VAST Data

1.1 Corporate Identity, Genesis, and Operating Model

VAST Data was established in 2016 by Renen Hallak and Shachar Fienblit to address the critical input/output bottlenecks emerging from modern deep learning and distributed data systems. The enterprise operates under a remote-first organizational structure with official global headquarters located in New York City, supported by major specialized research and development centers in Tel Aviv and Haifa, Israel. Unlike conventional enterprise storage vendors that silo file, block, and object access into disparate hardware appliances, VAST Data engineered its platform around a breakthrough Disaggregated Shared-Everything (DASE) software architecture. The DASE model unifies unstructured file and object storage, high-throughput tabular databases, and containerized compute engine capabilities into a single software-defined substrate termed the VAST AI Operating System. This architecture decouples stateless compute logic from globally accessible, shared non-volatile flash media, enabling thousands of graphics processing units (GPUs) and AI accelerator nodes to simultaneously query and ingest petabyte-scale datasets without lock contention, latency spikes, or filesystem fragmentation. The company's recent technology extensions, including confidential AI runtimes and the Polaris global control plane developed in close alignment with NVIDIA's computing ecosystem, target large-scale frontier model training laboratories, national research centers, and enterprise AI factories requiring continuous real-time data access.[CO001, CO010, CO011, CO014]

1.2 Executive Leadership, Governance, and Human Capital

VAST Data's strategic direction and architectural execution remain anchored by its technical co-founders. Founder and Chief Executive Officer Renen Hallak previously led core engineering teams at XtremIO prior to its acquisition by EMC, bringing extensive experience in high-performance flash-optimized storage systems to VAST's commercial vision. Co-founder and Chief R&D Officer Shachar Fienblit directs the company's extensive engineering apparatus across its Israeli development centers, overseeing core distributed systems software development, hardware certification, and protocol optimizations. Co-founder and Chief Marketing Officer Jeff Denworth leads product strategy and ecosystem positioning, serving as the primary architectural evangelist across major technology conferences. In 2024, the company appointed former Shopify Chief Financial Officer Amy Shapero as its CFO, a critical governance milestone that significantly bolstered corporate reporting, internal fiscal discipline, and audit readiness ahead of potential initial public offering filings. Total organizational headcount has scaled to approximately 1,100 employees globally, with roughly one-third concentrated in Israeli engineering hubs and the remainder distributed across global enterprise sales, technical support, and partnership engineering. Key-person dependency remains high around CEO Renen Hallak, whose architectural vision and executive relationships with strategic investors such as NVIDIA remain central to company momentum.[CO001, CO009, CO010, CO011]

1.3 Capitalization Trajectory, Valuation, and Investor Syndicate

VAST Data's institutional capitalization reflects an exceptional valuation trajectory driven by accelerating demand for generative AI infrastructure. In May 2021, the company announced an $83 million Series D funding round led by Tiger Global Management at a post-money valuation of $3.7 billion, tripling its valuation within twelve months on the back of rapid customer adoption. By December 2023, VAST completed a Series E financing round that nearly tripled its valuation again to $9.1 billion while confirming $200 million in ARR and over three years of sustained cash-flow positivity. In April 2026, VAST closed a landmark Series F financing that valued the company at $30 billion, more than tripling its Series E benchmark and establishing VAST as the highest-valued private software enterprise originating from the Israeli tech ecosystem. The Series F transaction totaled approximately $1 billion, comprising new primary growth capital alongside more than $500 million in secondary share liquidity for early employees and seed investors, transacted at an estimated 15% discount to headline valuation. The round was led by Drive Capital with Access Industries serving as co-lead, alongside continued participation from strategic investor NVIDIA, Fidelity Management & Research, and New Enterprise Associates. Prior to Series F, VAST had raised nearly $400 million in cumulative equity capital, demonstrating capital efficiency uncommon among mega-cap enterprise infrastructure unicorns.[CO002, CO003, CO004, CO005, CO006]

1.4 Commercial Scale, Ecosystem Alliances, and Counterparty Risks

VAST Data has expanded its commercial footprint rapidly across artificial intelligence model builders, enterprise hyperscalers, and sovereign computing initiatives. The company reported surpassing $4 billion in cumulative contract bookings and exiting its previous fiscal year with more than $500 million in Committed Annual Recurring Revenue (CARR), while sustaining positive operating margins and free cash flow. In contrast to company-disclosed Committed ARR, third-party press reporting from Calcalist in March 2026 estimated annualized run-rate figures approaching $2 billion by year-end 2025, highlighting a material reporting variance that requires rigorous audit verification across multi-year booking schedules versus ratable revenue recognition. A central pillar of VAST's commercial growth is its landmark multi-year partnership with neocloud leader CoreWeave, valued at $1.17 billion, which embeds VAST's storage layer into thousands of GPU clusters powering client workloads including Mistral AI and xAI. However, secondary market research from EquityZen underscores substantial counterparty concentration risks tied to this reliance on specialized AI hyperscalers; any capital expenditure moderation or architectural shift by CoreWeave could adversely disrupt projected bookings. Concurrently, legacy hardware incumbents including Dell Technologies have escalated direct competitive counter-positioning with 2025 PowerScale solutions, and platform encroachment remains a long-term risk if primary partner NVIDIA expands internal software storage capabilities.[CO007, CO008, CO012, CO013, CO014]

1.5 Exhibits

2.1 Market Boundary, Workload Scope, and Architectural Shifts

The enterprise storage and data management market is experiencing a structural architectural transformation driven by the scaling demands of accelerated artificial intelligence computing. Legacy storage infrastructures historically relied on dual-controller storage area networks (SAN) and network-attached storage (NAS) designed around hierarchical tiering, where expensive high-speed flash was strictly isolated from high-capacity hard disk drives. Under generative AI training and large-scale inference workloads, this legacy model fails: thousands of GPUs running parallel checkpointing and token ingestion rapidly saturate traditional storage controllers, leaving multi-million-dollar accelerator clusters idle during data fetch cycles. VAST Data defines its addressable market boundary around the AI Operating System layer, utilizing a Disaggregated Shared Everything (DASE) architecture that physically separates stateless compute nodes (CNodes) from high-density flash storage media (DBoxes) over an NVMe-over-Fabrics network fabric. Included spend within VAST's core market comprises high-throughput flash file and object storage, real-time vector databases, embedded runtime task engines, and hybrid-cloud data fabric software. Conversely, commodity bulk backup for legacy virtual machines, physical tape archiving, and commodity enterprise file shares that do not interact with accelerated compute fall outside the primary target spend boundary.[CM001, CM002, CM003, CM004]

2.2 Addressable Market Sizing, Growth Drivers, and Sizing Lenses

Evaluating the addressable opportunity for modern AI data platforms requires balancing expansive top-down macroeconomic forecasts with evidence-constrained commercial realities. Industry forecasters and ARK Invest have posited that the market capitalization of AI hardware and software vendors could scale at a 50% annualized pace toward $87 trillion by 2030, with data platforms representing an addressable market exceeding $400 billion over the next decade. More immediately, Tech Insider identified a $76 billion wave in AI data infrastructure spending powering large-scale training and inference deployments. Concurrently, independent research from Manhattan Venture Partners projected VAST Data's annual software revenues between $540 million and $700 million in 2025, charting a five-year trajectory toward $2.49 billion to $3.35 billion by 2030. In practice, the serviceable addressable market (SAM) is bounded by organizations operating high-density GPU clusters and enterprise datasets exceeding 10 petabytes, such as specialized neoclouds, national research supercomputers, and Fortune 500 AI labs. VAST Data's commercial disclosures—exceeding $500 million in committed ARR at the close of fiscal year 2025 alongside more than $4 billion in cumulative software bookings—demonstrate substantial serviceable obtained market (SOM) penetration within these high-capacity environments, even as broad enterprise production adoption remains in early phases.[CM005, CM006, CM007, CM008, CM009]

2.3 Buyer Personas, Procurement Dynamics, and Value Justification

Adoption of VAST Data's platform spans three principal customer segments, each characterized by distinct procurement drivers and budget ownership structures. The primary spending engine consists of specialized AI neoclouds and GPU-as-a-service providers, exemplified by CoreWeave's multi-year $1.17 billion commercial agreement. In these accounts, procurement is led by cloud infrastructure executives and Chief Technology Officers whose primary objective is maximizing GPU utilization and eliminating I/O starvation across massive multi-tenant clusters. The second segment encompasses frontier AI labs, sovereign AI initiatives, and federal defense agencies—such as xAI's 200,000-GPU Colossus facility and the US Air Force—where research directors and program managers require exabyte-scale throughput, deterministic multi-tenancy, and strict data sovereignty. The third segment consists of data-intensive Fortune 500 enterprises across financial trading, autonomous driving, life sciences, and retail analytics, such as Lowe's. To address traditional enterprise procurement inertia, VAST introduced the Gemini commercial model, which decouples software licensing from physical hardware procurement. Customers acquire commodity hardware at cost via manufacturing partners like AVNET while paying VAST for capacity-based software licensing, allowing enterprise IT buyers to avoid conventional hardware markups while delivering software gross margins approaching 90% for VAST.[CM010, CM011, CM012, CM013]

