Compute Pricing Calculator — Tonomia

Pricing — Sovereign AI Infrastructure

Buy AI capacity the way your
organisation needs it.

Host your own hardware, rent GPU compute on demand, consume managed inference APIs, or deploy a ready-to-use enterprise AI assistant — all on TonoForge™ and TonoFabric™.

⚡

Rent GPU Capacity On Demand

Access advanced AI compute without waiting for procurement, installation or facility build-out. Designed for startups, enterprise AI teams, research organisations and platform builders that need immediate access to high-performance GPU infrastructure. Whether your workload is inference, fine-tuning, training or HPC — hourly, monthly or yearly — fully managed with TonoFabric orchestration and 24/7 SOC included.

AI StartupsResearch OrgsMLOps TeamsPlatform BuildersNo Hardware CapExImmediate Capacity

30%+

cheaper vs
hyperscalers

4.20 €

NVIDIA B300
per GPU/hr

2.90 €

AMD MI355X
per GPU/hr

0 €

orchestration,
NOC & SOC

01 — GPU Platform

NVIDIA

B300 Blackwell Ultra

4.20 EUR/GPU-hr

288 GB HBM3e · 8 TB/s
~5 PFLOPS FP8 · 15 PFLOPS FP4
NVLink 5 · ConnectX-8 800Gb/s

4nm BlackwellNVL72 rack

AMD

Instinct MI355X

2.90 EUR/GPU-hr

288 GB HBM3E · 8 TB/s
~5 PFLOPS FP8 · 10 PFLOPS MXFP4
Infinity Fabric · Pensando 400G

3nm CDNA 464-GPU rack

02 — Configuration

Number of GPUs8 GPUs

Utilization80%

Billing Period

03 — Workload

LLM Inference

Training

Fine-Tuning

RAG / Search

Vision / Multimodal

HPC / Simulation

—

Total PFLOPS (FP8)

—

Total HBM Memory

—

Power Draw (kW)

GPU Compute Estimate

NVIDIA B300 · 4.20 EUR/GPU-hr

💸

—% cheaper

vs equivalent hyperscaler GPU capacity — TonoFabric orchestration, 800 Gbps networking & 24/7 SOC all included

Tonomia Price

EUR 0

per month

GPU compute—

TonoFabric orchestrationIncluded

800 Gbps networkingIncluded

NVMe storage—

24/7 SOC + SLAIncluded

RL efficiency saved—

vs Hyperscalers

Tonomia

—

Hyperscaler 1

—

Hyperscaler 2

—

Hyperscaler 3

—

—vs hyperscalers

TonoFabric orchestration included — removes infrastructure friction while giving full flexibility in scale, duration and usage model. PUE 1.05 vs 1.3–1.6 for hyperscalers.

🧠

Consume Sovereign AI Inference by Token

Build AI products and internal services without operating the underlying model-serving infrastructure. Designed for software vendors, enterprise AI teams and regulated organisations that need predictable API access to sovereign inference — with flexibility in model strategy, deployment architecture and EU data residency. Managed, scalable, auditable.

SaaS PlatformsDocument AIRAG SystemsEnterprise CopilotsRegulated SectorsEU Data Residency

Up to
70%

cheaper per token
vs hyperscalers

0 €

egress fees &
model-serving overhead

sovereign residency
GDPR-native included

99.9%

API uptime
SLA guaranteed

Estimation Mode

Select model → adjust volume below

Volume & Billing

Tokens per Month10M tokens

Billing

≤8ms

p99 API Latency SLA

99.9%

API Uptime SLA

GDPR

EU Sovereign

Token Cost Estimate

EUR/MT = Euro per Million Tokens

💸

—% cheaper

per token vs hyperscaler reference — smart MaaS routing, GDPR compliance & EU data residency included

Tonomia Cost

EUR 0

per year

Token processing—

Model—

Smart MaaS routingIncluded

Audit logsIncluded

GDPR complianceIncluded

Tonomia vs Reference

Tonomia

—

Hyperscaler

—

— cheaperYou save —

Managed model access · inference API layer · routing and orchestration · versioned deployment · EU residency options. No egress fees. No model-serving overhead.

💬

Toomi™ for Teams — Sovereign AI, Ready to Deploy

A multilingual AI assistant built for organisations that want immediate value from AI across chat, document interaction, language tasks and knowledge workflows — without infrastructure procurement or API integration. Ideal for enterprises, SMEs and public sector teams that need a finished, secure, EU-sovereign application for everyday work.

