VAST Data raises $1B at a $30B valuation after 10 years building AI infrastructure for hedge funds, AI labs, and hyperscalers

VAST Data raises $1B at $30B valuation

VAST Data has closed a $1 billion round at a $30 billion valuation, capping a decade-long bet that AI would eventually need a purpose-built infrastructure stack. Renen Hallak founded the company in 2016, when GPU clusters were still largely the province of quant hedge funds and life science institutes rather than the labs and hyperscalers that dominate today.

“We started in 2016. We thought AI required a new infrastructure stack way back then. We want to abstract this new hardware away from these new models and applications — make it easy, make it secure for everybody to generate AI agents and use AI in production. Jensen gave a really good analogy of a five layer cake: power, chips, infrastructure, models, and applications. We're that middle layer.”

What VAST actually does

Hallak describes VAST as the middle layer of what Jensen Huang calls a five-layer AI stack — power, chips, infrastructure, models, and applications. The pitch is that as AI moves from frontier labs into enterprise production, someone needs to abstract the underlying hardware away and make it consumable. VAST calls its abstraction layer a "data space," a single interface that lets customers run across on-premises, cloud, and hyperscale environments simultaneously.

CoreWeave is VAST's most prominent distribution partner. VAST supplies software that CoreWeave turns into cloud services — storage, database, streaming — that end users consume without necessarily having a direct relationship with VAST. Adobe is one example where the direct relationship does exist; Hallak says most large enterprises end up with both a cloud and an on-prem footprint, which is exactly where VAST's abstraction layer earns its value.

The hedge fund angle

Before generative AI, hedge funds were VAST's most important customers, and Hallak argues they remain underappreciated as a source of compute demand. He says the largest quant funds had bigger GPU clusters than any AI lab before the generative AI wave hit. Jane Street is a customer both directly and through CoreWeave. Hallak declines to get into specifics — the funds are deliberately opaque — but suggests the aggregate compute footprint of the major quants is larger than that of the social platforms, largely because there are far more of them than there are Metas.

Supply chain

Hallak flags a NAND shortage and a memory shortage as live constraints right now, alongside the better-publicized GPU scarcity. DPUs (data processing units) are also in short supply. His read is that no component category is in abundant supply, that nobody anticipated this growth curve, and that the hardware buildout will remain constrained for at least several years.

What the $1B is for

The capital is going toward filling out VAST's software infrastructure layer. As the company's customer base shifts from frontier labs toward broader enterprise adoption, Hallak frames the next phase as a simplification problem — making AI agents and production AI easier and more secure for organizations that don't have lab-scale engineering teams.