Crusoe, the AI data-center builder with supply contracts for Meta, Oracle, Microsoft and Google, is in talks to raise roughly $3 billion in a round that could triple its valuation to around $30 billion, Bloomberg reported. The company was valued at about $10 billion as recently as October, making this one of the fastest valuation re-ratings among AI infrastructure "neoclouds" this year.
Crusoe's path here is unusual: the company launched in 2018 as a cryptocurrency-mining operation designed to capture stranded natural gas that would otherwise be flared, before pivoting entirely toward AI data centers as the economics of GPU hosting outpaced crypto mining. That pivot now looks prescient -- Crusoe has contracts for nearly 5 gigawatts of committed AI computing capacity, with more than 40 gigawatts in its total project pipeline, serving four of the largest AI compute buyers in the world.
The round would follow Crusoe's $1.38 billion Series E in 2025, led by Valor Equity Partners and Mubadala Capital with participation from Nvidia and Salesforce Ventures, among more than a dozen other investors. A jump from that round's implied valuation to a reported $30 billion target places Crusoe in the same conversation as other AI-infrastructure names like CoreWeave, Together AI (recently valued at $8.3 billion) and Lambda, all of which have seen valuations climb sharply as AI labs scramble to lock in compute capacity years in advance.
“Deals like National Grid's $1.75 billion investment in Joulent's Texas gas-fired power project reflect the same underlying scarcity.”
The comparison to Together AI is instructive: Together AI's $800 million Series C valued it at $8.3 billion for its GPU-cloud and open-model-serving business, while Crusoe's reported target is nearly 4x that on the strength of its physical data-center footprint and long-term power contracts -- a reminder that investors are currently paying a premium for AI infrastructure with hard physical assets (land, power, buildings) over software-layer plays.
For infrastructure-focused VCs and LPs, Crusoe's trajectory reinforces a now-dominant 2026 thesis: the scarcest resource in AI isn't model talent or even GPUs directly, it's power -- specifically, the ability to secure gigawatts of contracted electricity fast enough to build data centers before hyperscaler compute demand outpaces available supply. Deals like National Grid's $1.75 billion investment in Joulent's Texas gas-fired power project reflect the same underlying scarcity.
The bear case: a valuation tripling in under a year on the back of contracts rather than proven long-term unit economics is exactly the kind of setup that has burned infrastructure investors before, and AI compute demand assumptions embedded in multi-gigawatt pipelines remain unproven if any major AI lab's growth slows.
What to watch: whether the round closes at the reported terms, who leads it, and whether Crusoe's crypto-to-AI-infrastructure pivot becomes a template other stranded-energy-adjacent startups try to replicate.