EnterprisesSMEsPublic SectorHR & LegalNo Build RequiredEU Sovereign

−80%

vs market avg
25 €/user/month

5 €

Toomi Pro
per user/month

9 €

Toomi Pro+
per user/month

0 €

model serving &
GDPR residency

Choose Plan

Standard

TOOMI PRO

5 EUR

/ user / month

Model agnostic

Text & document translation

File Q&A · Audio/video transcription

Image analysis

Image generation (optional)

Report generation · Code execution

RAG grounded answers

Premium

TOOMI PRO+

9 EUR

/ user / month

Everything in TOOMI PRO

Image generation & edit included

Natural Continuous Conversation

Advanced bot customization

Workspace prompts/templates

Speech transcription · Extensions

Configuration

Number of Users100 users

Billing

Feature Comparison

Feature	TOOMI Chat	5EUR TOOMI PRO	9EUR TOOMI PRO+
Model agnostic	S	S	S
Text translation	S	S	S
Document translation	intact fmt	S	S
File Q&A	S	S	S
Audio/video transcription	S	S	S
Image analysis	S	S	S
Image generation & edit	S	O	S
Report generation	S	S	S
Code execution	S	S	S
Bot customization	S	S	S
RAG grounded answers	S	S	S
Multi-language chat	S	S	S
Enterprise data isolation	on-prem	S	S
Built-in web search	S	S	S
Speech transcription	S	S	S
Natural Continuous Conv.	S	O	S
Public API / connectors	X	—	—

S = Included · O = Optional · X = Not available

-80%

vs market avg (25EUR/user)

≤30s

Failover · ≥95% recovery

GDPR

EU Sovereign native

Toomi™ Deployment Estimate

TOOMI PRO · 5EUR/user/month

💸

—% cheaper

vs enterprise AI assistant market avg (25 EUR/user) — model serving, GDPR residency & 99.9% uptime included

Tonomia Total

EUR 0

per month

User licenses—

PlanTOOMI PRO

Model servingIncluded

GDPR data residencyIncluded

Tonomia vs Hyperscaler (25EUR/user)

Tonomia

—

Hyperscaler

—

–% cheaperYou save —

Powered by Toomi™ on TonoFabric™ — EU sovereign hosting, ≤30s failover, 99.9% uptime. Fast path to adoption, no infrastructure build required.

📊

Similar cost. Vastly more performance. Fraction of the cost per token.

The H100 remains the world’s most deployed AI GPU — but at Tonomia, the same budget buys you an NVIDIA GB300 or AMD MI355X, both delivering 3 to 5× more throughput at comparable hardware cost. More compute per euro means your cost per token drops dramatically — making Tonomia the most efficient platform for inference, training and fine-tuning at scale, without spending more than you already would on H100 infrastructure.

GB300: ~3.5× H100 FP8 throughputMI355X: ~4× H100 FP8 throughputSimilar hardware costUp to 70% lower cost per tokenPUE 1.05 · 100% renewable

3–5×

more throughput
vs H100 FP8

Up to
70%

lower cost
per token

288 GB

HBM3e per GPU
vs 80 GB H100

PUE
1.05

vs 1.4 industry
100% renewable

H100 Reference Price (your cloud rate)

H100 SXM5 EUR/GPU-hr

EUR

Tonomia H100 reference: 2.70 EUR/GPU-hr

Precision – MI355X

Precision – B300

NVIDIA – Reference

H100 SXM5 (FP8)

TFLOPS FP81,979 TFLOPS

Memory80 GB HBM3

Bandwidth3.35 TB/s

EUR / 1000 TFLOPS/hr—

AMD – Tonomia 2.90 EUR/hr

MI355X (MXFP4)

TFLOPS MXFP410,000 TFLOPS

Memory288 GB HBM3E

Bandwidth8.0 TB/s

EUR / 1000 TFLOPS/hr—

NVIDIA – Tonomia 4.20 EUR/hr

B300 Blackwell Ultra (FP4)

TFLOPS FP415,000 TFLOPS

Memory288 GB HBM3e

Bandwidth8.0 TB/s

EUR / 1000 TFLOPS/hr—

⚡

— × cheaper/TFLOP —

MI355X & GB300 vs H100 FP8 — same budget, 3–5× more throughput, dramatically lower cost per token

Cost Efficiency vs H100

MI355X cheaper per 1000 TFLOPS—

B300 cheaper per 1000 TFLOPS—

EUR / 1000 TFLOPS/hr — lower is better

H100 FP8

—

MI355X

—

B300

—

GB300 and MI355X deliver 3–5× more throughput than H100 at a comparable hardware price point — making Tonomia the most cost-efficient platform for inference, training and fine-tuning at scale.

DeepSeek V3 685B — Est. Cost per Million Tokens

Cost per MT from GPU-hour rate, minimum GPUs to fit model, bandwidth-weighted throughput. H100 price adjustable above.

