VAST Data

1.1 Identity, Stage, and Product Scope

VAST Data is a private AI infrastructure company founded in 2016 that now presents itself as the “AI Operating System” for data-intensive applications rather than as a point-storage vendor. The core company story in retained sources is consistent on direction: VAST says it unifies storage, database, and compute in a single data platform, while independent coverage describes the same transition as a three-phase move from flash-oriented storage into database services and then into broader AI operating-system orchestration. Product materials describe VAST DataStore as a universal data store that removes traditional storage tiers and scales from terabytes to exabytes, while partner materials from Supermicro frame the platform as the data-management layer for enterprise AI and GPU cloud deployments. Headquarters is less cleanly disclosed than product scope. Official 2023–2026 press materials alternately use “No-Headquarters” or “Remote-First Company” while anchoring announcements to New York City, and Reuters-described coverage calls the company New York-headquartered. For later chapters, the safest ground truth is that VAST operates as a remote-first, New York-anchored private company with a US-facing commercial surface and Israeli operational roots reflected in later location reporting.[CO001, CO002, CO003, CO004, CO005, CO006]

1.2 Founders, Leadership, and Governance Signals

Renen Hallak is the clearest single point of leadership continuity across the retained record. He appears as founder and CEO across official financing releases, product positioning, and press interviews, making him the most important reusable identity anchor for later diligence. Jeff Denworth is also directly corroborated as a co-founder and remains a visible product spokesperson in both company and third-party materials, especially around the evolution from unified data platform to full AI operating system. Finance leadership is also now material: Amy Shapero, VAST’s first CFO and formerly Shopify’s CFO, is explicitly tied to IPO readiness in both CRN and Reuters-reported coverage. That combination suggests a governance stack that is maturing toward public-market discipline even though the company has not started a formal IPO process. Cloud leadership adds a more complicated signal. Jonsi Stefansson joined via the Red Stapler acqui-hire and was publicly cited by VAST during the Polaris launch, but the same executive remains tied to NetApp litigation risk. Public governance disclosure remains thinner than funding disclosure: the retained sources confirm a Fidelity board observer role, but they do not provide a clean, current full board list, which is a carry-forward diligence gap rather than a reason to invent governance detail.[CO009, CO010, CO011, CO012, CO013, CO014]

1.3 Funding History, Valuation, and Public Scale Metrics

VAST’s financing arc is unusually steep even by AI-infrastructure standards. The company went from an $83 million Series D at a $3.7 billion valuation in May 2021 to a $118 million Series E at a $9.1 billion valuation in December 2023, and then to a roughly $1 billion Series F transaction at a $30 billion valuation in April 2026. The Series F round matters for three reasons beyond price. First, it included both primary and secondary capital, so some of the round monetized existing holders rather than simply funding balance-sheet growth. Second, the syndicate now blends financial investors such as Drive Capital, Access Industries, Fidelity, and NEA with NVIDIA as a strategic participant. Third, the company paired the valuation step-up with unusually strong company-claimed operating metrics: more than $4 billion in cumulative bookings, more than $500 million in prior-year committed ARR, and positive operating margin and free cash flow. Those are powerful signals but still not substitutes for audited financial statements. Public scale disclosure is uneven. Late-2023 sources support more-than-700 employees and offices in the US, Israel, and London, while 2026 exact headcount, customer count, and any debt or credit instruments remain unsupported in the retained source pack.[CO016, CO017, CO018, CO019, CO020, CO021]

1.4 Milestones, Partnerships, and Adverse Context

The milestone record shows why later chapters should treat VAST as more than a storage vendor. After founding in 2016, the company spent its first phase establishing a high-performance storage base, then layered database and compute functions on top, and by 2026 was launching control-plane and agentic-computing features rather than only hardware-adjacent software. The partner map is also central to the story. VAST’s customer and channel footprint runs through AI clouds, OEMs, and strategic technology partners, with named relationships across CoreWeave, Google Cloud, Lambda, Core42, Supermicro, HPE, Cisco, and NVIDIA-linked deployments. That ecosystem expands reach, but it also creates dependence on partner routes and hyperscale AI-capex cycles. The main adverse item worth preserving is the Red Stapler/NetApp dispute. VAST was not named as a defendant, and NetApp’s Florida case was dismissed on venue grounds, but the episode still introduced reputational and integration risk precisely while VAST was raising another mega-round. For ground-truth purposes, the legal issue is not thesis-breaking on current evidence; it is a real diligence memory item that should stay live until related Icelandic or appeal processes are clearly resolved.[CO035, CO036, CO037, CO038, CO039, CO040]

