VAST Data's $30B mark is a bet on the middle layer of AI, not storage

VAST Data already had about $1 billion in the bank before it raised another one. Co-founder Jeff Denworth said so from the VAST Forward 2026 stage in Salt Lake City. CEO Renen Hallak has said the quiet part plainly in on-record interviews: the company is generating more than $100 million of cash per quarter and hasn't raised since December 2023. That's the context for today's $1 billion Series F at a $30 billion post-money. The round isn't fuel, it's a scoreboard.

The question worth asking is what the scoreboard is actually measuring.

The deal

Drive Capital and Access Industries led as new investors. NVIDIA, Fidelity, NEA, Tiger Global, Goldman Sachs, Norwest, General Atlantic, Dell Technologies Capital, BOND, Next47, Greenfield, and 83North participated from the existing cap table. Alphabet's CapitalG, reported in August 2025 to be in talks to co-lead, did not appear in the closing list.

Structure matters more than the headline. VAST confirmed the financing is both primary and secondary, with the tender letting employees and early investors sell alongside the new money. Liquidity rather than operations reads a specific way. With Amy Shapero, Shopify's former CFO, already in the seat and Hallak floating an IPO at VAST Forward, the tender is a pre-IPO release valve. Stripe and Databricks ran the same play.

VAST has raised about $1.38 billion lifetime and sits at roughly 3.3x its December 2023 Series E mark of $9.1 billion.

Denworth spelled it out today: the new capital will sit in reserves earning interest. The round funds validation and optionality, not operations.

The multiple

Pick your denominator and the story changes.

VAST's own announcement discloses committed ARR over $500 million and cumulative bookings above $4 billion. Against projected 2026 ARR near $600 million, $30 billion is somewhere between 50x and 60x forward. Trailing ARR estimates land closer to $2 billion after a better-than-3x growth year, dropping the multiple to about 15x. Both can be true. CARR is booked and signed, recognized ARR is billing, and the gap is where the growth story lives.

For contrast, Everpure (the rebranded Pure Storage, NYSE: PSTG) trades publicly at roughly $17 billion on several billion in revenue. Low single-digit multiples. VAST is priced at 1.7x Everpure on a fraction of the revenue. Not an indictment. The bet: VAST becomes a category, and the category is bigger than storage.

Do the math on capital efficiency and the case sharpens. The company has raised about $1.38 billion lifetime against $4 billion-plus in cumulative bookings and $500 million-plus committed ARR, and is operating cash flow positive at a reported $100 million-plus per quarter. Denworth pegs VAST's Rule of X at 228%, against a 61% average for the top ten public software companies. Bookings are roughly three times lifetime capital raised. Infrastructure startups don't usually look like that.

The pivot that matters

Hallak is no longer selling a storage product. He has framed it as storage to data platform to AI operating system — a better description of what the company actually does in 2026. The framing appears in the VAST Forward 2026 keynote recap on VAST's own blog: VAST as the middle layer, the software infrastructure between GPUs and applications.

The positioning is where the $30 billion has to earn its keep. As a storage company, $30 billion is indefensible against any reasonable comp. As the operating system for AI infrastructure, VAST is being priced against Databricks and Snowflake, and the math bends. Denworth's post-raise framing is the clearest statement of the ambition yet: "VAST is not a category killer. The broad reach of VAST's AI OS establishes us as a killer of categories."

The product stack is six engines on VAST's Disaggregated Shared-Everything (DASE) architecture: DataStore for unified object/block/file, DataBase for structured query and native trillion-vector search, DataSpace for global namespace federation, DataEngine for serverless compute, InsightEngine for NVIDIA-integrated embeddings and RAG, and AgentEngine for a Kubernetes-based agentic runtime. At VAST Forward in February, the company unveiled PolicyEngine and TuningEngine for end-of-year release, the latter running LoRA, supervised fine-tuning, and RL pipelines inside the platform and feeding agent outcomes back into continuous improvement loops.

Hallak frames DASE as the moat. The company can't take open source or acquire competitors and bolt on their tech, because every piece of the stack descends from the same shared-everything architecture. Denworth's VAST FWD 2026 recap makes the longer version of the point: continuous AI systems can't be built on fragmented infrastructure. Read it as a pre-buttal to anyone expecting VAST to go shopping with the new cash. The architecture forecloses it.

