Strategic PartnersLatest editionIssue 06
VAST Data
TACC's Horizon User Guide, updated November 10, documents that early users of the NSF Leadership-Class Computing Facility system at UT Austin work on a shared VAST file system for $HOME and $SCRATCH, without Lustre striping, with a 10-day unbacked scratch purge and a $WORK tier deferred to late 2026.
Full research summary
TACC's Horizon User Guide, updated November 10, documents that early users of the NSF Leadership-Class Computing Facility system at UT Austin work on a shared VAST file system for $HOME and $SCRATCH, without Lustre striping, with a 10-day unbacked scratch purge and a $WORK tier deferred to late 2026. TACC's October 28 report says Horizon has entered only its first phase of early operations as Frontera retires. For public research universities this is an operator design reference, not yet measured performance or reliability evidence.
Read the edition Previous: Issue 05, November 5All VAST Data editions
What this stream covers
VAST Data AI infrastructure, storage platforms and data management solutions. Separate vendor claims from independent deployments, benchmarks and limitations. Explain Government/Education applicability where supported, without forcing irrelevant news into a SLED use case. Other storage vendors belong in their respective streams.
- Evidence records
- 2
- Cross-source patterns
- 1
- Also published November 10
- AnthropicAWSCampus OperationsCiscoCitizen Engagement & ServicesCrowdStrikeDellEmergency ServicesEverpureGoogleHealth & Human ServicesK–12Local GovernmentMicrosoftNetAppNVIDIAOpenAIPalo Alto NetworksPublic SafetyRed HatResearch UniversitiesServiceNowState GovernmentStudent SuccessTransportationUtility Management
- Subscribe
- Atom feed for VAST Data
What the evidence supports, and what it leaves open
1 pattern this edition will stand behind, each carried by at least two independent sources. Every one arrives with the question it has not answered.
- Horizon's VAST user file systems are in early operations, not yet production evidence
Operating questionBefore citing Horizon in a university or state research-computing storage decision, which measured early-user throughput, availability and migration results will TACC publish, and when will $WORK and production status arrive?
Research through your lens
Every resource includes source evidence and takeaways for all three roles.
Evidence in this micro-vertical
14 resources
Follow the outcomes
14 resources across outcomes in your selection. Counts include all outcomes.
Refine by evidence type and topic
TACC's Horizon User Guide (updated Nov 10, 2026) puts NSF LCCF early users on shared VAST $HOME and $SCRATCH, with $WORK deferred to late 2026
The Horizon User Guide, last updated November 10, 2026 and labeled a draft in progress, states that Horizon, the NSF Leadership-Class Computing Facility system at UT Austin, uses a shared VAST file system for $HOME and $SCRATCH. These file systems are not Lustre and do not support stripe count or stripe size settings. $WORK (VAST) is unavailable to early users and planned for late 2026. $SCRATCH has no quota, about 400 PB overall capacity, no backup and a 10-day access-time purge. A notice dated 07/24/2026 still limits Horizon to internal users.
Limitations & uncertainty
Draft documentation; internal inconsistency between 'up to 200PB' local solid-state storage in the specification table and '~400 PB' scratch capacity in the file-system table; quotas listed as TBD; no measured throughput, uptime or user outcomes; operator rather than independent evaluator.
Blocks & Files qualifies VAST's $10B figure as commitments not necessarily orders, and records allocation-by-commitment under flash shortage
Same-day Blocks & Files coverage restates VAST's 14-customer / $10B / 200 EB claims, then qualifies them: the fiscal window is February 1–October 31, 2026; the $10B is aggregate commitments for software and supporting hardware rather than necessarily orders already placed; there is no official top-10 frontier-lab ranking; 'significant percentage' of enterprise NAND is undefined (editor suggests 2–5% as a possible floor); and a VAST spokesperson said demand exceeds supply so aligned capacity is allocated by customers' advance commitments and delivery schedules, not an unlimited guarantee. The piece also sketches forward supply planning with SSD partners without per-order triggers.
Limitations & uncertainty
Trade press, not a formal audit; some hypothetical supplier dialogue is illustrative. Author has previously been criticized by StorageMath for amplifying VAST claims, so read qualifications carefully but do not treat the article as adversarial research.
VAST says 14 customers committed over $10B in FY2026 YTD and aligns 200 EB of flash supply through 2029
On November 5, 2026 VAST Data announced that in the first three quarters of its current fiscal year, 14 of more than 1,000 customers made more than $10 billion in aggregate commitments for VAST software and supporting platform hardware; each of the 14 committed over $100M and four over $1B, and VAST-powered systems are said to support 9 of the top 10 frontier AI labs. To meet demand, VAST says it structured agreements aligning more than 200 exabytes of flash capacity through 2029, describing partner deliveries as a significant percentage of global NAND for enterprise data centers, and states the planning helps AI clouds, labs, enterprises and government organizations building on the VAST AI OS.
Limitations & uncertainty
Vendor-claim only; commitments are not broken out as software vs hardware revenue or as booked orders. 'Significant percentage' of enterprise NAND is undefined. '9 of top 10 frontier AI labs' has no cited ranking. Government mention is aspirational planning language, not a deployment case study.
- Source
- VAST Data blog (Jason Gaudreau, Senior Product Marketing Manager)
- Published
- 2026-10-30 (page metadata 19:20 UTC; modified 2026-10-31 02:04 UTC; on-page byline shows Nov 3, 2026)
VAST models VAST on Cloud Event Broker at $238,387 a month versus $1.1 million for Confluent Cloud Enterprise at 4 GB/s, and concedes long retention can cost more
VAST's own TCO model prices a Kafka-compatible VAST Event Broker running on VAST on Cloud virtual EBoxes against Confluent Cloud Enterprise list pricing (October 2026) for a 4 GB/s, 2:1 fanout, five-day-retention workload: $238,387 versus $1,106,530 per month ($23.78 versus $110.38 per TB ingested). The gap is attributed to Confluent's per-GB ingress/egress and post-replication storage fees versus erasure coding with 2.7% to 20% parity overhead. VAST concedes long retention on flash EBoxes can cost more than archive-tier cloud offerings and says value is greatest for 1 to 14 days. Full-AZ recovery takes 3 to 15 minutes, with data lazy-loaded from object storage.
Limitations & uncertainty
Vendor claim. Single modeled scenario; no independent reproduction; publication date ambiguous (metadata Oct 30, byline Nov 3). Event Broker performance claims have previously been criticized for missing methodology (see StorageMath item). Treat savings as a hypothesis to test, not an expected outcome.
- Source
- GovTribe (text version of Empire AI RFI No. EAI-2026-102902)
- Published
- 2026-10-29
New York's Empire AI issues a $160M Gamma RFI that names Beta's 20 PB of VAST as incumbent storage Gamma must interoperate with
Empire AI's Gamma RFI (issued October 29, 2026; responses due December 4) seeks input on up to 5 MW of AI/HPC compute and white space for its UB North Campus facility, with a roughly $160 million budget split across Gamma and Delta tranches. It records that Beta (online August 2026) includes 20 PB of VAST and 10 PB of DDN storage, and asks respondents to describe storage capacity, sustained throughput, metadata performance, data protection and integration with the existing VAST and DDN systems, whether or not Beta is relocated. It reports 1,631 active users, 314 projects and over 80 percent utilization on Alpha and Beta.
Limitations & uncertainty
The RFI is not a solicitation and awards nothing; naming VAST reflects the installed base, not a preference. The text was read through a third-party portal's extracted version, which showed only part of the 40-page document; the original portal posting requires an account. Interpretation (not source finding): interoperability language may favour incumbents in a later RFP, but Empire AI explicitly states it does not prescribe a manufacturer.
TACC reports Frontera retired in October 2026 and Horizon in its first phase of early operations with exabyte-capacity storage in place
TACC's Oct. 28, 2026 feature on the ninth TACCSTER symposium reports that Frontera was retired in October 2026 after more than eight million jobs and that Horizon, the NSF LCCF centerpiece, has entered its first phase of early operations. Executive Director Dan Stanzione announced that power infrastructure is complete, an exabyte-capacity storage system is in place, and initial deployment of Horizon's AI-focused hardware is underway. The article does not name VAST or give a production date.