GPU	Precision	Min GPUs	EUR/GPU-hr	Cluster/hr	Throughput	EUR/MT	vs H100

* H100: 80 GB, min 17 GPUs (FP8). MI355X/B300: 288 GB, min 5 (FP8), 3 (FP4). Throughput: bandwidth-limited decode model.

🏢

Bring Your Own Hardware into TonoForge™

Deploy your own servers inside a sovereign, high-density AI environment designed for speed, efficiency and operational simplicity. Ideal for enterprises, telecom operators, AI labs and cloud providers that want to deploy faster without the cost, delay and complexity of conventional AI facility development. Power, cooling, connectivity and 24/7 NOC — all included.

EnterprisesTelecom OperatorsAI LabsCloud ProvidersBYOH / BYODPilot → Production

52%

cheaper infra fee
vs classic colo

95 €

per kW/month
vs 200 € market

0.18 €

per kWh renewable
vs 0.24 € grid avg

PUE
1.05

vs industry
avg 1.4

01 — Rack Configuration

■

Small

150 kW

1 TonoForge rack

■■

Medium

300 kW

2 TonoForge racks

🛮

Cluster

600 kW

4 racks · 2 TonoForges

Number of Racks ×150 kW each1 rack · 150 kW

1 rack (150 kW) 50 racks (7.5 MW) 100 racks (15 MW)

02 — Deployment Timeline

1 week

1w2w3w 4w5w6w 7w8w / 2mo

Includes site survey, power connection, liquid cooling commissioning, network integration & TonoFabric onboarding. Factory-tested & pre-racked — plug-in ready from day one.

03 — Billing Period

04 — Energy & Sustainability

Tonomia · Renewable

PUE 1.05

0.18 EUR/kWh · 100% renewable
On-site battery buffer
Waste-heat recovery ready
Zero carbon-offset surcharge

Classic DC · Grid mix

PUE 1.4

0.24 EUR/kWh Belgian enterprise avg
PUE 1.4 · Mixed-source energy
No heat-recovery
Carbon offset often extra

150 kW

IT Power

0.15 MW

Total Capacity

0 t

CO₂ avoided/yr

05 — SLA Conditions

99.95%

Power Uptime

Dual UPS feeds + diesel generator. <4.4 h downtime/year contractually guaranteed.

N+1

Cooling Redundancy

Full N+1 CRAC/CRAH units. Direct liquid cooling on all TonoForge racks.

15 min

On-site Response

24/7 NOC monitoring. Engineer dispatched within 15 min for P1 incidents.

400 Gb/s

Network Uplink

Dual-carrier 400 Gb/s uplinks. BGP failover <60s. Zero inter-rack egress.

SOC 2

Security & Compliance

SOC 2 Type II · ISO 27001 aligned. Biometric access & 24/7 CCTV.

GDPR

Data Sovereignty

Belgium jurisdiction. EU data residency guaranteed. DPA signed at contract.

Infra-as-a-Service Estimate

TonoForge · 95 EUR/kW/month · BYOH

💸

52% cheaper

infrastructure fee vs classic DC colocation — 95 EUR/kW/month incl. power, liquid cooling, 400 Gbps, NOC & SOC

Tonomia IaaS Total

EUR 0

per year

Infrastructure fee (95 EUR/kW)—

Energy (0.18 EUR/kWh · PUE 1.05)—

TonoFabric orchestrationIncluded

Remote hands (8 h/month)Included

24/7 NOC + SOC monitoringIncluded

400 Gb/s dual-carrier uplinkIncluded

Your hardware (BYOH)You bring it

Tonomia vs Classic DC Colocation

Tonomia

—

Classic DC

—

— cheaperYou save —

Infrastructure fee: 95 EUR/kW/month (Tonomia) vs 200 EUR/kW/month (classic colo baseline). Energy billed separately: 0.18 EUR/kWh renewable (Tonomia) vs 0.24 EUR/kWh Belgian enterprise grid average. Includes power, liquid cooling, 400 Gb/s connectivity, rack integration, 24/7 NOC & SOC.

— saved vs classic DC colocation

GPU: AMD MI355X 2.90 EUR/GPU-hr · NVIDIA B300 4.20 EUR/GPU-hr · H100 ref 2.70 EUR/GPU-hr. Token pricing from Tonomia reference table. Chatbot: TOOMI PRO 5EUR · TOOMI PRO+ 9EUR/user/month. IaaS: 95 EUR/kW/month · 0.18 EUR/kWh renewable · BYOH. Sources: tonomia.com/pricing · tonomia.com/tonoforge. Estimates are indicative — contact for a tailored proposal.

Buy AI capacity the way yourorganisation needs it.

GPU Compute Estimate

Token Cost Estimate

Toomi™ Deployment Estimate

Infra-as-a-Service Estimate

Generate Non-Binding Letter of Intent

Buy AI capacity the way your
organisation needs it.