1.5 Exhibits

2.1 Market boundary, adjacencies, and substitutes

VAST is not best framed as a generic AI-infrastructure company. Its own positioning and white paper place it in the shared data layer that sits between raw storage media and higher-level AI application software: a unified platform for file, object, block, tabular, vector, and streaming data that can serve training, inference, analytics, and HPC workloads from a common control plane. That means the core market boundary includes high-performance shared storage, metadata and vector retrieval, global namespace and governance, and the software needed to keep those data services available across on-prem, cloud, and edge environments. It does not include GPU compute, networking fabric, general-purpose application software, or large swaths of generic IT services spend. The practical competitive frame is therefore narrower than total AI infrastructure but broader than a traditional NAS array. VAST’s own solution pages push into adjacent analytics, compliance, backup, container, and data-lakehouse use cases, while the white paper describes replacing separate object stores, analytics clusters, and orchestration layers with one data-centric platform. Status-quo substitutes remain incumbent all-flash arrays, scale-out NAS, object stores, and multi-product stacks assembled from discrete storage, database, and orchestration tools. The market-definition question matters because using a broader AI-infrastructure headline as TAM would overstate the spend pool VAST can realistically capture, while using only classic external-storage categories understates the software and orchestration layer VAST is trying to sell.[CM001, CM003, CM004, CM005, CM007, CM008]

2.2 Sizing lenses and methodology conflict

Public market data supports a range of defensible lenses, not a single clean TAM. At the narrow end, IDC’s external OEM enterprise storage forecast reaches about $37.7 billion in 2026 and is explicitly tied to branded storage systems, with all-flash arrays benefiting from AI and analytics demand. Fortune Business Insights places AI-powered storage at $44.94 billion in 2026, a broader cut that includes storage architectures optimized for AI workloads and separates file- and object-based approaches. Research and Markets and Coherent Market Insights both put broad AI infrastructure at roughly $90 billion in 2026, but those reports bundle much more than storage, spanning offerings, functions, technologies, deployment models, and end users. Gartner’s $1.366 trillion AI-infrastructure number is broader still and is only useful as context for how much total ecosystem capex can surround VAST’s category. For valuation work, the correct conclusion is not that one of these numbers is right and the rest are wrong. The right conclusion is that VAST sits inside several nested spend pools, and each publisher chooses a different boundary. The most defensible public framing is an evidence-constrained core lens bounded by enterprise external storage on the low side and AI-powered-storage categories on the high side, with broader AI-infrastructure figures treated as adjacency rather than TAM. Public sources do not disclose enough VAST-specific pricing, customer, or win-rate data to isolate a clean SAM or SOM, so contradictory estimates and methodology limits should be preserved rather than averaged away.[CM009, CM011, CM013, CM017, CM018, CM019]

2.3 Buyer, user, payer, and adoption path

The buyer map is segment-dependent, but the pattern is consistent: technical champions emerge from infrastructure, data-platform, HPC, or AI-engineering teams, while budget approval and risk review move upward into CIO, procurement, finance, and broader buying committees. Market reports and vendor reference architectures show why. Fortune’s end-user cuts point to enterprises, CSPs, telecom, and government as distinct demand pools. Coherent’s on-prem bias for sensitive sectors, plus VAST’s own enterprise AI-factory framing, suggests regulated enterprises and public-sector or HPC users remain important for buyers that need control, governance, and data locality. In these accounts, storage administrators, ML-platform teams, or research-computing leaders may specify the architecture, but they rarely buy alone. IDC says procurement, finance, and revenue-operations functions have become central decision-makers, while Forrester says the typical complex purchase now involves 13 internal stakeholders and nine external influencers. That governance-heavy buying process also explains the adoption path. Trials and referenceability matter because buyers must prove business value before scaling. Forrester says more than 60% of buyers use a trial and 78% of buyers above $10 million do so. Supermicro and VAST sell a turnkey route from design to deployment, while VAST’s reference architecture emphasizes starting small and scaling with demand. The practical funnel is therefore problem discovery, architecture validation, proof of value, initial production cluster, and then broader rollout across more workloads or sites. What is still missing publicly is the actual conversion rate through that funnel for VAST accounts.[CM016, CM018, CM022, CM026, CM029, CM031]

2.4 Growth drivers, adoption constraints, and diligence gaps

Demand drivers are real and multi-layered. IDC’s storage forecast, VAST’s own AI-factory messaging, and the reference architecture all point to a common theme: more AI production workloads mean more shared data infrastructure, more all-flash demand, and higher sensitivity to time-to-token and GPU utilization. Deloitte shows enterprise AI activity moving from pilot to production, while Cloudera describes 2026 as the shift from experimentation to operational intelligence orchestration. NetApp and Pure both underline that incumbents are already monetizing the same need for AI-ready, accessible enterprise data. In other words, VAST is selling into a real category, not a purely theoretical one. But the constraints are equally material. Gartner says enterprises remain ROI-sensitive and often prefer incumbent providers during the current AI cycle. Informatica and Deloitte both show governance, data quality, and skills gaps slowing scale-out. Electronic Design and SNIA treat storage as a first-order design constraint because it affects performance, power, cooling, and deployment risk, while Avnet shows that SSD and DRAM tightness can push up cost and extend supply risk through 2027. INFUSE adds a buyer-trust problem: more information does not necessarily produce more purchase confidence. The combination means VAST’s valuation should not assume frictionless adoption. The biggest unresolved diligence questions are company-specific rather than market-wide: pricing, customer count, workload mix, trial-to-production conversion, and how much of the broad AI capex wave actually lands in the shared data layer VAST serves.[CM010, CM012, CM024, CM028, CM034, CM036]

