Vast Data positions the information layer for AI scale

The AI arms race has centered on compute: Who has essentially the most graphics processing units, the fastest chips and the most important clusters? But a distinct pressure point is emerging as enterprises move AI from pilot programs into continuous, production-grade operations. Vast Data Inc. has built its strategy around that gap, offering a knowledge infrastructure platform designed to serve because the operating layer between enterprise datasets and the AI workloads that depend upon them.

That ambition is backed by measurable momentum. Vast crossed $2 billion in cumulative software bookings in under six years, has maintained positive money flow for 12 consecutive quarters and holds a gross margin approaching 90%, in keeping with Paul Nashawaty, principal analyst at theCUBE Research.

“Vast Data is de facto carving out a novel spot within the cloud-native world,” he said. “They’re not only one other storage vendor; they’re going after the role of a cloud-native AI infrastructure platform.”

Partnerships with Nvidia Corp., Microsoft Corp., CoreWeave Inc., Google Cloud and Solidigm, a trademark of SK Hynix NAND Product Solutions Corp., signal that major infrastructure players increasingly view the platform as foundational to AI at scale, in keeping with Dave Vellante, chief analyst at theCUBE Research.

“We’re witnessing the evolution of the ‘AI Factory,’ where the main target is shifting from experimental model training to continuous, high-scale inference,” he said. “For the enterprise, the challenge is increasingly around data velocity and architectural efficiency. Vast Data is at the middle of this transition, providing the high-performance foundation required to run agent-based systems in production.”

That shift might be on display at Vast Forward. Join theCUBE, SiliconANGLE Media’s livestreaming studio, on Feb. 25 for exclusive coverage of Vast’s evolving platform architecture, the operational demands of agentic AI and the way the corporate’s expanding partner ecosystem supports production-scale deployments. (* Disclosure below.)

Inside Vast Data’s AI operating system

Vast’s platform comprises 4 integrated layers — DataStore, DataSpace, DataBase and DataEngine — each addressing a distinct stage of the AI data pipeline, from storage and cataloging to real-time analytics and distributed processing. The Disaggregated Shared-The whole lot Architecture, or DASE, separates compute logic from storage, enabling organizations to scale capability and processing power independently. The approach positions the platform as an alternative choice to traditional data lake and lakehouse models, with an architecture designed to administer AI workloads end-to-end.

“In our opinion, success in AI requires greater than just access to GPUs,” Vellante said. “Keeping those invaluable resources might be just as essential. Vast is playing directly into that reality with an architecture and an operating model built around utilization.”

Vast Forward is predicted to construct on themes introduced at the corporate’s Cosmos event and reinforced at KubeCon + CloudNativeCon NA, where Vast detailed how its platform embeds compute directly into the information layer. That capability enables AI models — including embedding models for video and audio — to operate across cloud and on-premises environments with Kubernetes because the orchestration layer. Product updates around DataEngine and DataSpace are among the many areas to observe, together with expanded support for edge processing and autonomous agent workflows, in keeping with Vellante.

“Especially in a market where flash and memory efficiency have gotten tactical mandates, Vast’s ability to optimize the information layer is a critical differentiator for any organization moving toward an always-on AI reality,” he said. “Our research indicates the industry’s memory constraints are forcing efficiency to the forefront — and Vast is leaning in tactically with flash efficiency programs designed to do more work per byte and per watt.”

An ecosystem for production-scale AI

Vast’s partner strategy spans cloud, compute and storage, extending earlier collaborations with Nvidia, Microsoft and CoreWeave while adding a recent alliance with Google Cloud. The Microsoft partnership, announced in late 2025, brings the Vast AI OS to Azure, giving enterprise customers access to tools resembling InsightEngine for real-time data evaluation and AgentEngine for orchestrating autonomous AI workflows across hybrid cloud environments, in keeping with Jeff Denworth, co-founder of Vast Data.

“The target is to principally bring the entire goodness of what we call the Vast AI Operating System to Microsoft customers,” he said. “We’ve been collaborating with them for, I’d say, the higher a part of two years now. The target is to principally open up data access to all of the client’s data sets for all different compute platforms that Microsoft offers.”

On the compute infrastructure side, Vast and CoreWeave formalized a $1.17 billion business agreement that positions the Vast AI OS as a primary data foundation across CoreWeave’s network of GPU-dense cloud data centers. The deal allows CoreWeave to deploy Vast’s platform on demand, giving joint customers low-latency access to massive datasets for AI training and inference without the bottlenecks related to conventional storage architectures.

“Our deep integration with CoreWeave is the results of a long-term commitment to working side by side at each the business and technical level,” said Renen Hallak, founder and chief executive officer of Vast. “By aligning our roadmaps, we’re delivering an AI platform that organizations cannot find anywhere else out there.”

Solidigm, one other key Vast partner expected to feature prominently at Vast Forward, adds a storage economics dimension. A joint whitepaper from the 2 firms found that all-flash architectures deliver a 58.9% lower total cost of ownership than traditional hard-disk-drive tiers, difficult long-held assumptions in regards to the need for tiered storage, in keeping with the paper.

“We imagine the timing of Vast’s customer event is fortuitous since it spotlights an actual inflection,” Vellante said. “Specifically, AI infrastructure is moving from experimentation to production, and efficiency — especially in memory and flash — has develop into a first-order key performance indicator.”

TheCUBE event livestream

Don’t miss theCUBE’s coverage of Vast Forward on Feb. 25. Plus, you’ll be able to watch theCUBE’s event coverage on demand after the event.

Easy methods to watch theCUBE interviews

We provide you various ways to observe theCUBE’s coverage of Vast Forward, including theCUBE’s dedicated website and YouTube channel. It’s also possible to get all of the coverage from this yr’s events on SiliconANGLE.

TheCUBE podcasts

SiliconANGLE’s “theCUBE Pod” is on the market on Apple Podcasts, Spotify and YouTube, which you’ll enjoy while on the go. During each podcast, SiliconANGLE’s John Furrier and Dave Vellante unpack the most important trends in enterprise tech — from AI and cloud to regulation and workplace culture — with exclusive context and evaluation.

SiliconANGLE also produces our weekly “Breaking Evaluation” program, where Dave Vellante examines the highest stories in enterprise tech, combining insights from theCUBE with spending data from Enterprise Technology Research, available on Apple Podcasts, Spotify and YouTube.

Guests

During Vast Forward, theCUBE will feature interviews with executive leaders at Vast, Nvidia, Solidigm, CoreWeave and more. Discussions will concentrate on the impact of agentic AI on operations and the way platform architectures are evolving to handle these challenges.

(* Disclosure: TheCUBE is a media partner for Vast Forward. Sponsors of theCUBE’s coverage, including presenting sponsor Solidigm, don’t have editorial control over content on theCUBE or SiliconANGLE.)

Image: SiliconANGLE