2.4 Market Constraints, Switching Barriers, and Sizing Uncertainties

Despite compelling architectural efficiencies—such as Solidigm and VAST evaluations demonstrating a 58.9% total-cost-of-ownership reduction relative to hard disk drive tiers—the market faces significant adoption constraints and diligence risks. Enterprise IT departments remain constrained by organizational inertia, complex existing data management workflows, and long-term enterprise agreements with incumbent storage vendors like NetApp and Dell Technologies. Many corporate generative AI initiatives remain stalled in pilot phases, slowing the transition from experimental GPU capacity to enterprise-wide data platform rollouts. Furthermore, high customer concentration in specialized GPU clouds introduces revenue volatility linked to broader macroeconomic GPU financing conditions. Top-down TAM figures of $400 billion rely on aggressive assumptions regarding the displacement of traditional enterprise storage, while actual enterprise spend remains split across compute, networking, and legacy data repositories. Maintaining long-term revenue growth will require VAST to expand beyond hyperscale training deployments into sustained, high-margin enterprise inference pipelines and multi-cloud database workloads.[CM014, CM015]

2.5 Exhibits

3.1 Competitive Landscape and Architectural Taxonomy

The enterprise AI storage and infrastructure market has bifurcated into legacy enterprise storage providers adapting traditional architectures for GPU workloads and specialized, AI-native platforms engineered for massive parallel throughput. VAST Data competes across multiple fronts, confronting established enterprise all-flash and hybrid incumbents such as NetApp and Pure Storage, high-performance parallel file systems like Weka and open-source Lustre, and full-stack OEM systems from Dell Technologies and HPE. Historically, traditional enterprise storage systems relied on dual-controller or clustered shared-nothing architectures where data access was tightly coupled to specific controller nodes, creating inter-server communication bottlenecks and complex failover coordination when scaled to support thousands of GPUs. VAST Data's Disaggregated Shared-Everything (DASE) architecture breaks these constraints by separating compute nodes from shared storage media across high-speed NVMe-over-Fabrics, allowing stateless compute servers to access all QLC flash devices in parallel without inter-controller coordination. At its current $30 billion valuation, supported by a $1 billion Series F funding round, VAST Data commands a private market valuation exceeding the market capitalizations of long-standing public storage leaders such as NetApp and Pure Storage, signaling strong investor conviction in its platform-level divergence from commodity storage hardware.[CP001, CP002, CP003, CP004]

3.2 Architectural Differentiation and Core Capability Matrix

VAST Data differentiates its offering through a combination of storage media economics, protocol unification, and native AI security integrations. By deploying consumer-grade quad-level cell (QLC) flash for more than 90% of storage capacity, backed by storage-class memory for write buffering and global similarity reduction, VAST achieves high-performance all-flash throughput at economics competitive with tiered hybrid storage. Unlike single-purpose storage architectures that segregate file, object, and structured data, VAST unifies file systems, S3 object storage, and structured tabular databases within a single global namespace. This architectural consolidation prevents the latency and data duplication inherent in moving vast datasets across separate storage silos during AI training and inference. In response to growing enterprise security requirements for generative AI and agentic workflows, VAST previewed DataEnclave, an integrated runtime utilizing NVIDIA Confidential Computing to run models in hardware-isolated environments with cryptographic attestation. However, this confidential compute domain is becoming an active competitive battleground, as server OEMs including Dell Technologies and HPE are certifying native confidential AI stacks on Dell AI Factory and HPE Private Cloud AI ahead of VAST's planned Q1 2027 general availability.[CP005, CP006, CP007, CP008]

3.3 Commercial Packaging, Procurement Models, and Switching Barriers

Commercial packaging represents a key axis of competition between VAST Data and legacy hardware appliance vendors. Through its Gemini commercial model, VAST decouples software subscription licensing from physical hardware procurement. Customers purchase or lease certified commodity hardware directly from manufacturers or integration partners while licensing the VAST software platform based on capacity in 100TB increments. Because software licenses are fully transferable across hardware enclosures, enterprise buyers avoid the vendor-imposed hardware refresh taxes that historically characterized enterprise storage renewals. However, VAST does not publish transparent enterprise price books or volume discount tiers on its public website, requiring customers to negotiate customized multi-year contracts with sales teams. Switching costs for established VAST customers are substantial: once an enterprise consolidates its petabyte-scale training datasets, vector stores, and unstructured pipelines into VAST's unified namespace, migrating exabytes of data and rewiring active GPU pipeline integrations creates significant operational friction, reinforcing customer retention across leading AI cloud operators and enterprise adopters.[CP009, CP010, CP011, CP013]

3.4 Moat Defensibility, Commoditization Pressure, and Displacement Vectors

VAST Data's long-term defensibility relies on a combination of proprietary intellectual property, strategic ecosystem partnerships, and deep customer lock-in among premier AI infrastructure operators. The company maintains an active intellectual property portfolio of 43 global patents protecting foundational techniques in erasure coding, distributed scalable storage, and global metadata management, achieving a 94.7% US patent grant rate. Deep technical alignment with NVIDIA—spanning early adoption of BlueField DPUs, GPUDirect Storage, and NVIDIA Confidential Computing—positions VAST as a preferred data layer across major GPU clouds like CoreWeave and enterprise leaders including Mistral AI and JPMorganChase. Despite these advantages, VAST faces meaningful competitive and commoditization headwinds. Major cloud hyperscalers maintain native storage offerings with deep ecosystem integration and strong vendor consolidation appeal for corporate buyers. Concurrently, rapid standardization of AI data interfaces and hardware reference stacks by NVIDIA and open-source communities could allow server OEMs and well-funded storage peers to narrow VAST's software performance premium over time.[CP012, CP013, CP014]

3.5 Exhibits

4.1 Revenue Streams, Pricing Model, and Monetization Architecture

VAST Data operates as an enterprise software platform company monetizing a unified AI Operating System that collapses storage, database, and compute orchestration into a single software tier. Rather than selling proprietary storage appliances, VAST licenses its software under multi-year or annual subscription agreements based on the raw petabyte capacity deployed across customer clusters. This architecture enables enterprise buyers to procure certified commodity flash hardware directly from OEM partners such as Dell, Cisco, and Supermicro while paying VAST pure-play software licensing fees. Pricing is not publicly disclosed on official web properties; commercial terms reflect custom enterprise quotations with tiered volume discounting for multi-petabyte deployments. Major public cloud marketplace integrations, including managed offerings on Google Cloud, provide supplementary consumption-based channels while preserving VAST's software-centric revenue characteristics.[CI001, CI002, CI003, CI004]

4.2 Commercial Traction, Enterprise Scale, and GTM Motion

VAST Data has demonstrated unprecedented commercial acceleration within enterprise infrastructure, scaling from early commercialization to reported committed annual recurring revenue exceeding $500 million and cumulative bookings surpassing $4 billion by early 2026. The company reached $100 million in ARR within three years of launch and achieved $200 million in ARR by December 2023, reflecting rapid adoption among data-intensive organizations. VAST's enterprise go-to-market motion focuses on accounts with petabyte-scale unstructured data requirements across AI research, life sciences, quantitative finance, and GPU cloud infrastructure. Among its top 100 new accounts, average annual commitments exceed $1.2 million, with top customer engagements scaling beyond $100 million in total platform spend. Reported penetration exceeds 25% of the Fortune 100 alongside major GPU cloud service providers, establishing strong enterprise validation despite an opaque private reporting environment.[CI005, CI006, CI007, CI008, CI009]

4.3 Cost Structure, Gross Margin Dynamics, and Cash Flow Profile

VAST Data's cost structure reflects the operational advantages of an asset-light, software-defined business model. By offloading server and drive manufacturing, logistics, and hardware inventory to partner OEMs, the company avoids physical cost of goods sold and achieves reported software gross margins of approximately 90%. Cost of delivery is primarily composed of engineering support, cloud telemetry infrastructure, and technical customer success. The company reports maintaining positive operating margins and positive free cash flow for several consecutive years, eliminating operational burn and external capital dependency. However, because VAST's disaggregated architecture is optimized exclusively for all-flash QLC NAND media, enterprise customer hardware procurement remains sensitive to cyclical flash memory pricing and supply chain constraints, which can indirectly compress enterprise software budget allocations.[CI010, CI011, CI012, CI013]

4.4 Capital Adequacy, Concentration Risks, and Underwriting Diligence Gaps

In 2026, VAST Data finalized a $1 billion Series F financing round at a $30 billion valuation led by Drive Capital and Access Industries, providing growth capital and secondary shareholder liquidity while positioning the company for potential IPO readiness. Despite robust reported capitalization and self-funding operating cash flows, institutional underwriting faces substantial diligence hurdles. VAST's forward bookings exhibit heavy customer concentration, underscored by a landmark $1.17 billion multi-year agreement with neocloud provider CoreWeave. Any plateau in generative AI infrastructure spending or shift in CoreWeave's hardware roadmap poses direct downside risk to projected revenue. Furthermore, enterprise hardware incumbents including Dell Technologies have launched competitive counter-campaigns, specifically advertising power efficiency advantages for PowerScale AI storage. Without audited financial statements reconciling committed ARR to GAAP revenue recognition, investors must withhold definitive financial clearance.[CI014, CI015, CI016]