2.5 Exhibits

3.1 Landscape, Peer Classes, and Substitute Paths

VAST’s competitive field is wider than “other flash array companies.” Coldago’s 2025 file-storage maps and StorageNewsletter’s methodology summary place VAST in overlapping enterprise and high-performance storage frames with Pure, NetApp, IBM, DDN, Qumulo, Hammerspace, and WEKA. The direct peer set is therefore AI-first file/object or high-performance data-platform vendors, not just one vendor with a similar appliance shape. The next ring is public-storage incumbents. Pure and NetApp now sell consumption-oriented commercial models on top of broad product portfolios, while IBM still positions Storage Scale as a software-defined AI/HPC and hybrid-data platform. The adjacent ring matters too: Hammerspace and Qumulo push orchestration, global namespace, and hybrid-cloud mobility, which can let a buyer de-emphasize VAST even without copying its DASE architecture. Substitutes are real rather than theoretical. AWS FSx for Lustre, S3, Azure Managed Lustre, and Google Filestore expose usage-based pricing, native billing, and existing cloud contracts. Internal-build and status-quo options also persist because Ceph and MinIO can unify key storage semantics without handing more leverage to a single proprietary vendor. The net result is a market where VAST sells into direct product rivalry, incumbent bundling, orchestration alternatives, hyperscaler-native substitutes, and open-source anti-lock-in narratives at the same time.[CP001, CP002, CP003, CP006, CP011, CP015]

3.2 Competitor Profiles and Capability Tradeoffs

The retained profiles show three distinct ways to attack the same budget. Pure and NetApp compete from the incumbent end by combining multiprotocol storage breadth with packaging that looks safer to procurement teams already standardizing on public-company vendors. WEKA, DDN, and IBM compete by arguing that AI performance, GPU efficiency, or scale-out data services matter more than appliance familiarity. Hammerspace and Qumulo instead lean into orchestration and data locality: Hammerspace’s Tier 0 story is that local NVMe plus shared access beats external storage bottlenecks, while Qumulo emphasizes global namespace, cloud-native deployment modes, and managed Azure routes. VAST’s public story is broader than raw storage hardware because it now claims an AI operating system spanning storage, database, and compute. That broader story is helpful, but the competitive downside is obvious: once VAST says it is the data and execution substrate for agentic AI, it is no longer compared only with filers. It is compared with every vendor claiming unified data services, every orchestrator promising data-local execution, and every cloud substitute offering transparent metering plus existing contracts. Capability therefore matters on two axes at once: the depth of the hot data plane for AI workloads, and the breadth of workflow, namespace, deployment, and procurement options around it.[CP001, CP003, CP006, CP008, CP011, CP013]

3.3 Pricing Transparency, Switching Friction, and Distribution Power

Pricing visibility is one of the clearest structural differences in the corpus. AWS, Azure, and Google all publish explicit or calculator-driven usage constructs for file or object services, while specialist storage vendors mostly route buyers to sales, demos, or negotiated subscription constructs. That does not make hyperscaler-native substitutes cheaper in every real workload, but it does lower procurement friction and makes them easier to compare inside an existing cloud contract. Pure and NetApp narrow the gap further by wrapping their platforms in STaaS models rather than only capex appliance sales. VAST’s switching costs still exist because petabyte-scale datasets, tuned AI pipelines, and operational familiarity are sticky. But the evidence suggests those costs are softer than monopoly-style lock-in: buyers can multi-home across cloud object, managed filesystems, orchestration layers, and open-source tiers, and several alternatives sell explicit anti-lock-in or portability stories. Distribution magnifies the point. Public incumbents and hyperscalers already have large account control, while WEKA and Hammerspace are still investing to expand sales or partnerships. VAST therefore needs more than speed claims to stay differentiated; it needs the product, migration path, and commercial model to remain convincing once buyers compare them against cloud billing convenience and incumbent account leverage.[CP004, CP007, CP018, CP021, CP022, CP023]

3.4 Moat Durability, Multi-Homing, and Adverse Read-Throughs

VAST’s moat is real, but the retained evidence says it is conditional rather than absolute. Coldago still treats VAST as a leader, and VAST’s own scale narrative now extends to a $30 billion valuation and an AI-operating-system pitch. That supports the bull case: a vendor with credible scale, strong AI positioning, and a shared data plane can consolidate storage, metadata, and parts of the AI workflow. The adverse evidence matters because it attacks the exact same thesis. theCUBE argues that VAST’s AI-OS ambition is ahead of current platform maturity, while StorageMath says VAST’s data-reduction narrative overstates reality and nudges customers toward deeper ecosystem lock-in. Hammerspace’s competitive brief attacks VAST on Tier 0 and cloud-scale orchestration. Meanwhile, Pure and NetApp are not standing still; they already pair technical breadth with safer commercial packaging. Cloud substitutes keep pressuring the cost and convenience end of the decision, and internal-build options keep the anti-lock-in alternative alive. The realistic read is that VAST’s moat depends on staying meaningfully better on hot AI data performance and integrated workflow value while preventing buyers from deciding that incumbent packaging, orchestration overlays, or cloud-native substitutes are “good enough.” Multi-homing is the key release valve: if customers can keep VAST only for the hottest tier and solve the rest elsewhere, moat durability weakens.[CP018, CP027, CP028, CP029, CP030, CP031]

3.5 Exhibits

4.1 Revenue model, pricing, and recognition issues

VAST's monetization is best described as infrastructure software sold through enterprise and partner channels rather than as a pure self-serve storage product. The company markets a unified AI Operating System that combines storage, database, and compute, and older company material explicitly says Gemini is a software consumption model that disaggregates hardware from software economics. Public commercial pages emphasize contact-sales and demo motions, not self-serve pricing. That means there is enough evidence to identify the revenue mechanism, but not enough to observe realized price cards or discounting behavior directly. Cisco's Global Price List provides one verified partner procurement path, while Polaris being included at no extra cost suggests at least part of the control plane is bundled into the platform. The CoreWeave agreement demonstrates that VAST can land very large partner-linked commitments, but the public record still does not bridge how bookings, committed recurring revenue, and any usage or system-delivery components are recognized into a single revenue-definition stack.[CI001, CI002, CI003, CI004, CI005, CI006]

4.2 GTM motion and sales-efficiency proxies

VAST's public GTM evidence points to a high-touch enterprise motion with unusually large average land sizes and strong expansion dynamics. In 2021 the company said initial investments averaged about $1 million and expansion averaged 328 percent; in FY22 it said NRR exceeded 300 percent. More recent third-party reporting suggests top-100 new customers spend more than $1.2 million on average and that contract duration often runs five to seven years. Those are strong proxies for durable enterprise revenue, especially when paired with Cisco GPL access and cloud-partner distribution into AI infrastructure budgets. The company also appears to benefit from giant lighthouse accounts: CoreWeave alone signed a $1.17 billion agreement and multiple outlets reference xAI and other AI-cloud customers. What the public record does not disclose is CAC, CAC payback, the average sales cycle, or partner take-rates. So the chapter can argue that VAST likely has very efficient land-and-expand dynamics, but it cannot convert those demand signals into a closed sales-efficiency model.[CI006, CI007, CI008, CI017, CI018, CI024]

4.3 Cost structure, gross margin drivers, and capital intensity

The strongest public margin signals point toward a software-led cost structure even though VAST sells into a hardware-heavy market. Company disclosures from 2022 and 2023 described the business as built on software gross margins and later cited nearly 90 percent gross margin with positive cash flow. The architectural and commercial model matters here: Gemini separates hardware and software economics, the white paper says OEMs such as Cisco and SuperMicro can sell standard-server VAST clusters, and Cisco now carries the platform on its GPL. Those factors imply that some inventory, manufacturing, and balance-sheet working-capital burden may sit with OEM partners rather than with VAST itself. At the same time, bundled features such as Polaris and installed-base programs such as Amplify mean that support, control-plane delivery, and optimization services are embedded in the platform economics but not separately disclosed. The result is a favorable-looking gross-margin path with lower apparent capital intensity than a proprietary-array vendor, but one that still lacks a current segment-margin bridge and working-capital disclosure.[CI003, CI019, CI020, CI021, CI022, CI023]

4.4 Public traction and capital adequacy

Public traction is unusually strong for a private infrastructure company, but capital adequacy still has to be read through partial disclosures. Official sources show a staircase from $200 million ARR at the end of 2023 to more than $500 million of committed ARR at the exit of the prior fiscal year, alongside $4 billion of cumulative bookings, positive operating margin, and free cash flow. Independent outlets stretch that picture much further, with Reuters-linked reporting discussing $600 million of ARR and Calcalist/TNW discussing up to roughly $2 billion of ARR-like revenue including non-committed components. That gap is not just noise; it tells investors that VAST's publicly quoted revenue metrics are using different definitions. On capital adequacy, the 2026 Series F valued the company at $30 billion and brought in about $1 billion of total transaction value, but public reporting indicates that a large share of that headline number was secondary liquidity. Fidelity's SEC marks independently corroborate the valuation step-up, yet the company still has not publicly disclosed cash on hand, burn, runway, debt, or financing triggers.[CI010, CI011, CI012, CI014, CI015, CI016]

4.5 Financial verdict and diligence blockers

The financial read-through is favorable but not fully underwriteable. On the positive side, VAST combines software-like margin signals, multi-year contract evidence, very large average commitments, positive free cash flow claims, and exceptional demand from AI-cloud customers. That is a stronger revenue-quality profile than most late-stage infrastructure startups show publicly. The caution is that several of the most important inputs remain private: customer concentration, actual cash balance and runway, CAC and payback, realized pricing, debt obligations, and a reconciled bridge across bookings, CARR, and any broader ARR or revenue constructs. Large AI-cloud accounts also introduce concentration and renewal-timing risk precisely because they can move results so dramatically. The right underwriting stance is therefore not to dismiss the business model, but to separate what the public record can support from what still needs company disclosure. VAST looks financially strong enough to justify serious diligence, yet the evidence set still stops short of a closing-grade financial model.[CI030, CI036, CI037, CI038, CI039, CI040]

4.6 Exhibits

5.1 Product scope in customer workflow terms

VAST sells the platform as an AI Operating System rather than as a single storage SKU. In customer workflow terms, the retained record says the buyer is trying to collapse multiple stages of the data path onto one shared substrate: ingest data through file, object, streaming, or Kubernetes interfaces; catalog and query it through DataBase; trigger event-driven processing in DataEngine; and keep the same data visible across sites through DataSpace rather than copying it into separate silos. The AI reference architecture is explicit that the company wants one flash-native platform to cover data capture, refinement, model training, model serving, audit logging, and RAG-style retrieval. Kubernetes material makes that concrete at operator level, because the same cluster can be exposed either as NFS or as CSI-provisioned persistent volumes. The workflow value proposition is therefore not merely speed; it is reducing data movement, avoiding separate storage/database/streaming stacks, and presenting one control model to infrastructure teams.[CE001, CE002, CE003, CE004, CE005, CE007]

5.2 Architecture and operating model

The architectural center of gravity is DASE, VAST's disaggregated shared-everything model. Public technical documents describe stateless compute nodes handling protocol and software services while persistent data and system state sit on NVMe-backed storage enclosures, which allows performance and capacity to scale independently. That matters because the platform pitch depends on all protocols seeing the same underlying data plane: DataStore exposes file and object interfaces, DataBase adds tabular and query services, Event Broker extends the platform toward Kafka-style streaming, and block access is now being added through NVMe/TCP. The architecture is also presented as deployment-flexible. The older reference architecture still shows classic CBox and DBox layouts, while current partner material from Supermicro and Cisco shows packaged solutions that keep VAST's software abstractions but adapt them to OEM hardware and AI-factory reference designs. The net read-through is that VAST's product is not a loose bundle of separate products; it is one layered software platform whose value depends on the shared data plane actually behaving consistently across multiple access methods and node topologies.[CE003, CE014, CE015, CE016, CE017, CE018]

5.3 Deployment, integration, and roadmap maturity

VAST has enough public integration surface to look operable in modern infrastructure environments rather than merely demoable. The Juniper validated design exposes concrete cluster topology, VIP rebalancing behavior, GUI and CLI management, and a REST API under the hood. The Kubernetes knowledge-base article plus the CSI repo and Helm documentation show that container platforms are not an afterthought. The GitHub organization, Python SDK, Terraform provider, DataEngine CLI, and Go client together create a public operator toolchain that can be used for provisioning, querying, automation, and pipeline management. The roadmap, however, is moving faster than the deepest public documentation. Block storage and Event Broker were surfaced in 2025, and the 2026 narrative adds Polaris, CNode-X, PolicyEngine, TuningEngine, and open DataEngine blueprints. Those moves expand the buyer story from storage into orchestration and agentic-AI operations, but they also mean the newest layers are backed more by launch materials and commentary than by the kind of mature administrator documentation visible for the core platform and Kubernetes integrations.[CE009, CE010, CE011, CE012, CE013, CE020]

5.4 Differentiation, dependencies, and technical risks

The supportable differentiation thesis is stronger on shared data-plane breadth than on every new AI-OS layer being fully mature today. VAST does have a real product distinction in the retained corpus: one platform claims multiprotocol file, object, table, streaming, and now block access; DASE is repeatedly positioned as the reason the platform can scale without the classical trade-offs of shared-nothing or tiered storage; and the NVIDIA collaboration brings GPU acceleration into SQL, vector search, retrieval, and long-context inference paths. That said, the same evidence surfaces meaningful dependencies. Accelerated deployments depend on NVIDIA libraries, DPUs, and networking assumptions; Cisco and Supermicro are explicit go-to-market and hardware routes for CNode-X; and the newest governance and tuning layers still read as announced capabilities rather than widely evidenced production defaults. The product therefore looks technologically differentiated, but the risk register sits at the seams: partner hardware availability, execution on forward-looking agentic-AI features, and the challenge of making a much broader AI-OS claim feel as mature as the underlying storage core.[CE016, CE018, CE021, CE025, CE026, CE027]

5.5 Trust, compliance, and control posture

Public trust evidence is real but uneven. The strongest artifacts are not a polished public trust-center package; they are deployment and hardening materials. VAST's Security Configuration Guide says the platform is a STIG-hardened appliance, documents federal procurement orientation, and states the product is listed on the DoDIN Approved Products List. It also lays out concrete control surfaces: HTTPS-based VMS management, CLI access, predefined roles including CSI-specific permissions, identity integration with Active Directory, LDAP, and NIS, and trusted TLS prerequisites before SSO and MFA are configured. The AI reference architecture adds the broader design claim of zero trust, RBAC, ABAC, encryption, auditing, network isolation, and tenant-specific key management. Commvault's integration guide adds cyber-resilience evidence by documenting WORM-mode immutable backup copies on VAST-backed cloud-library workflows. The caution is that this chapter's retained corpus still does not provide a broad public certification catalog, clear SLA or incident-history evidence, or strong public reference proof that the newest Polaris and Policy/Tuning layers are already widely deployed in production.[CE032, CE033, CE034, CE035, CE036, CE037]

5.6 Exhibits

6.1 Customer segmentation and buyer / user / payer map

The public record supports a broad but not fully enumerated VAST customer base. Official customer-proof pages and roster disclosures show VAST reaching AI-cloud operators, telecom sovereign-AI builders, financial-services institutions, healthcare and life-sciences teams, government research bodies, and media organizations. The visible buyer-user-payer pattern is segment specific rather than uniform. In media, the NHL frames the purchase around production and archive workflows run by media-operations teams for downstream editors, producers, broadcasters, and eventually AI-assisted content discovery. In sovereign AI and AI-cloud contexts, the technical operator is a platform or infrastructure team, but the platform is procured so enterprise, research, and public-sector tenants can consume GPU and data services. In healthcare, genomics, and quant trading, the end users are researchers, scientists, and quants, while central infrastructure budgets fund the platform. That means VAST is not selling to a single universal persona; it is selling a deep infrastructure layer whose commercial sponsor is usually central IT or platform engineering, while the economic justification comes from higher-level data teams and business workflows.[CU001, CU002, CU003, CU004, CU024, CU037]

6.2 Production adoption trajectory and named customer proof

The strongest proof is where VAST or the customer describes a live deployment with scope, chronology, or measurable outcomes. The NHL relationship is the clearest long-duration example: VAST says the league first adopted the platform about six years ago, moved more than 20 PB of archive content onto it, extended the footprint to all 32 arenas plus headquarters, and in 2026 successfully piloted a portable node for outdoor events that securely replicated more than 2 TB of footage. SK Telecom provides the clearest sovereign-AI and telco example, with VAST integrated into Petasus AI Cloud so GPU environments can be provisioned in about 5–10 minutes and scaled across a 1,000-Blackwell-GPU stack. CoreWeave is the highest-stakes AI-cloud example: VAST announced a $1.17 billion commercial agreement in late 2025, and both official and third-party sources describe VAST as a key data foundation inside CoreWeave’s cloud. Beyond those flagship cases, the retained set includes quantified healthcare outcomes from Invitae and PacBio and title-level case pages across HSBC, CACEIS, Jump Trading, HHS, LLNL, and Chan Zuckerberg Initiative, which together show real vertical spread even when not every page has the same evidentiary depth.[CU005, CU006, CU007, CU008, CU009, CU010]

6.3 Retention, repeat usage, and satisfaction signals

Durability is supportable, but only indirectly. VAST’s own 2021 Series D release remains the cleanest historical expansion datapoint: average initial customer investment of about $1 million, average net-revenue expansion of 328%, and several customers already above $10 million of spend. More recent public proof shows continued repeat usage and expansion in named accounts rather than disclosed portfolio-wide retention metrics. The NHL relationship expanded from archive modernization into real-time arena-to-headquarters workflows and then into a 2026 outdoor-event pilot. CoreWeave evolved from an existing partnership into a disclosed $1.17 billion agreement. PacBio explicitly says it added another 2 PB to its VAST cluster with no issues, and Invitae cites 6x faster data access plus 30x IOPS improvement. Independent satisfaction evidence is directionally positive but thin: FeaturedCustomers aggregates a large body of testimonials and reference ratings, while PeerSpot reports 5.0/5 and 100% willingness to recommend from a small sample. The caution is that no retained source discloses current GRR, NRR, churn, cohort renewal, or top-customer share, so the public durability case is positive but incomplete.[CU020, CU021, CU022, CU029, CU030, CU031]

