Fluidstack, Valued at $18 Billion, Partners with Google to Expand AI Data Center Capacity
Fluidstack, a startup recently valued at $18 billion, is playing a crucial role in Google's strategy to expand its AI chip business and compete with Nvidia. The company is building and operating data centers for Anthropic's large-scale TPU deployment, securing gigawatts of power resources from Bitcoin miners across multiple US locations, including Glendale, Arizona, and sites in Texas. Fluidstack is rapidly expanding its computing power capacity with significant new financing.
Fluidstack, a previously low-profile British startup, has achieved an $18 billion valuation, positioning itself as a key partner in Google's efforts to expand its Tensor Processing Unit (TPU) business and challenge Nvidia's dominance in AI chips. Fluidstack's core business involves building and operating AI data centers, notably becoming the first publicly known TPU data center operator outside Google itself. The company is facilitating Anthropic's million-scale TPU deployment project, which involves a $500 billion investment in data center construction, equating to 1 gigawatt of computing power.
Fluidstack is aggressively expanding its footprint, operating more than 10 sites this year with a target of 1.3 gigawatts of computing power, and an ambitious goal to exceed 17 gigawatts by 2030. To achieve this, Fluidstack has signed multi-billion dollar long-term agreements to secure gigawatts of power resources from Bitcoin miners such as TeraWulf, Cipher Mining, and Hut 8. The company is also making significant moves in real estate, having quietly leased over 1 million square feet for two buildings in Glendale, Arizona, and establishing new data center sites in Texas. Fluidstack has also secured substantial financing, including a $1.5 billion round led by Jane Street, alongside other major investors like BlackRock, Atreides Management, and Spark Capital, and has raised over $15 billion through debt financing for project construction.
This rapid expansion is underpinned by Fluidstack's claim of completing data center construction in three months, significantly faster than tech giants like Google, Amazon, or Meta. While Fluidstack is deeply tied to Google, it operates within a competitive landscape where Anthropic also utilizes hardware from other providers like Nvidia and Amazon Web Services. The article also touches upon a legal dispute involving SemiAnalysis and a former employee concerning alleged misuse of sensitive TPU data, underscoring the high stakes and competitive environment of the AI chip and data center market.