Limitations & uncertainty
Predates the last completed run (used for why-now context); institutional news, not an evaluation; VAST not named; 'exabyte-capacity storage system' likely aggregates tiers and should not be attributed wholly to VAST.
VAST Data announces an Agent Sandbox Foundation Stack with NVIDIA NemoClaw and OpenShell, claiming persistent agent state and durable audit records
VAST announced a Foundation Stack in which NVIDIA OpenShell provides isolated execution and policy enforcement, while the VAST AI OS supplies zero-copy-clone sandbox workspaces (DataStore), shared agent memory/context (DataBase), and a persistent record of agent tool calls, prompts, tokens, costs and decisions, plus VAST Management System observability. VAST says it will demo at GTC Berlin and publish an automated deployment on GitHub. No customer deployment, benchmark or independent validation is cited.
Limitations & uncertainty
Single vendor source with partner quotes; no performance, scale or customer data. GitHub deployment was described as planned. Page date label shows Oct 20 while the dateline is October 21, 2026 (dateline used). Independent findings concern NVIDIA OpenShell, not VAST components; NVIDIA runtime coverage belongs primarily to the NVIDIA stream and is cross-referenced here only because it underpins this stack.
VAST Data and Hyperfusion plan containerized sovereign AI inference (Sentinel) with VAST AI OS as preferred data platform; technical validation still in progress
VAST and UAE-based Hyperfusion announced plans to integrate the VAST AI OS into Sentinel, a factory-built 45-foot container AI data center (reference 96 GPUs, about 2-3 PB per unit; HyperPod about 20 PB), marketed for rapid, sovereign and optionally air-gapped inference with confidential computing. The release states joint go-to-market will begin while technical validation of VAST within Sentinel and HyperPod is completed.
Limitations & uncertainty
Announcement of intent; technical validation incomplete; no customers named; non-U.S. vendor context limits transferability to U.S. state procurement. Page date label shows Oct 18 while the dateline is October 19, 2026 (dateline used).
VAST blog maps AI OS encryption, retention locks, WORM, VIP isolation and immutable snapshots to HIPAA, while conceding compliance needs policy and training
VAST's October 15, 2026 blog (re-shared on its community site October 27) describes software AES-256 encryption at rest using FIPS 140-3 validated libraries, internal keys or KMIP external key managers (per-tenant or per-folder keys only with an external EKM), in-flight encryption for NFS, SMB, S3 and replication, retention locks and WORM, VIP-based client isolation with LDAP/ABAC, and immutable snapshots. It states compliance requires training, policies and technology, yet closes by saying healthcare IT on VAST is compliant.
Limitations & uncertainty
Single vendor source; the concluding 'compliant' statement is a marketing claim not supported by an assessment. No mention of business associate agreements, audit logging or AI-specific data flows. Interpretation: SLED buyers should treat this as a feature checklist to verify, not evidence of compliance.
Georgia Tech's NSF-funded Nexus AI supercomputer opens nationally on an eight-appliance VAST Data all-flash storage layer, still pre-production as of Oct 10
Supercomputing News reports that Nexus, a $20M NSF Category I capacity system at Georgia Tech's Coda data center, uses eight VAST Data Ceres V2 all-flash appliances (about 10.8 PB raw NVMe) on a separate storage fabric that is non-blocking for DGX B200 nodes and oversubscribed for RTX PRO 6000 nodes. At least 90% of capacity is allocated nationally through ACCESS/NAIRR peer review. As of October 10, ACCESS listed Nexus as pre-production; the ribbon cutting was October 12. No storage performance results are published.
Limitations & uncertainty
Secondary reporting (AI-assisted drafting per SCN disclosure) with no storage performance data. Operator portal and SCN disagree on storage-fabric type and RTX node/GPU counts; treat specifications as unconfirmed until the hardware guide is checked directly. System was pre-production at reporting time; applications were 'opening soon'. VAST's own social post about the ribbon cutting was not used.