6.4 Expansion, concentration, channel dependence, and unresolved gaps

The same public evidence that proves adoption also sharpens risk. CoreWeave is simultaneously VAST’s best public expansion case and clearest concentration flag because one disclosed partner-linked agreement is worth $1.17 billion, and third-party reporting says VAST underpins workloads CoreWeave delivers to Meta, OpenAI, and Microsoft. Public VAST materials also frame adoption through partners such as NVIDIA, Supermicro, HPE, Lambda, and Core42, which suggests partner-led distribution matters structurally for customer acquisition and deployment. What public sources do not show is equally important: there is no retained disclosure of current total customer count, channel mix, sales-cycle length, partner take-rates, cohort retention, or top-customer revenue share. The user-specified names BMW, Nasdaq, Goldman Sachs, and Dell are also revealing. The retained NVIDIA pages on BMW and Nasdaq prove adjacent AI infrastructure credibility, but they do not mention VAST. Sacra supports Goldman Sachs and Dell Technologies Capital as investors, not paying customers. So the chapter can support real adoption, real expansion, and real partner leverage, but it cannot close the loop on direct proof for every marquee name or on the concentration math needed for underwriting-grade customer durability.[CU018, CU019, CU033, CU034, CU035, CU036]

6.5 Exhibits

7.1 Ranked risk stack and investment implication

VAST is not a scandal or obvious product-failure case; the risk is mostly underwriting risk at a peak-cycle AI infrastructure price before public disclosure catches up. The company does have real proof points: it closed a $30 billion Series F, says it has more than $4 billion of cumulative bookings and more than $500 million of committed ARR, landed a $1.17 billion CoreWeave partnership, and now ships through validated NVIDIA, Cisco, and Supermicro architectures. But those same facts create the top residual stack. First is valuation-reset risk, because the price assumes durable AI build-out and continued acceptance of private metrics. Second is concentration risk across Nvidia, Cisco, OEM hardware, and a small set of marquee AI-cloud customers. Third is execution breadth as VAST stretches from storage into a full AI operating system and cloud control plane. Fourth is legal and compliance exposure around Red Stapler and new data rules. Fifth is Israel-linked continuity and talent exposure. Investment implication: the deal is still diligencable, but only if legal provenance, concentration, and compliance evidence become bounded rather than open-ended risks.[CR001, CR002, CR005, CR006, CR007, CR011]

7.2 Legal, regulatory, privacy, and sovereignty risk

The clearest live legal overhang is the Red Stapler chain. NetApp sued former CTO Jonsi Stefansson over alleged trade-secret misuse tied to the startup VAST acquired, and although the Florida complaint was dismissed without prejudice, the dismissal was venue-based and NetApp appealed. That means VAST is not yet facing a proven liability event, but the matter is also not fully extinguished. Separately, the public legal surface shows a company that processes contact, device, and candidate data through a website and cloud stack while pushing deeper into regulated sectors through Cisco reference designs for healthcare and financial-services use cases. That makes the 2025 bulk-data regime and broader U.S. privacy enforcement more relevant than they would be for a simple storage vendor. Israel adds another layer. VAST publicly tied itself to a sovereign Israeli AI cloud and operates a large Tel Aviv-centered development footprint, while Israeli legal commentary highlights export-control, dual-use, and IIA-grant constraints that can matter in financings, IP transfers, and exits. Public evidence is enough to rank these as real legal and regulatory risks, but not enough to close them without document review.[CR021, CR022, CR023, CR024, CR025, CR026]

7.3 Operational, quality, and dependency risk

Operational risk is less about a publicly documented outage history than about how many critical layers VAST does not fully control. The current product story relies on NVIDIA acceleration, Cisco distribution, Supermicro reference systems, and broader OEM hardware from vendors that also work with VAST competitors. The CoreWeave contract and SDS sovereign-cloud build prove that VAST can win very large deployments, but they also show how revenue and roadmap execution can become tied to a handful of capital-intensive counterparties whose own GPU budgets, data-center timing, and procurement priorities can shift abruptly. Competitive pressure sharpens this problem. Dell is not a partner in the evidence reviewed; it is an active rival publicly attacking VAST on power, space, and benchmark framing, while StorageMath has criticized one of VAST's optimization claims. That does not prove product weakness, but it does mean VAST must keep both technical credibility and partner alignment high while scaling across mission-critical environments.[CR007, CR008, CR009, CR011, CR012, CR013]

7.4 Financial, commercial, and competitive risk

The financial-model risk is not an imminent cash crunch in the public record; it is valuation fragility and concentration fragility. VAST's financing access is clearly strong, but the round structure mixed primary and secondary capital, so the headline size exaggerates fresh operating cash. At the same time, the strongest scale metrics remain company-claimed rather than publicly reconciled. That matters because macro context is no longer uniformly supportive. Gartner still forecasts huge AI infrastructure spend in 2026, but it also says the sector is in a trough of disillusionment and that incumbents will capture much of enterprise AI demand. Goldman frames the build-out debate as assumption-sensitive, not guaranteed. Those two facts can coexist: spend can grow while the market becomes less forgiving of premium private valuations. For VAST, that means the downside case is not just slower top-line growth; it is slower top-line growth combined with harder competition from Dell and other incumbents and a public-market multiple reset before VAST closes its disclosure gap.[CR001, CR002, CR004, CR005, CR006, CR018]

