AI Infrastructure

Planner

TCO Analysis Dashboard

Comprehensive total cost of ownership comparison between on-premises and cloud AI Clusters. Select a preset or enter your own configuration to see the cost breakdown.

TCO Comparison: Cloud vs. On-Premises

3 YEARS ANALYSIS

Analysis Period

Calculating TCO Analysis

Fetching cloud pricing data and computing cost comparisons for your configuration...

Cloud Total Cost

$246,227.00

$6,839.64/month avg.

On-Premises Total Cost

$2,527,591.00

$70,210.86/month avg.

Recommended Option

CLOUD

Cloud infrastructure is more cost-effective for your use case, saving approximately $2,281,364.00 (90.3%) over 3 years.

Savings
90.3%
Payback Period
Beyond analysis period
On-Premises Cost Breakdown
Compute
GPU Servers • Control Servers
$400,000/GPU server • $30,000/control server
$0.00
Network
Switches • Optics
$100,000/switch • $1,000/optic
$0.00
Storage
Storage Servers
$30,000/storage server
$0.00
Power & Cooling
24/7/365 Operation • PUE: 1.5
$0.120/kWh
$0.00
IT Labor & Admin
2x Platform Engineers • 1x Network Engineers • 30% Benefits
Platform: $150,000/year • Network: $130,000/year
$0.00
Facilities
Data Center Rack Space (13 racks, limited by power)
$1,200/rack/month • 30kW per rack
$0.00
Maintenance & Support
Hardware & Software Support
10% of hardware cost
$0.00
Software
Data Platform • Misc Software
Data Platform: $30,000/year • Misc: $20,000/year
$0.00
Total$0.00
Cloud Cost Breakdown
EC2 Compute (p5en.48xlarge)
32x instances • on-demand
On-Demand: $63.296/hour per instance
Effective: $63.296/hour per instance
$0.00
EBS Storage (gp3)
64TB • General Purpose SSD
$0.080/GB/month
$0.00
Additional Storage Services
FSx for Lustre
1024TB • Scratch Storage
$0.140/GB/month
$5,284,823.04
Enterprise Support
AWS Enterprise Support Plan
10% of net AWS usage
$0.00
Cloud Architecture & DevOps
Cloud architects • 30% Benefits
$180,000/year per engineer
$0.00
Total$0.00

Workload Presets

Configuration Details

Extra large cluster for enterprise training - 32 nodes with 256 H200 GPUs

GPU Type:H200
Total GPUs:256
Nodes:32
RAM:16384GB
Storage:64000GB

GPU Configuration

GPU Specifications

This will require 32 GPU servers (8 GPUs per server)

Affects both hardware pricing and cloud instance selection

GPU Configuration Summary

Total GPUs
256
H200 GPUs
GPU Servers Required
32
8 GPUs per server
Cloud Instance Type
p5en.48xlarge
32 instances needed

On-Premises Configuration

Hardware Breakdown

Calculated: 256 total GPUs ÷ 8 GPUs per server = 32 servers

Hardware Maintenance

On-Premises Software

Personnel

Facilities Configuration

Cloud Configuration

Instance Configuration

Pricing Configuration

Choose your preferred pricing model

Enter your negotiated enterprise discount

AWS Enterprise Support fee percentage

Additional Storage

Personnel

Cloud Software Configuration

Calculating Costs

Updating summary...

Cloud Infrastructure

Total Cost$0.00
Monthly Average$0.00
NPV$0.00

On-Premises Infrastructure

Total Cost$0.00
Monthly Average$0.00
NPV$0.00

Comparison

Cost Difference$0.00
Savings PercentageNaN%
Break-even PointBeyond analysis period

Key Assumptions

GPU Servers and AWS instances are all assumed to have 8 GPUs. 2-4 GPU configurations per node is roadmapped.

Management software for compute and networking is included in the per unit cost.

Personnel costs include benefits and overhead at 30% of base salary.

Power consumption of 10kW per GPU Server, 2.4kW for Control & Storage Servers. 2.4kW for Switches.

On-premises infrastructure has 24/7/365 100% utilization.

Management network not included in the cost calculation.

B200 GPU AWS Instances assumed at the same price as H200. Pricing currently unavailable via API.

Roadmap

Launch v1.0!

2-4 GPU configurations per node support

Add cost per 1M token pricing

Add Azure & GCP pricing integrations

Advanced networking cost modelling