Ready to get started?
Deploy sovereign AI on your infrastructure - in weeks, not months.
HPE × Katonic · AI Factory & Private Cloud AI Partner
Katonic runs natively on HPE Private Cloud AI and HPE AI Factory, turning ProLiant compute, Aruba networking and Ezmeral data fabric into production agents your business actually uses.
Hardware from HPE. GPUs from NVIDIA. The agent platform from Katonic. One sovereign stack.
Katonic is a Day-1 partner for both HPE AI go-to-market motions. Whether you compose your own AI factory or unleash a Private Cloud AI appliance, the same governed agent layer sits on top.
Build it your way. Modular HPE compute, storage and networking sized for the workload, with Katonic as the agent runtime above the stack.
Turnkey AI cloud, on-prem. HPE GreenLake-managed. Pre-validated NVIDIA AI Enterprise stack. Katonic shipped as the agent layer at day one.
GreenLake
HPE GreenLake Cloud · control plane
Consumption-based control plane. Pay for what you use. Katonic billed alongside your other GreenLake services.
Unified observability, identity, billing. Katonic agents inherit GreenLake org & cost tags. One bill across compute, storage and the agent layer.
ProLiant
ProLiant Gen12 + Cray XD · validated
Reference compute for HPE AI Factory. DL380a Gen12 with HGX, or Cray XD675 dense liquid-cooled GPU nodes.
Validated with NVIDIA H100/H200/B200. Katonic K8s control plane installs on HPE iLO-managed nodes. Zero-touch provisioning supported.
Ezmeral
HPE Ezmeral Data Fabric · + S3 + Kafka
Unified file, object and stream namespace. Katonic RAG ingests from Ezmeral with row-level entitlements.
Posix + S3 + Kafka in one namespace. Katonic connector reads tables, mirrors permissions to vector index, supports incremental refresh and TTL.
Alletra MP
HPE Alletra MP Storage · all-flash
NVMe-class storage for vector indexes, training datasets and agent state. Disaggregated, scales independently.
Validated for Private Cloud AI. Katonic deploys Milvus / pgvector on Alletra block. Snapshots, replication and air-gap aware.
Aruba CX
Aruba CX 10000 Networking · in-fabric
Distributed services switch. East-west security policy enforced in the fabric, not in the agent.
Microsegmentation for agent tenants. Katonic Control Room maps Aruba policy groups to org units. Audit trail unified with platform logs.
GreenLake AI
GreenLake for Large Language Models · in-region
Sovereign LLM inference on HPE infrastructure. Katonic AI Gateway routes to GreenLake LLM endpoints first.
Models stay in-region. Katonic routing tier prefers GreenLake / Private Cloud AI; cloud providers as failover. Per-call attribution back to HPE infra.
HPE and NVIDIA co-engineered the full stack from networking fabric to the AI Enterprise software layer. Katonic adds the agent runtime, guardrails and the end-user surfaces.
One procurement. One support path. One sovereign deployment.
Pre-validated, jointly supported
Configurations validated by HPE and NVIDIA engineering. Katonic certified on each PCAI t-shirt size (S / M / L / XL). Single escalation path.
Networking through Spectrum-X & Aruba
Microsegmentation per agent tenant in the fabric. East-west visibility surfaced into Control Room. Zero-trust between business units.
Sovereign by default
Models, vectors, prompts and traces never leave the HPE infrastructure. NVIDIA NIM serves inference locally. Katonic AI Gateway routes on-prem first.
Katonic Workroom · Studio · Control Room
Use · Build · Govern
Katonic Agent Runtime + Guardrails
21 platform services
NVIDIA NIM · NeMo · NAT
Microservices, safety, agents
NVIDIA AI Enterprise
Curated NVIDIA software stack
HPE Private Cloud AI control plane
GreenLake-managed
HPE Ezmeral Data Fabric + Alletra MP
Data + storage layer
HPE ProLiant / Cray XD + NVIDIA HGX
Compute · H100 / H200 / B200
Aruba CX 10000 + NVIDIA Spectrum-X
Networking fabric
HPE customers are buying the substrate. They still need the agentic application layer on top. That’s where joint HPE × Katonic deals close.
Saudi Arabia, the UAE and Gulf nations are building sovereign AI infrastructure on HPE. They need agents, not just infrastructure. Katonic ships the eighty pre-built business agents that justify the factory's existence.
Vision 2030 · Sovereign cloud programs · National AI initiatives
Service providers building agentic AI offerings on HPE hardware need a multi-tenant agent platform that can be white-labeled. Katonic's AI Cloud delivers exactly this, with per-tenant FinOps and Distributor Console.
Regional telcos · ePLDT · e& enterprise
Banks, governments, healthcare and defense organisations buying HPE ProLiant and Cray XD for AI workloads need governance, isolation and the agentic depth that hyperscaler platforms can't match inside their perimeter.
Banking · Public sector · Healthcare · Defense
Three steps. From first joint conversation to production. Joint account planning, joint engineering, joint customer success, one accountable team.
HPE and Katonic SEs jointly map the customer's existing ProLiant, Cray XD or Aruba footprint to the Katonic agentic platform. One session. Reference architecture in the room.
Deliverables
A focused POC on the customer's most painful agent use case. Pre-validated on HPE Private Cloud AI, Katonic deployed in days not months, joint troubleshooting through both vendors' tier-2.
Deliverables
Single procurement vehicle through the HPE channel. Joint deployment. HPE AI Factory hardware, Katonic platform license, joint customer success motion.
Deliverables
HPE customers do not start from a blank canvas. Eighty production-grade agents ship with the platform. Adopt one as-is, or swap in your model, your MCP tools, your knowledge sources and your guardrails, and put it in production the same week.
End-to-end loan origination assistant. Pulls borrower history, runs policy checks, drafts the credit memo.
Document extraction, sanctions screening, risk scoring and SAR drafting on customer-owned hardware.
Ambient scribe + structured EHR draft. PHI stays on the HPE rack. NeMo guardrails enforced inline.
Multi-lingual citizen-services concierge with full audit trail. Deployable on air-gapped Cray XD.
Voice-first network ops assistant. Runbook recall, ticket drafting, escalation routing on Ezmeral fabric.
Multi-lingual care agent grounded on BSS/OSS, with per-tenant FinOps for telco neoclouds.
Contract review, vendor comparison, PO drafting. Plugs into ERP via MCP. Full spend lineage.
Employee self-service for policies, benefits and leave. PII scanned and redacted on every turn.
§ Compose · Model · MCP · Knowledge · Guardrails
Llama, Mistral, Granite or any HuggingFace OSS model served on NVIDIA NIM and vLLM on HPE GPUs. Or call Azure OpenAI, Anthropic, Bedrock. Katonic AI Gateway routes intelligently.
Native MCP server registry. Wire the agent into ServiceNow, Salesforce, SAP, Workday, Jira, Confluence, custom internal APIs, without writing tool-calling glue code.
One-click ingest from SharePoint, S3, Alletra MP, Confluence, Snowflake. Hybrid retrieval, per-tenant indexes, citations on every answer. Ezmeral fabric for unified governance.
NeMo guardrails plus Presidio PII, jailbreak detection, topic control and grounding checks. Eight rail types, all configurable per agent, all running on HPE.
† Every prebuilt agent ships with sensible defaults across all four slots. Customers override only what matters to them.
“We needed an on-prem AI platform with the operating discipline of a public cloud. HPE Private Cloud AI delivered the rack. Katonic delivered the agents. From PO to production in one quarter, without our data leaving the building.
NVIDIA technology: NIM, NeMo, NAT, KAI, MIG and Dynamo, deployed on HPE infrastructure and exposed as three surfaces by Katonic. One platform. Three jobs to be done.
NIM
Inference microservices
NeMo
Guardrails + RAG
NAT
Agent toolkit runtime
KAI
GPU scheduler
MIG
GPU partitioning
Dynamo
Inference orchestration
A 30-minute joint session with your HPE account team and Katonic solution architects.
Use cases, sizing, reference architecture and a 90-day delivery plan.