Massachusetts AI Hub's state-funded AICR cluster runs on ~14.3 PiB of VAST all-flash storage, with snapshot-only protection and no off-site backups
AICR's system description (page dated October 8, 2026) lists 31 B200 nodes (248 GPUs), 19 RTX PRO 6000 nodes (152 GPUs) and 32 service nodes, with ~14.3 PiB effective VAST all-flash storage rated at 360 GB/s, encryption at rest and GPU Direct Storage over NFS-RDMA on InfiniBand NDR400. Home and work areas have 7-day snapshots; scratch has none and is purged; the operator warns there are no off-site backups. UMass's AICR page (modified October 20, 2026) marks the system Beta, proposal-only with per-project caps, and bars interactive inference.
Limitations & uncertainty
Documentation can change without notice and is not an independent evaluation. Hardware vendor for compute/service nodes is Dell (covered in the Dell stream); this record concerns only the VAST storage layer. The $31M figure and university match come from MGHPCC's December 2025 post, not the October pages. Interpretation (not source finding): snapshot-only protection is an operator design choice, not a VAST limitation.
LLNL study on the Tuolumne production system finds VAST avoids all-reduce contention but still shows heavy-tailed DataLoader stalls in 12% of iterations
On LLNL's Tuolumne (HPE Slingshot-11, four nodes with 16 AMD MI300A GPUs, a real GNN training workload), DataLoader waits were 15 ms median on VAST but spiked to multiple seconds in 12% of VAST iterations versus 28% for Lustre. VAST, mounted over NFSv3/TCP, sits in a separate Slingshot traffic class, which largely spared all-reduce from contention, but its data path still crosses the shared fabric. Node-local NVMe staging (DYAD) was 1.06x faster than VAST over one epoch and 1.31x once the cache warmed, and 7.4x faster than Lustre.
Limitations & uncertainty
Single workload and small scale; results depend on Slingshot traffic classes and NFSv3/TCP mounts and may differ with RDMA or other fabrics. Authors evaluate their own mitigation. Not a SLED deployment; transferability to university or state clusters is an interpretation. Original submission September 6, 2026; v2 September 19.
Lasso Security research shows data exfiltration from NVIDIA NemoClaw/OpenShell sandboxes through default policy-permitted egress (npm, node, git/gh)
Lasso demonstrated two attack scenarios in which an agent inside an OpenShell sandbox, using default egress policy, installed a malicious npm package and exfiltrated credentials and files (including plaintext API keys) via allowlisted gh/git paths, with cron-based persistence in the second scenario. Lasso reports NVIDIA's disclosure program deemed the scenarios out of scope. Researchers manually adjusted some default policies (e.g., adding api.openai.com, node path, installing gh) to make the environment functional.
Limitations & uncertainty
Tested an early/alpha NemoClaw/OpenShell release in April 2026 with some manual configuration changes; later versions may differ. Lasso sells AI security products (potential interest). Findings concern NVIDIA components, not VAST software; NVIDIA-runtime coverage belongs primarily to the NVIDIA stream.
StorageMath argues VAST's 99.9991% uptime and 10x-over-Kafka Event Broker claims lack the methodology needed for verification
StorageMath examines VAST's claimed 99.9991% measured uptime (about 4.7 minutes a year) and Event Broker claims of a 10x+ advantage over Kafka on like-for-like hardware, later stated as 604% more throughput. It argues neither claim can be verified because VAST has not published definitions of downtime, sample size, exclusions, hardware and configuration, replication and acks settings, latency distributions, or reproducible scripts. It concludes the claims should be treated as marketing until methodology is published, while allowing the Event Broker may genuinely be faster.
Limitations & uncertainty
Ten months old and polemical in tone; anonymous authorship; no independent test data. It references 2025 claims as reported by Blocks and Files; VAST may since have published more methodology (VAST's OpenMessaging test blogs were seen in search results but not inspected for this edition).
Stream editions
Each edition carries its own synthesis and evidence ledger.
November 10, 20261 edition
November 5, 20261 edition
November 1, 20261 edition
October 31, 20261 edition
October 30, 20261 edition
October 29, 20261 edition