Cohesity at NVIDIA GTC 2026

At NVIDIA GTC 2026 this week, Cohesity will showcase how Cohesity Gaia, built on NVIDIA AI Enterprise, delivers trusted sovereign AI and unlocks historical, governed enterprise data for agents and AI platforms—wherever that data resides. No extra data copies to manage. No complex ETL pipelines to build. No need to rebuild permissions.

Here are the top things we’re highlighting at NVIDIA GTC 2026:

• Backup data, now AI-ready and accelerated by NVIDIA AI Enterprise: Gaia delivers faster model deployment, higher answer precision, and fewer unsafe responses – helping organizations deploy AI faster, improve answer quality, and strengthen policy compliance.
• Advancing agentic workflows with NVIDIA: Cohesity is exploring deeper integration with NVIDIA Agent Toolkit, including NVIDIA AI-Q, NVIDIA NeMo Agent Toolkit profiling and optimization, and NVIDIA cuVS for vector indexing to expand Gaia into more advanced agentic workflows—potentially enabling up to 33% more multi-hop questions answered and up to 6x acceleration in vector indexing throughput of large-scale unstructured data on two NVIDIA H100 GPUs.
• Better answers from your agentic AI tools: Gaia is designed to enhance Microsoft Copilot, Glean, and Google Gemini by grounding them in governed enterprise data.
• Frictionless data access for AI pipelines*: Cohesity Gaia Catalog will expose curated datasets through a secure, read-only, S3-compatible endpoint—no data duplication, no new ETL pipelines.
• Run AI where the data lives: Gaia is designed to support deployment across major cloud environments as well as on-premises, enabling organizations to run AI workloads close to their data without requiring code changes.
• Protect AI infrastructure and recover fast: Cohesity supports secure data isolation and rapid recovery, so AI pipelines and agents remain resilient when incidents occur.

Gaia: Now available for AI agents

Powered by NVIDIA AI Enterprise, Gaia can deliver measurable performance gains:

• Faster model deployment
• Higher answer precision
• Fewer unsafe responses

Most organizations have struggled to analyze unstructured data spread across cloud and on-premises environments—let alone reason over how that data has changed over time. Gaia makes both possible. 

By running the industry’s first retrieval-augmented generation (RAG) directly on backup data, Gaia unlocks unstructured enterprise data for agentic reasoning. It builds a governed semantic layer by extracting, chunking, and embedding enterprise content for high-performance, policy-aligned retrieval. 

To power this capability, Cohesity uses NVIDIA AI Enterprise—including NVIDIA NIM LLM, NVIDIA Nemotron Reranking NIM, and NVIDIA NeMo Guardrails—to deliver fast, accurate, and secure responses.

Through integrations with Microsoft Copilot, Glean, and Google Gemini, Gaia will allow agents to query curated datasets directly—without requiring users to manage additional data copies or rebuild permissions—bringing governed enterprise history into the AI tools teams already use.

Cohesity is also evaluating the use of NVIDIA AI-Q Blueprint, the NVIDIA NeMo Agent Toolkit, and NVIDIA cuVS to power more advanced agentic workflows. NVIDIA AI-Q Blueprint provides a ready-to-use agent architecture to Cohesity data can power richer, multi-step assistants. NVIDIA NeMo Agent Toolkit keeps those agents high-quality and reliable through built-in monitoring and optimization while NVIDIA cuVS makes ingestion of Cohesity’s large scale unstructured data fast enough for real-time, fresh interactive experiences. Together this enables up to 33% more multi-hop questions answered, dramatically accelerating how enterprise data becomes AI-ready for investigation and insight.

“Secure agentic AI requires both performance and governance at scale,” said Pat Lee, vice president, Strategic Enterprise Partnerships, NVIDIA. “By leveraging NVIDIA AI Enterprise and Agent Toolkit, Cohesity Gaia demonstrates how accelerated inference, policy-enforced guardrails, and high-performance indexing can deliver measurable gains in deployment speed, answer precision, and ingestion throughput for enterprise AI.”

Gaia Catalog*: AI-ready data for AI pipelines

Data engineers need governed enterprise datasets for custom RAG stacks, agent training, and analytics pipelines. Traditionally, this meant building complex ingestion workflows and duplicating massive volumes of unstructured data—adding cost and risk.

Gaia Catalog will be the secure gateway to backup data, eliminating data duplication, permission rebuilding, and refresh maintenance.

Cohesity Data Cloud continuously refreshes datasets while preserving immutability and auditability. This results in faster experimentation and production deployment without months of data engineering overhead.

Gaia Catalog will enable teams to:

• Discover and curate time-series unstructured data using natural language or SQL.
• Filter by file type, ownership, permissions, and time range.
• Enrich datasets with intelligent classification and business context.
• Publish approved datasets through a secure, read-only S3-compatible endpoint.

*Sign up for Gaia Catalog Early Access at GTC.

Sovereign AI: Run AI where the data lives

For many enterprises, especially in regulated industries, AI must run inside their own environment. Data residency and sovereignty requirements prohibit moving sensitive data to public cloud AI services.

Cohesity supports a sovereign AI architecture that combines:

• Cisco AI Pod infrastructure
• NVIDIA AI Enterprise (including NIM, NeMo Guardrails, NVIDIA Nemotron reranking NIM)
• NVIDIA AI Infrastructure (NVIDIA L40S, NVIDIA Hopper, NVIDIA RTX PRO)
• Cohesity DataProtect for secure, immutable data capture
• Gaia for governed data activation

Gaia self-managed deployments allow AI processing to run locally while accessing protected backup data, maintaining governance and auditability while benefiting from NVIDIA-accelerated AI performance.

These deployments enable organizations to build once and deploy anywhere—from on-prem data centers to sovereign cloud environments—without refactoring applications or retraining models.

Cohesity booth details at NVIDIA GTC 2026

Visit Cohesity at booth #3003 to speak with our experts and see live demos. Learn how the Cohesity Data Cloud helps organizations build trusted sovereign AI for the agentic enterprise.
 

This blog post may contain forward-looking statements that are subject to risks, uncertainties, and assumptions. You should not rely upon forward-looking statements as predictions of future events. All statements other than statements of historical fact could be deemed forward-looking. Forward-looking statements include statements of belief or expectations or plans of management for future operations or statements concerning new or planned product or service availability and technology developments.

Although Cohesity believes that the expectations reflected in the forward-looking statements are reasonable, Cohesity cannot guarantee that the future results, performance or events reflected in the forward-looking statements will be achieved and such future results, performance or events are subject to change. Any unreleased services or features referenced in this blog post are not currently available and may not be made generally available on time or at all, as may be determined in our sole discretion. Any such referenced services or features do not represent promises to deliver, commitments, or obligations of Cohesity and may not be incorporated into any contract. Customers should make their purchase decisions based upon services and features that are currently generally available.