Rognix
Pricing

Built for AI infrastructure. Affordable for everyone underneath.

The AI tiers below are sales-led, every deployment starts with a 30-minute review and a read-only assessment. The standard server tiers are self-serve, with a free forever plan for monitoring up to 3 nodes.

AI Infrastructure · Sales-led

For 16+ GPU clusters and dedicated AI data centers.

These tiers are built for organizations where GPU infrastructure is the business, tier-2 GPU clouds, sovereign AI deployments, ML platform teams, and dedicated AI data centers operating tens to hundreds of accelerators across multiple racks. If you're running a single training box or a handful of dev VMs, the standard server tiers below are the right fit. Every AI deployment starts with a 30-minute review and written assessment, final pricing factors in node count, support tier, deployment model (SaaS / dedicated / air-gapped), and data residency.

Most teams start here
AI INFRASTRUCTURE

AI Cluster

Built for serious GPU operators. 16+ accelerators, multi-tenant training fleets, ML platform teams running production clusters. Not for individual developers.

$999/mo starting
$9,990/yr · final pricing after deployment review
GPU nodes
up to 16
Standard nodes
+30
  • Up to 16 GPU nodes (H100 / A100 / L40S / RTX-class)
  • Plus 30 supporting servers
  • Per-GPU utilization, thermals, power
  • Thermal-runaway prediction
  • Cluster cooling imbalance detection
  • Underutilized GPU detection ($/hr wasted)
  • AI workload failure prediction
  • Slurm / Kubernetes / Ray integration
  • Priority support, 4h response
Request review
AI INFRASTRUCTURE

AI Datacenter

For dedicated AI data centers, sovereign clouds, and tier-2 GPU cloud operators. Hundreds of GPU nodes across multiple racks and regions, full BMC / IPMI / Redfish integration, per-tenant chargeback. This is the tier you buy when GPU infrastructure is the business.

$4,999/mo starting
$49,990/yr · final pricing after deployment review
GPU nodes
up to 128
Standard nodes
+250
  • Up to 128 GPU nodes
  • Plus 250 supporting servers
  • Rack-level thermal + power view
  • PUE / cooling efficiency analytics
  • BMC / IPMI / Redfish integration
  • Per-tenant chargeback reporting
  • NVLink / InfiniBand fabric health
  • SOC2 audit export + SSO
  • Dedicated support engineer
  • 99.95% uptime SLA
Request review
Book deployment review See AI capabilities
Standard servers · Self-serve

Running standard infrastructure too?

For VPS, on-premise, and cloud servers without GPUs. Free forever for the first 3 servers. Card on file required only at the Starter tier and above.

Free

Monitor standard servers, forever.

$0/forever
  • Up to 3 standard servers
  • Real-time monitoring
  • 10 AI insights / month
  • Manual mode
Start monitoring free

Starter

For solo builders running standard infra.

$19/mo
$190/yr · save 17%
  • Up to 10 standard servers
  • 200 AI insights / month
  • Assisted mode
  • Slack + email + webhook
Start with Starter

Team

For ops teams with standard production fleets.

$79/mo
$790/yr · save 17%
  • Up to 50 standard servers
  • 1,500 AI insights / month
  • Autonomous mode + approval queue
  • Per-template action permissions
Start with Team

Business

Multi-environment standard ops at scale.

$299/mo
$2990/yr · save 17%
  • Up to 200 standard servers
  • 8,000 AI insights / month
  • SSO (OIDC + SAML)
  • SOC2 audit log export
Start with Business
Enterprise · Frontier-scale

Enterprise

For hyperscalers and sovereign AI deployments. Self-hosted control plane, sovereign / air-gapped deployments, FedRAMP & SOC2 review support, dedicated solutions engineer.

Talk to founders

Common questions

What's a deployment review?+
A 30-minute call with our team plus a read-only assessment of one rack of your fleet. We ship you a written report within 5 business days identifying three opportunities, thermal, utilization, and workload. No commitment.
Why is the AI tier sales-led but standard servers are self-serve?+
AI deployments have specific topology, fabric, and workload requirements that benefit from a brief conversation up front. Standard servers don't, drop the agent, you're monitoring.
What does the GPU collector actually do?+
Reads NVIDIA-SMI / DCGM / ROCm equivalents, utilization, memory pressure, temperatures, power draw, NVLink/PCIe error counts. Read-only. No process lists, no model weights, no training data, no checkpoints. Optional Slurm / Kubernetes / Ray hooks bring workload context.
Can we run Rognix in our own VPC or air-gapped?+
Yes, on the AI Datacenter and Enterprise tiers. Self-hosted control plane image, signed container, with documented update procedure for sovereign / classified deployments.
What about a free tier for AI servers?+
The standard Free tier monitors any Linux server including those with GPUs, but GPU telemetry, thermal analytics, and workload failure prediction unlock at the AI Cluster tier.
Discount for research labs / universities / open-source AI?+
Yes, email hello@rognix.com with a one-paragraph case. Usually approved within a day.