4.5 Exhibits

5.1 Disaggregated Shared-Everything (DASE) Core Architecture

VAST Data's foundational innovation is its Disaggregated Shared-Everything (DASE) architecture, designed to eliminate the performance bottlenecks, stranded capacity, and metadata synchronization overhead endemic to traditional scale-out NAS and shared-disk storage systems. In conventional clustered storage, compute and storage are tightly coupled within individual appliances, requiring complex inter-node chatter, distributed cache coherence protocols, and data re-balancing during scale-out operations. DASE resolves these structural constraints by separating stateless front-end compute nodes (CNodes) from highly dense, persistent flash storage enclosures (DNodes or EBoxes) over a low-latency NVMe over Fabrics (NVMe-oF) network fabric using RoCE or InfiniBand. Because front-end compute nodes are entirely stateless and maintain no persistent local cache, any compute node can directly address any NVMe solid-state drive across the entire cluster without routing requests through an owning node or engaging in distributed locking. A global transactional element store provides Atomicity, Consistency, Isolation, and Durability (ACID) semantics across all operations using atomic RDMA primitives, enabling simultaneous multi-protocol access to files, objects, and tabular records within a unified global namespace (DataSpace).[CE001, CE002, CE003]

5.2 Software Modules, InsightEngine, and Workflow Execution

The VAST AI Operating System consolidates four primary functional engines into an integrated software stack, moving enterprise data management beyond passive file storage into active computational infrastructure. The underlying DataStore delivers high-throughput multi-protocol file and object access supporting NFSv3/4.1, SMB 2.1/3.1, S3 object APIs, and NVMe-over-TCP without duplicating data across separate functional pools. Layered directly atop this substrate, VAST DataBase provides a low-latency columnar and tabular database supporting real-time SQL analytics, feature store operations, and native vector storage for retrieval-augmented generation (RAG). VAST DataEngine functions as an event-driven serverless processing tier that automates ingestion, transformation, and indexing directly as new data arrives on flash. Building upon these services, VAST InsightEngine continuously classifies incoming enterprise content, extracts contextual features, and generates vector embeddings via deep integration with NVIDIA Inference Microservices (NIM). This automated vectorization allows enterprises and foundation model builders to query exabyte-scale unstructured datasets in real time without engineering complex external ETL data pipelines. In media and entertainment environments, deployment partners such as Scale Logic utilize this unified namespace to eliminate traditional tiered hierarchies, providing simultaneous real-time multi-stream playback for 4K/8K video editing and seamless long-term archival.[CE002, CE011, CE012, CE014]

5.3 Hardware Appliance Packaging, Performance, and Commercial Licensing

To support enterprise on-premises and sovereign AI factory deployments, VAST packages its software stack onto turnkey hardware platforms through strategic original equipment manufacturer (OEM) and cloud partnerships. The Cisco AI Data Platform with VAST AI OS pairs Cisco CNode-X compute systems (UCS C845A M8 rack servers equipped with dual AMD EPYC 5th Generation processors and eight NVIDIA RTX PRO 6000 GPUs) with Cisco EBox persistent storage nodes (UCS C225 M8 rack servers housing all-flash NVMe SSDs and storage-class memory). In validated Cisco configurations, each individual EBox delivers up to 21 GB/s read throughput, 10 GB/s peak burst write throughput, and 2.6 GB/s sustained write throughput, backed by dual Mellanox ConnectX-7 400GbE network interfaces. However, initial cluster dimensioning imposes critical performance constraints: deployments configured with fewer than 12 EBox storage nodes suffer an approximate 6% to 15% reduction in maximum write bandwidth because the write buffer stripe width is permanently fixed at installation time and cannot be dynamically widened post-deployment. From a commercial standpoint, VAST operates purely as a software subscription model with terms spanning 36 to 60 months, pricing licenses based on usable capacity (per terabyte) and orchestrated CPU cores, with no perpetual licensing option available.[CE004, CE005, CE006, CE007]

