Real assets, measured
GPUs, CPUs, storage, networks, power, cooling and water. Cyclops reads the resource managers and meters a datacenter already runs.
Physical AI fintech · Engineered in Switzerland
Fintech for physical AI infrastructure.
AI runs on physical assets: GPUs, servers, power and cooling. Cyclops measures every GPU-hour, token and kilowatt-hour once, and uses it three ways: to bill your customers, to report your energy, and to price each unit above its true cost.
Billed usageCHF 418k96,400 GPU-h
PUE1.3268 % renewable
Cost to serve$2.51per GPU-hour
Physical AI fintech
Financial infrastructure for the machines that run AI
AI is built on hardware that is expensive to buy, power and cool. Cyclops connects that hardware to billing, reporting and pricing. Each GPU-hour, token and kilowatt-hour is measured once and used for the invoice, the ESG report and the price list.
AI workloads, accounted for
Training jobs, inference and LLM tokens on GPU clusters, supercomputers and sovereign clouds, with quantum platforms added as they come online.
Usage turned into money and margin
Usage is invoiced with a full audit trail and costed hour by hour, so operators earn revenue, price above cost and report with figures that match their books.
Our products
Three products on one platform
Datacenter operators need to earn a return on their compute, report the energy it uses and know what each unit costs. Cyclops does all three from the same data.
Bill every resource you sell
Meter, rate and bill HPC, GPU and cloud resources by the job, the hour or the token.
- GPU, CPU, storage, network and token usage
- Rating plans, credits and budgets
- Reseller and multi-party settlement
- Slurm, HEAppE, OpenStack and Kubernetes
Report the energy behind it
Measure facility and IT energy, and produce the data for regulation, disclosure and certification.
- Continuous PUE, with partial PUE by zone
- PUE, WUE, ERF and REF for EED Article 12
- Scope 1 and 2 emissions for CSRD
- Energy and emissions per tenant and job
Price every unit for margin
Know the true cost of every GPU-hour and token, hour by hour, and set prices above it.
- Cost to serve per unit, tenant and contract
- Price floors per GPU type and token tier
- Time-of-use tariffs from hourly power prices
- What-if scenarios for PPAs and refreshes
The challenge
Datacenters face pressure on revenue, reporting and margin
AI hardware has to earn back its cost, operators must report the energy it uses, and power prices change every hour. Most still handle all three with spreadsheets, flat fees and manual meter readings.
Unbilled usage
Shared GPU and HPC capacity is often charged by flat fee or estimate. Heavy users pay too little and costs are not recovered.
EED Article 12
Operators with 500 kW or more of IT power report energy and water KPIs to an EU database every year, by 15 May.
Hidden cost to serve
Leases, staff, support and depreciation sit in separate systems. Contracts can sell GPU-hours below their full cost without anyone noticing.
New units of sale
Customers now buy GPU-hours, tokens and model access. Billing systems built for VMs and storage do not handle these units.
CSRD and certification
Emissions, energy mix and measured evidence are needed for CSRD, the EU Taxonomy, LEED, BREEAM and green finance.
Power prices change every hour
On 14 October 2026 the German day-ahead price moved from €97 to €296 per MWh within one day. Flat GPU prices cannot follow that.
One pipeline, three outputs
From meter reading to invoice, ESG report and price list
Usage, energy and cost data share the same collection and accounting stages, then split into billing, ESG reporting and pricing. All three outputs use the same numbers.
Why Cyclops
Built for the scale of modern compute
Arbitrarily scalable
A horizontally scalable billing stack that grows with your infrastructure — no ceiling, no rewrite.
Exascale-ready
Track hundreds of millions of real and virtual resources across a true exascale system without losing fidelity.
Flexible integration
Plugs into Slurm, HEAppE, OpenStack, OpenNebula, Kubernetes and bespoke frameworks alike.
Every asset, metered
GPU clusters, CPU cores, networking, storage, IP addresses, firewalls — if it can be used, it can be billed.
Future-optimal
Already extending to quantum computing platforms and token-based LLM accounting in sovereign datacenters.
Built on European research
Born of European collaborative research projects in supercomputing and cloud, and built for operators across Europe and beyond.
CHF 1.5M
client invoices processed per quarter
4
EED KPIs: PUE, WUE, ERF, REF
Hourly
cost to serve per GPU-hour
engineered in Switzerland
Beyond classical HPC
Ready for the next era of compute
The same pipeline that meters a GPU cluster today is being integrated to monetize quantum computing platforms and to track LLM usage by the token in sovereign datacenters.
"See your compute clearly."
From classical cores to qubits, Cyclops turns the complexity of modern compute into something measurable, accountable, and genuinely approachable.
See supported technologies
Bill your compute. Report its energy. Price it for margin.
Start with the product you need most. All three run on the same data, so you can add the others later.