The customer list is the thesis

Every neocloud that matters runs on VAST.

CoreWeave signed a five-year commercial agreement estimated at $1.17 billion in November covering 23 data centers, more than 250,000 GPUs, and 500-plus petabytes of capacity. xAI's Colossus cluster in Memphis, north of 200,000 NVIDIA GPUs, runs on VAST. Denworth has said publicly that xAI is "training and checkpointing and storing their data on VAST," and VAST's own account of the xAI relationship points to a significant TCO advantage at Colossus scale. Lambda, Crusoe, Nebius, Nscale, and G42's Core42 are all on the list.

The hyperscaler arc is documented in VAST's own press releases. Google Cloud integration landed at Google Cloud Next in April 2024, expanded to deeper collaboration in April 2025, and arrived in November 2025 as a fully managed AI OS on Google Cloud Marketplace. A week later at Ignite, Microsoft and VAST announced a strategic collaboration to power agentic AI on Azure. Under that sits a conventional enterprise book: NASA, DOE, Verizon, Zoom, Booking Holdings, Pixar.

Brian Venturo, CoreWeave's co-founder and CSO, called VAST's AI Operating System the underpinning of how CoreWeave designs and delivers its AI cloud. Not boilerplate. It's the architecture admission that decides the multiple.

The competitive frame is louder than it looks

Everpure launched FlashBlade//EXA to go head-to-head. WEKA, last valued near $1.6 billion, still owns the POSIX/parallel-filesystem story for training throughput. NetApp trades at roughly $20 billion public cap. DDN is pushing Infinia onto NVIDIA's AI Data Platform partner roster, where VAST also sits.

Dell is the loudest. In a September 2025 engineering post, Dell's PowerScale product VP David Noy argued PowerScale meets NVIDIA Cloud Partner certification for 16,384 GPUs in 168 rack units, claiming VAST needs roughly 1.8x more rack space and Everpure nearly 5x more. Noy also pinned PowerScale at 41% less power than VAST and 72% less than Everpure, with four backend switches against VAST's and Everpure's 28 to 32. Everpure hit back on its own channel. In a November 2025 blog, Hari Kannan said FlashBlade//EXA testing ran "almost two times faster than our closest competitor (in less than half a rack of storage)." Everpure also told press that a 16,000-GPU Flashblade//EXA configuration fits in 37 rack units at roughly 40 kW.

Hammerspace attacks from a different angle: a Tier 0 model that turns GPU-local NVMe into shared storage and pitches VAST as bottlenecked by its network. Vendor roundups flag VAST's setup complexity and pricing as adoption friction.

Five vendors sharpening knives for a single private competitor is itself a signal. You don't get attacked this hard at $3 billion.

What to watch

Two things decide whether $30 billion prints.

First, neocloud concentration. The CoreWeave contract is a material share of VAST's bookings surge, and November reporting put CoreWeave past xAI as VAST's largest customer. If neocloud capex decelerates, or if AWS, Azure, and GCP build competitive data platforms in-house, the growth curve meets a direct test. The Google and Microsoft wins are the hedge, not the cure.

Second, flash supply. NAND shortages and price moves squeeze VAST's hardware-dependent deployments. The DASE thesis says shared flash pools beat siloed ones on economics, and all-flash TCO sits at the center of the DataStore pitch. That thesis gets harder when NAND pricing moves against you.

Then there's the IPO watch. Denworth said the company doesn't need the money; the cash-generation numbers back him up. The secondary tender gave early investors a clean exit window. Shapero is already CFO. The Series F reads as the last privately-negotiated benchmark before an S-1, and the lead choice (Drive Capital and Access Industries, not a crossover fund) says VAST wasn't forced to accept IPO-overhang terms to close.

When Denworth described the Series D in 2021 in Breaking the Mold, he said the cash would sit in the bank like the B and C rounds before it. Five years later, VAST pulled in another billion and a large share is going to people who already own shares. Not desperation. Not hype. A company with nothing to prove operationally setting a public-market reference price before it has to.

Whether $30 billion survives contact with an IPO tape is the next chapter. The middle layer of AI is the bet. The scoreboard is set.