5.4 Write-Path Mechanics, Network Fabric Pressure, and Architectural Risks

Despite VAST's substantial throughput and architectural simplicity advantages for read-heavy deep learning workloads, independent storage architectural analysis highlights meaningful engineering trade-offs inherent in the DASE write path and shared-everything topology. Because VAST utilizes commodity quad-level cell (QLC) NAND flash to achieve archive-like storage economics, incoming writes cannot be written directly to flash in small granular blocks without causing severe write amplification and premature drive wear. Instead, incoming client writes are routed to storage-class memory (3D XPoint) buffers and triplicated across three distinct physical media targets to guarantee power-fail safety before being acknowledged. Consequently, writes experience an estimated 3:1 performance penalty compared to reads, which is further exacerbated by the higher power consumption and lower write bandwidth of solid-state media. Furthermore, because every stateless compute node mounts every storage drive across the cluster via NVMe-oF, large-scale deployments create dense all-to-all communication patterns that place severe pressure on the storage network fabric, making cluster performance highly vulnerable to network topology congestion and packet loss. Additionally, client-side NFS access enforces close-to-open consistency rather than strict POSIX locking, meaning concurrent overlapping writes to identical file extents from multiple client hosts can produce data corruption unless applications enforce direct I/O semantics.[CE003, CE008, CE009, CE010, CE013]

5.5 Exhibits

6.1 Customer Base Segmentation and Buyer Profiles

VAST Data's customer base is structured across four primary segments: hyperscale and specialized GPU neocloud providers, sovereign AI infrastructure operators, regulated enterprise accounts, and high-performance computing or research institutions. In the specialized neocloud segment, buyers are infrastructure engineering leaders and cloud architects whose economic model requires maximizing GPU utilization and eliminating storage I/O bottlenecks. In the sovereign and regional cloud segment, operators such as Nscale in Europe, BUZZ HPC in Canada, and Sharon AI in Australia purchase VAST infrastructure to deliver hardware-isolated, compliant compute environments that adhere to local data residency requirements. Regulated enterprise accounts—spanning global financial institutions, healthcare networks like Boston Children's Hospital, and commercial enterprises like Booking Holdings—procure VAST through CIO and Chief AI Officer mandates to unify unstructured file repositories and vector embeddings under strict zero-trust governance. Finally, research and public-sector institutions, including NASA and the US Department of Energy, utilize VAST for multi-exabyte scientific simulations and petabyte-scale data pipelines. Across these groups, procurement is shifting from standalone storage evaluations to comprehensive AI platform architectures co-engineered with server OEMs and compute partners.[CU001, CU002, CU003, CU004]

6.2 Adoption Trajectory, Deal Sizing, and Revenue Scale

The adoption trajectory of VAST Data has expanded dramatically alongside the generative AI infrastructure buildout. By fiscal year 2026, the company reported accumulating over $4 billion in cumulative software bookings and more than $500 million in committed annual recurring revenue (ARR), while industry reporting indicates that total annualized run-rate revenue crossed $2 billion when factoring in non-committed hardware-coupled expansions. Enterprise adoption has deepened among large-scale organizations: the company's customer base among Fortune 1000 accounts quadrupled, and more than 25% of the Fortune 100 have deployed VAST. Furthermore, commercial expansion within enterprise accounts has generated strong ACV growth, with the top 100 new customer cohorts averaging more than $1.2 million in initial contract spend. Customer deployments increasingly operate at unprecedented physical and compute scale, supporting multi-exabyte storage footprints that feed clusters spanning hundreds of thousands of GPUs globally. Despite these impressive high-level velocity milestones, VAST does not publicly report total active customer counts or volume-weighted usage tiers, leaving an analytical gap regarding the exact distribution of revenue across its broader non-marquee client roster.[CU005, CU006, CU007, CU008]

6.3 Named Customer Proof and Production Workloads

Empirical proof of VAST Data's platform capabilities is anchored in large-scale production deployments across commercial AI, entertainment, and life sciences. The most prominent AI cluster deployment is xAI's Colossus supercomputer, where VAST powers the data tier across more than 200,000 Nvidia GPUs; according to published disclosures, VAST's disaggregated architecture reduced total cost of ownership by 50% compared to legacy storage architectures. In commercial cloud services, CoreWeave solidified VAST as its core data foundation through a landmark $1.17 billion commercial agreement signed in November 2025. In digital media, Pixar Animation Studios utilizes VAST to manage petabytes of rendered assets that serve as multimodal training data for animation pipelines. In the life sciences vertical, Boston Children's Hospital relies on VAST for high-throughput genomic sequencing data analysis, while enterprise commerce leader Booking Holdings deploys the platform for real-time recommendation data processing. These named references establish that VAST has progressed beyond exploratory laboratory pilots into primary operational production, though the reference base remains heavily weighted toward technology-forward pioneers and capital-rich AI infrastructure builders.[CU009, CU010, CU011, CU012]