7.5 People, execution, mitigations, and kill criteria

People risk is meaningful because VAST's market narrative still runs heavily through founder CEO Renen Hallak, while the cloud-control-plane expansion vector depends on the Red Stapler team whose arrival is precisely the area touched by the NetApp dispute. There are also classic late-stage private-company diligence limits: the public source set still does not provide top-customer concentration, renewal schedules, debt, current security certifications, or a clean treasury view. The good news is that some mitigations are real. VAST has deep financing access, partner validation from Cisco and Supermicro, and credible adoption in large AI environments. The bad news is that those mitigations do not answer the most thesis-sensitive questions on their own. Investors should therefore treat several items as explicit post-signing monitors rather than vague watchpoints: any new litigation step that drags VAST directly into the Red Stapler fight, evidence of customer concentration worse than expected, inability to produce current compliance attestations for regulated workloads, or any financing event that implies a valuation reset. Those are clearer kill criteria than generic concerns about competition or macro noise.[CR022, CR037, CR038, CR039, CR042, CR048]

7.6 Exhibits

8.1 Price-sensitive recommendation, not a company-quality verdict

VAST Data looks like a real category leader rather than a hype-only asset. The company has an unusually strong combination of AI-infrastructure relevance, marquee customers, rapid growth, and positive margin signals for a private storage-and-data platform. The problem is not whether the business matters; the problem is whether fresh capital at a $30 billion valuation can still earn an attractive return using only public evidence. At the official floor of more than $500 million of committed ARR, the round price implies roughly 60x. Even the older Reuters-style $600 million ARR projection still implies about 50x. Only the most aggressive external estimate of roughly $2 billion of ARR pulls the entry price down to about 15x, which is still rich versus public storage peers. That makes the recommendation price-sensitive and evidence-sensitive: track or research more for new money, hold only if the basis is materially lower, and do not confuse a strong company with a justified entry price.[CV001, CV007, CV009, CV019, CV020, CV021]

8.2 Financing context, marks, and entry discipline

The financing record is directionally supportive but still incomplete where underwriting matters most. March 2026 reporting surfaced the raise early, while the company officially disclosed the closed Series F on April 22, 2026. Management confirmed that the roughly $1 billion transaction combined primary and secondary capital, and outside reporting suggests that liquidity for early holders may have represented around half the headline amount. That matters because a secondary-heavy round is a weaker signal on fresh balance-sheet strength than a pure primary raise. Fidelity portfolio filings show an independent mark-up in VAST preferred shares from roughly $18.16 per share in August 2024 to roughly $59.22 by February 2026, so the valuation step-up did not come from nowhere. But those same sources still do not disclose the exact preference stack, dilution, or revenue-definition bridge. Entry discipline therefore has to be simple: assume the public record supports a valuation below the round unless diligence proves that recurring revenue is much closer to the high-end estimates and that late-stage financing terms are clean.[CV004, CV005, CV006, CV017, CV018, CV019]

8.3 Scenarios and comparable valuation anchors

Public comparables frame the valuation problem clearly. Everpure, the rebranded Pure Storage, trades around 7.7x EV to trailing revenue; NetApp trades around 4.1x; and CoreWeave, the clearest public AI-infrastructure premium reference, trades around 10.9x EV to annualized first-quarter revenue while carrying much heavier leverage. Private AI infrastructure rounds such as Lambda, Crusoe, and Together AI confirm that investors are paying up for GPU-cloud and AI-platform assets, but none of those milestones close the gap between VAST's official $500 million-plus CARR floor and a $30 billion price. The base case therefore sits below the round, because a mid-point revenue estimate combined with a public AI-infrastructure premium still lands in the low-to-high teens. The bull case reaches the high twenties to mid-thirties only if the roughly $2 billion ARR estimate is directionally right, concentration is manageable, and public markets remain open to premium AI infrastructure multiples. The bear case collapses quickly toward public-storage bands if AI spending slows or the revenue bridge fails.[CV024, CV027, CV030, CV031, CV032, CV033]

8.4 Exit readiness, thesis-break triggers, and final diligence asks

The most supportable exit path is an eventual IPO, but management itself says that VAST has not started a formal IPO process or hired bankers yet. That means investors should not model a near-term public exit as though it were already underway. The public record also leaves open the questions that most often break late-stage deals: what fraction of revenue comes from a handful of hyperscaler or neocloud accounts, how repeatable those contracts are, what the post-Series-F cap table looks like, and whether audited reporting systems are ready for public-market scrutiny. Those are not cosmetic details at this valuation. If ARR definitions fail to reconcile, if customer concentration is high, or if the company remains far from IPO readiness despite demanding public-style pricing, the thesis should break immediately. The diligence agenda is therefore straightforward: ask for the private metric bridge, the concentration schedule, the financing waterfall, and the IPO-readiness roadmap before treating the $30 billion round as investable rather than merely impressive.[CV039, CV040, CV045, CV046, CV048, CV049]

8.5 Exhibits