The U.S. Department of Defense is in advanced talks to lend approximately $5 billion to Fluidstack, a London-founded AI cloud-computing startup that has since relocated its headquarters to New York, marking what would be the largest loan ever issued by the Pentagon's Office of Strategic Capital (OSC). The money would not fund a single new data center but rather shore up domestic supply chains and manufacturing capacity for data-center components — a strategic distinction that reflects growing anxiety over U.S. dependence on foreign-made infrastructure.
The OSC, created under the Biden administration in 2022 and significantly expanded by the Trump administration's One Big Beautiful Bill Act, now has authority to deploy more than $200 billion in financing. Previously it has backed rare-earth material suppliers and drone manufacturers, and it has at times taken equity stakes or warrants in recipient companies. The timing tracks with a Trump executive order issued last month declaring a national emergency over American reliance on foreign-made electrical equipment — gear central to AI data-center buildouts.
Fluidstack is advising its application through Erebor Bank, a newly chartered national bank founded by Palmer Luckey, the entrepreneur who co-founded defense firm Anduril Industries and was an early, vocal Trump supporter in Silicon Valley. Erebor is actively pitching its ability to navigate federal financing programs as a core business line, and is separately seeking to raise at least $1 billion in venture capital at an $8 billion valuation. The bank's involvement signals a broader push by defense-adjacent fintech players to become gatekeepers for the AI-infrastructure funding boom.
Fluidstack's financial profile has grown rapidly. Earlier this year it raised $830 million led by Leopold Aschenbrenner's Situational Awareness fund, and more recently closed $1.5 billion from Jane Street at an $18 billion valuation. Much of its demand is underpinned by a major partnership with Google and Anthropic that gives Fluidstack access to Google's tensor processing units, with Google backstopping its data-center leases to reduce borrowing costs. The company has signed multiyear leases in New York, Indiana, Louisiana, and Texas, and employs 700 people now based in the U.S.
The broader financing picture is staggering: JPMorgan analysts projected in June that lenders will supply $4.1 trillion in AI infrastructure financing through 2030. The Pentagon's potential entry into this capital stack — alongside private credit, public markets, sovereign wealth funds, and hyperscalers — underscores how thoroughly national security and AI investment strategy have become intertwined.
Gist is a free AI reader for your browser, iPhone, and Android. Get concise summaries and key takeaways from any article or podcast.
Get Gist — Free