6.4 Contract Durability, Retention Dynamics, and Concentration Risks

VAST Data's customer durability benefits from exceptional architectural switching barriers and extended commercial commitment terms. Typical enterprise contracts run for durations of five to seven years, creating high revenue visibility and insulation from short-term churn. Because VAST's Disaggregated Shared Everything (DASE) platform stores primary training data, operational metadata, and real-time vector indexes, displacing the platform requires complex, high-risk data migration across petabyte-to-exabyte volumes. Industry analyses note that once an organization deploys VAST into active production, customer churn remains remarkably low. However, this architectural lock-in is counterbalanced by acute commercial concentration risk. High-profile neocloud accounts such as CoreWeave and sovereign cluster buildouts represent a disproportionate share of contracted pipeline and committed ARR. Consequently, VAST faces meaningful counterparty exposure if leading GPU cloud providers experience capacity digestion, financing constraints, or renegotiations upon contract expiration. Additionally, independent research notes that while VAST's platform is highly capable, enterprise buyers face platform lock-in risks because migrating AI-focused vector and unstructured data out of VAST's integrated ecosystem requires substantial data engineering overhead.[CU013, CU014, CU015, CU016]

6.5 Exhibits

7.1 Ranked risk view and investment implication

The highest residual exposure is not a documented failure; it is the combination of migration friction, a vertically integrated operating model, and limited public evidence on concentration. Dell, an interested competitor rather than an independent benchmark, argues that VAST requires data to enter its namespace and depends on hardware and integration partners. VAST's own technical description confirms a shared-everything architecture in which every compute node mounts every SSD and system state is shared across the cluster. That supports a high-impact lock-in hypothesis, but not Dell's broader commercial conclusions. Financial disclosure is the second major risk: current reporting distinguishes cumulative bookings and committed ARR from cash-flow claims, while 2023 public materials used ARR and signed-contract language for the same headline amount. Proposed thesis-break triggers are a top customer above 20% of ARR without contractual protection, inability to reconcile reported commercial metrics within 10%, or a critical workload deployed without a tested recovery design. Those thresholds are investment controls, not reported facts.[CR001, CR002, CR008, CR009, CR010, CR011, CR012, CR016]

7.2 Operational, quality, and security exposure

VAST documents substantive mitigations, including DBox redundancy, DBox-HA, erasure coding, encryption, role-based and attribute-based access control, immutable snapshots, and external key-management support. The same white paper also identifies configuration boundaries that matter to underwriting: a classic cluster can go offline with a whole DBox failure; DBox-HA requires at least eight DBoxes and raises protection overhead in the stated example; cloud-local NVMe alone does not deliver the durability claimed for on-premises systems; and a single cloud instance has no failure protection. These are not evidence of incidents. They are reasons to verify the exact topology, availability tier, RPO/RTO tests, encryption settings, privileged-support process, and cloud persistence mode for each material deployment. A kill criterion should be any production-critical environment that cannot demonstrate recovery from its declared failure domain or any unremediated material security exception.[CR003, CR004, CR005, CR006, CR007, CR014, CR015]

7.3 Legal, regulatory, privacy, and governance diligence

The assigned evidence includes product-control descriptions but no regulator filing, court record, independent certification report, penetration-test summary, incident register, or privacy audit. The adverse search record returned no allocated URLs; that is a coverage limitation and must not be read as proof that no lawsuit, enforcement action, breach, or outage exists. VAST markets an isolated, encrypted, and verifiable confidential-AI runtime for sensitive data, while the technical paper describes FIPS 140-3 validated libraries, transport encryption, and granular access controls. Those controls lower design risk but do not establish operating effectiveness or compliance in each jurisdiction. Before investment, counsel should run litigation, sanctions, export-control, privacy, and IP searches in the United States, Israel, the EU, and other material sales jurisdictions, then reconcile results to management's claims, insurance, customer notifications, and security attestations.[CR001, CR006, CR007, CR015]

