Cache storage powers Vast Data’s AI infrastructure edge

As AI infrastructure investment scales globally and inference workloads multiply, cache storage is emerging as the critical data layer that makes AI factories functional, persistent and economically viable in an era of disaggregated computing.

VAST Data Inc. has positioned itself at the center of that shift, having closed its Series F financing at a $30 billion valuation — more than tripling its 2023 valuation — with NVIDIA Corp. among its investors. The round reflects both the company’s revenue trajectory and the architectural decisions it made years ago to build a unified AI data platform from the ground up, according to Jeff Denworth (pictured), co-founder of VAST Data.

“The valuation, from our perspective — there’s a lot of hyperbole, there’s a lot of hype in this market,” Denworth said. “And to have professional investors do the diligence to confirm that our market leadership is tangible and is real, it’s helpful for us to inspire confidence in our future customers.”

Denworth spoke with theCUBE’s John Furrier at RAISE Summit 2026, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed AI infrastructure scaling, the rise of Neo Clouds, KV cache storage for disaggregated inference, and VAST’s deepening partnership with NVIDIA. (* Disclosure below.)

KV cache storage and the disaggregated AI infrastructure shift

The AI infrastructure buildout has entered a new phase in which scale is no longer just about GPU clusters — it is about the data systems that keep those clusters fed. Flash cost surges, capacity constraints and the emergence of disaggregated inference are forcing a fundamental rethink, Denworth noted. VAST’s platform, which counts CoreWeave Inc. as a customer through a commercial agreement valued at $1.17 billion, is now being architected explicitly for this new class of workload.

“The biggest deals that we have in our pipeline are actually KV cache oriented,” Denworth said. “The rise of disaggregated computing and disaggregated inference has created a new market with respect to storage for very, very high throughput KV cache storage. We’ve been applying all the same rigor that we normally do from an enterprise data management perspective on top of this data.”

The scale claims backing that pipeline are significant. VAST recently built a single-system one-exabyte AI storage cluster — representing 1% of the world’s annual flash production — supporting an environment with 300,000 GPUs, a scale Denworth said exceeds anything built by public cloud providers. Beyond raw scale, VAST is expanding into what it calls Neo Clouds: a new class of AI-native service providers that use VAST’s platform to offer a full catalog of data services, from vector databases to event brokers, on top of GPU infrastructure.

“VAST is much more than a storage product,” Denworth said. “We have our own database services, our own runtime, our own pipelines, our own event broker — you put that all together, it turns out we’re working with Neo Clouds very, very quickly to up-level their service catalogs so that they can provide a comprehensive set of services to their users. If I pin my prices to a leading hyperscaler vector price, our customers who are the clouds make like 98% gross margin on this.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of RAISE Summit 2026:

(* Disclosure: TheCUBE is a paid media partner for the RAISE Summit event. Neither Solidigm, the headline sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE