Argentum AI just signed one of the bigger cloud GPU supply deals of the year, and oddly enough, nobody’s saying who the customer actually is.
Under the contract, Argentum will supply roughly 27,000 Nvidia GB300 GPUs over a long-term period, delivering all of it through its own cloud platform. The deal carries a $4.1 billion price tag. Still, the buyer shows up only as a “leading AI company,” which in this industry could mean almost anyone with deep pockets and an appetite for cloud compute it can’t build fast enough on its own.
Argentum plans to roll out the capacity in phases starting this year, though it hasn’t said where the infrastructure will actually sit. That’s fairly typical for deals this size, but it also means the public has to take the company’s word for both the math and the timeline.
“AI demand continues to outpace infrastructure availability across the market,” said Andrew Sobko, Argentum’s founder and CEO, framing the deal as proof that his company can move fast on large-scale cloud GPU delivery. He added that as foundation models keep growing and inference workloads pile up, companies increasingly can’t function without reliable access to scalable cloud GPU infrastructure.
This isn’t Argentum’s first big swing, either. Just last month, the company signed a separate $2.5 billion agreement with cloud gaming company Boosteroid and DL Invest Group to build a data center exceeding 300 megawatts in Bielsko-Biala, Poland. Combined, these two deals push Argentum’s infrastructure commitments to roughly $6.6 billion within a matter of weeks.
According to the company, it already runs operations, or has contracts in place, across more than 15 countries, and it’s reportedly courting US banks and global investment firms to fund further cloud expansion. Even so, the real question is whether that expansion can keep pace with the deals piling up. After all, promising 27,000 GPUs through the cloud is one thing. Actually installing, powering, and running them on schedule is a much harder problem, and one that has tripped up plenty of companies chasing this same AI infrastructure boom.