7.4 Partner, customer, and financial-model dependencies

VAST's partner network is simultaneously a moat and a correlated dependency. Public reporting identifies a $1.17 billion CoreWeave commercial agreement, while another source describes Nvidia as both investor and technology partner and CoreWeave as a large buyer. VAST also claims adoption across more than 25% of the Fortune 100 and leading GPU cloud providers, but this does not disclose revenue concentration, cancellation rights, renewal cohorts, or gross margin by deployment. The 2026 Series F reportedly included both primary and secondary capital; a lower-tier independent blog says more than half went to selling holders, but the assigned higher-quality source does not provide the split. Underwriting should therefore separate funding raised from cash available and bookings from ARR or revenue. Kill criteria are undisclosed related commercial terms with strategic investors, concentration above the approved threshold, or audited schedules that materially contradict public commercial or profitability claims.[CR002, CR008, CR009, CR010, CR011, CR012, CR013, CR016]

7.5 Exhibits

8.1 Investment Thesis, Recommendation, and Entry Discipline

VAST Data occupies a strategically critical position in the modern artificial intelligence stack, transforming from high-performance storage into a comprehensive AI Operating System that integrates storage, database functionality, and containerized compute. The company's technical architecture, based on Disaggregated Shared Everything (DASE), enables exabyte-scale deployments supporting millions of GPUs worldwide across major AI infrastructure providers including CoreWeave, Nscale, Firmus, and Fluidstack. Financial execution has been exceptionally rapid, with cumulative bookings surpassing $4 billion and committed annual recurring revenue (CARR) exceeding $500 million while maintaining operating profitability and positive free cash flow over several years. However, the April 2026 Series F funding round—which raised approximately $1 billion in combined primary and secondary capital led by Drive Capital and Access Industries—valued VAST Data at $30 billion. This headline price tags the company at roughly 60x committed ARR, leaving virtually zero margin of safety for external public shocks, hardware transition cycles, or AI multiple contraction. Consequently, our recommendation for uncommitted institutional capital is track, advising investors to maintain strict valuation discipline and await detailed private diligence on revenue composition, customer concentration, and contractual retention before participating at or near round pricing.[CV001, CV002, CV003, CV004, CV007, CV008, CV011, CV012]

8.2 Thesis, Anti-Thesis, and Late-Stage Financing Dynamics

Evaluating VAST Data at a $30 billion valuation requires balancing a compelling secular expansion thesis against substantial private-market anti-theses. The core thesis rests on VAST becoming the default data platform for hyperscale AI model development, inference clusters, and emerging agentic workflows that require unified low-latency data access across distributed hardware. Existing strategic backers such as NVIDIA, Fidelity Management & Research Company, and New Enterprise Associates (NEA) validate the company's platform centrality. Conversely, the anti-thesis highlights three material structural risks. First, VAST Data's reported revenue construct of committed ARR may diverge meaningfully from audited GAAP recognized revenue or trailing subscription ARR, obscuring true run-rate recurring cash flows. Second, significant commercial exposure to specialized GPU cloud providers creates pronounced counterparty concentration risk should neocloud buildouts slow or face capital rationing. Third, the Series F round incorporated substantial secondary liquidity for early investors and employees, alongside dual reported funding figures ($500 million primary equity versus $1 billion total transaction volume), creating potential liquidation preference overhang and dilution complexities. Diligence must establish full cap table transparency and evaluate whether common equity value matches the preferred headline valuation.[CV002, CV004, CV005, CV006, CV009, CV010, CV013, CV014]

8.3 Scenario Valuation, Peer Comparables, and Diligence Kill Triggers

A robust scenario framework demonstrates that VAST Data's current valuation embeds sustained high-growth execution and continued public-market enthusiasm for AI infrastructure assets. In our bull scenario, durable recurring revenue scales toward $1.5 billion to $2.0 billion as enterprise agentic deployments mature, supporting a 20x to 25x multiple and an enterprise valuation of $30 billion to $45 billion upon public listing. In our base scenario, audited recurring revenue stabilizes in the $800 million to $1.0 billion range with moderate enterprise expansion, yielding a valuation of $16 billion to $22 billion at a 18x to 22x software-infrastructure multiple—representing a 25% to 45% discount to the Series F mark. In our bear scenario, AI capital expenditure cycles cool, hyperscaler in-house storage alternatives compress pricing power, or major neocloud customers renegotiate capacity, compressing revenue to $600 million to $700 million and multiples to 10x to 12x, resulting in an equity value of $6.0 billion to $8.5 billion. Relative to public infrastructure peers such as Pure Storage and NetApp, VAST trades at a multiple premium exceeding 5x, placing the burden of proof squarely on private metrics. Investors must establish strict kill triggers: any audit finding that reveals top-three customer concentration exceeding 45% of ARR, trailing gross margins below 70%, or a prolonged stall in formal IPO governance preparations should immediately terminate further investment consideration.[CV003, CV007, CV008, CV011, CV014, CV015, CV016]

8.4 Exhibits