London-based data center and software stack startup Nscale is reportedly seeking to raise US$2.7 billion to develop AI infrastructure.

This information comes from documents reviewed by Bloomberg, which indicate that the funding would facilitate the construction of global data centers equipped with Nvidia chips for companies like ByteDance to train and operate AI models.

The proposed financing plan includes US$1.8 billion in private credit led by Goldman Sachs, as well as an additional US$900 million through preferred equity and convertible shares.

This funding effort is linked to a potential US$2 billion contract with ByteDance.

However, a spokesperson for ByteDance said that these figures “significantly overstate the scale of any possible cooperation.”

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ AI infrastructure startups are seeing unprecedented capital concentration at record speed

Nscale’s attempt to raise $2.7 billion less than a year after founding represents an extraordinary acceleration in the funding timeline for infrastructure companies.

This trend is evident across the industry, with CoreWeave accumulating $13.4 billion in total funding and reaching a $23 billion valuation despite being a relatively young company 1.

Similarly, Lambda Labs has sought three major funding rounds in 2024 alone, raising $500 million in April, $320 million in February, and now pursuing an additional $800 million 2.

The capital intensity of this sector directly relates to the enormous costs of AI model development, exemplified by Google’s Gemini 1.0 Ultra, which cost an estimated $192 million just to train 3.

These funding amounts significantly exceed traditional startup trajectories, as AI infrastructure companies must finance massive hardware acquisitions before generating significant revenue. This creates a capital-intensive race that favors early, aggressive fundraisers.

2️⃣ “Neocloud” providers are creating a specialized middleware layer in the AI economy

Companies like Nscale and CoreWeave represent the emergence of specialized AI infrastructure providers that sit between chip manufacturers (primarily Nvidia) and AI developers.

CoreWeave has expanded from three to fourteen data centers and quadrupled its workforce to meet the surging demand for GPU-accelerated cloud infrastructure 4.