From the NetApp edition of November 4, 2026
Futurum: NetApp Novus and PEAK:AIO pursue AI factories with design targets, not published benchmarks
The Futurum Group · Independent technology analyst firm · Global
- Publisher
- Futurum
- Original publication
- 30 September 2026
- Source retrieved
- 2026-11-05
What happened
Futurum analysts (30 September, newly relevant after the 2 November PEAK:AIO close) say NetApp Novus targets 100 TB/s and that NetApp has published design targets rather than benchmark results. They read PEAK:AIO as buying metadata engineering for Novus's hardest part, note NetApp had not said how PEAK:AIO relates to initial Novus systems, and flag governance questions for agent access via the Console MCP server: identity binding, least privilege, approvals and immutable audit logs. They also note a locally installed Console that runs air-gapped and replaces Active IQ Unified Manager extends to regulated and public-sector sites.
Why it matters
Limited direct SLED relevance: Novus targets AI clouds and neoclouds. University research computing centers evaluating NetApp for AI may care, and the air-gapped Console and agent-governance questions apply to regulated public-sector sites. Analyst opinion, not deployment evidence.
Evidence and measured results
Opened the full article; read the news summary, analyst take, PEAK:AIO, AI Data Engine, Keystone and what-to-watch sections. Futurum discloses it has engaged with many technology companies.
Limitations and uncertainty
Analyst opinion; predates the deal close; Futurum discloses vendor relationships; no measured results.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-04; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Research and AI program leads need storage that keeps GPUs busy, without relying on unproven vendor targets.
- Stakeholders
- Research computing director
- CIO
- CISO
- Procurement officer
- Discovery questions
- What GPU utilization and storage throughput do you measure today?
- Would you connect an LLM agent to storage operations, and under what controls?
- Value hypothesis
- A measured evaluation reduces the risk of overbuying AI storage.
- Potential engagement
- A bounded proof-of-concept plan with independent test criteria.
- Unsupported claims
- Novus reaches 100 TB/s in production
- Any SLED Novus deployment
- Guaranteed GPU utilization gains
Pre-sales engineering
Role takeaway
- Fit
- Large research universities or state AI centers with GPU clusters; poor fit for typical agency storage.
- Architecture and integration
- pNFS Flex Files with standard Linux NFS clients; ONTAP data nodes.
- Prerequisites
- GPU cluster workload profile
- Network fabric able to carry pNFS traffic
- Constraints
- No published benchmarks
- Initial hardware limited to AFF A90 variant
- PEAK:AIO integration path not stated
- Security
- Govern Console MCP agent access with least privilege, approvals and audit logs.
- Proof of value
- Metadata-heavy small-file benchmark
- Sustained read and write throughput test
- Failure and recovery test
Delivery
Role takeaway
- Work
- Design and run a time-boxed evaluation.
- Dependencies
- NetApp test hardware
- Representative workloads
- Security review
- Ownership
- Research computing or infrastructure lead with security.
- Skills and adoption
- pNFS, ONTAP and agent-governance skills.
- Governance checkpoints
- Architecture review before pilot
- Security review before agent access
- Acceptance criteria
- Sustained throughput meets an agreed target (proposal)
- Audit logs capture every agent action (proposal)
- Risks
- Relying on vendor targets
- Integration delays
- Agent misuse of storage operations
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
PNFS with standard Linux clients; initial systems on an AFF A90 variant, software-defined versions later per NetApp.
Governance
Who approves, reviews and stays accountable for outcomes?
Agent access to storage via MCP needs identity, least privilege, approvals and audit logs.
Security and privacy
What data, permissions and controls need testing?
Agentic storage operations widen the control-plane attack surface if not governed.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Research computing staff would need pNFS and ONTAP skills.
Procurement
What should contracts, pricing and exit terms secure?
Buyers should require published, workload-specific performance before committing.
Operating model
Which teams own the service once it runs?
Console MCP access needs human approval steps and audit trails.
Publication history
- 2026-11-04NetApp · Issue 062 resources
Stable resource ID: futurum-netapp-novus-peak-aio-two-market-ai-strategy-design-targets-not-benchmarks-20260930