London-based AI startup Nscale is seeking $2.7 billion to build global data centers in partnership with ByteDance, aiming to power the next wave of AI development.

Nscale, which launched from stealth in May 2024, is already setting its sights high. According to an offering document seen by Bloomberg, the company is looking to raise a massive $2.7 billion to build out artificial intelligence infrastructure. The plan is tied to a pending partnership with ByteDance Ltd., the Chinese tech giant behind TikTok.

“Nscale, a startup that is less than a year old, is already trying to raise about $2.7 billion to build artificial intelligence infrastructure on the back of a pending partnership with ByteDance Ltd., the Chinese owner of TikTok, Bloomberg reported.

If Nscale follows through, the money would go toward setting up data centers around the world, packed with Nvidia chips. These facilities would be leased to companies like ByteDance to train and operate their AI models.

This move comes just four months after Nscale pulled in $155 million in Series A funding. That round, led by Sandton Capital Partners with backing from Kestrel 0x1, Blue Sky Capital Managers Ltd, and Florence Capital, was aimed at fueling expansion across Europe and North America.

Nscale isn’t just another data center company. It builds sustainable AI-focused facilities, rolling out large-scale GPU infrastructure and offering AI cloud services. The company’s full-stack approach — handling both the data center design and the GPU supercluster infrastructure — lets it squeeze out better speed, performance, and efficiency for AI workloads.

Nscale’s platform is built to support the entire lifecycle of generative AI projects. Developers can run and train models on bare-metal or virtualized GPU nodes, tap into Kubernetes-native services (NKS), manage AI workloads with SLURM-powered scheduling, and choose between serverless or dedicated inferencing options. Nscale also plans to launch a public cloud offering in Q1 2025, giving developers flexible access to inference and training tools built specifically for AI.

Since coming out of stealth mode in May 2024, Nscale has seen huge demand. Its pipeline of greenfield data center projects across Europe and North America has ballooned from 300MW to 1.3GW, with 120MW planned for 2025.

These upcoming sites are being built specifically to handle large supercluster deployments for hyperscalers and companies running massive language models. Nscale is betting on next-gen tech like closed-loop direct liquid cooling to maximize performance while keeping environmental impact low. With this setup, the company says it’s ready to deliver custom GPU clusters at any scale, whether for governments, growing AI startups, or global enterprises.

“The AI market is scaling rapidly, and so are we,” comments Joshua Payne, CEO of Nscale. “Nscale manages every layer of infrastructure in the value chain to meet the intensive needs of large-scale AI customers. In particular, the largest risk to the market’s ability to scale is the large contiguous tranches of electricity required to power these large GPU superclusters.”

Payne added, “Nscale has a 1.3GW pipeline of sites in our portfolio, which allows us to design from the ground up, the data centre, the supercluster, and the cloud environment end-to-end for our customers. This means we can deploy bespoke GPU clusters tailored to customers’ requirements, do this faster than competitors, and at any scale, all while delivering superior unit economics. With the support of our investors, we can now reach more customers globally with bespoke-designed, sustainable, and cost-effective AI infrastructure that unlocks new AI capabilities, products, and services.”

If Nscale can pull off this $2.7 billion fundraising goal, it would instantly put the company on the map as one of the most ambitious new players in AI infrastructure — and a serious contender in a field that’s about to get even more competitive.