Pricing

Reserved clusters, priced per GPU-hour.

Sapience sells dedicated NVIDIA Blackwell capacity on reservation, at cluster level: 8, 24 or 32 servers. The rate is per GPU-hour for the term, in USD, excluding VAT, fixed in the contract. Rates depend on term, cluster size and support scope and are confirmed in a written quote.

Get a quote Configure a cluster
UnitPer GPU-hour, for the term
MinimumOne cluster: 8 servers, 64 GPUs
CurrencyUSD, excluding VAT
Cluster sizes8 · 24 · 32 servers
InvoicingPayment and invoicing terms are set out in your quote.
How pricing works

Two nodes, one way to buy.

Every cluster is reserved for a term at a per-GPU-hour rate fixed in the contract. Shorter terms carry higher rates; larger clusters and longer terms carry lower rates. Every rate is quoted in writing.

GPU node GPUs × memory per server Cluster sizes Reservation terms Rate Availability Action
NVIDIA RTX PRO 6000Blackwell Server Edition
8 × 96 GB GDDR7
768 GB per server
8 · 24 · 32 servers
64 · 192 · 256 GPUs
From 3 months
longer terms on request
Quote
per GPU-hour, fixed for the term
Fully reserved through 2026
reservations for 2027 open
NVIDIA HGX B300Blackwell Ultra
8 × 288 GB HBM3e
2.3 TB per server
8 · 24 · 32 servers
64 · 192 · 256 GPUs
Multi-year
term agreed per contract
Quote
per GPU-hour, fixed for the term
Reservations open
first deliveries from Dec 2026
Clusters are sold whole. The minimum reservation is one 8-server cluster (64 GPUs). Single servers and single GPUs are not offered; that follows from how the capacity is financed.
What sets the rate. Rates depend on term, cluster size and support scope and are confirmed in a written quote. Reseller and agent pricing is set separately.
What is included. Dedicated servers, local NVMe, the node fabric, encryption and 24/7 monitoring and support. Storage add-ons and private-cluster arrangements are quoted separately.
Configure a cluster

Size it, then ask for the rate.

Choose the node, the cluster size and the term. The configuration goes straight into the quote form; the rate comes back in writing.

Node
Cluster size
Term
HGX B300 is reserved on multi-year terms, so terms under a year are unavailable for it.
Your configuration NVIDIA RTX PRO 6000
Servers
8
GPUs
64
GPU memory
6.1 TB (96 GB per GPU)
Term
3 months
Earliest start
January 2027
Rate on quote

Rates depend on term, cluster size and support scope and are confirmed in a written quote.

Included in the rate

What the per-GPU-hour rate covers.

Dedicated servers
The whole cluster, single-tenant, with direct hardware access and no virtualisation layer.
Local NVMe storage
Encrypted NVMe on every server (AES-256 at rest).
Cluster fabric
Node-to-node networking on the Bratislava fabric; 400G for multi-node scaling.
Operations
24/7 monitoring and technical support from the Sapience team.
Encryption
TLS 1.3 in transit, AES-256 at rest.
Quoted separately

What is confirmed in the quote.

The rate
Per GPU-hour for your term, cluster size and support scope, fixed in the contract.
Additional storage
Capacity beyond local NVMe, sized to the workload.
Private clusters
Isolated environments with contractual data residency for regulated customers.
Connectivity
Dedicated connectivity and data transfer arrangements.
Managed platform layers
Virtual machines, Kubernetes and SLURM are in development on the same infrastructure. Platform roadmap →
Reservation terms

Short terms for RTX PRO 6000. Multi-year for HGX B300.

RTX PRO 6000 reservations

From 3 months

Clusters of 8, 24 or 32 servers reserved for three months or longer at a rate fixed for the term. The current capacity is fully reserved through 2026; reservations for 2027 are being taken now, with the start date confirmed in the quote.

Inference, agentic AI and visualisation workloads Longer terms on request

HGX B300 reservations

Multi-year

Training-scale clusters committed for a multi-year term at a rate fixed in the contract. Longer terms carry lower rates; the rate for your term, cluster size and support scope is confirmed in the quote.

Reservations open now First deliveries from December 2026
Questions

Pricing questions.

Anything else: sales@sapienceai.eu or +421 233 329 562.

Is there a minimum?

Yes: one cluster of 8 servers (64 GPUs). Clusters come in 8-, 24- and 32-server sizes and are sold whole; single servers and single GPUs are not offered.

Can I buy on demand, by the hour?

Not at present. Capacity is sold on reservation for a term, which is what lets us finance and commission dedicated clusters. The rate is still expressed per GPU-hour so you can compare it.

Why is there no price list?

Rates move with term, cluster size, support scope and the market, and we would rather quote you the current rate in writing than publish one that is out of date.

Which currency, and is VAT included?

Rates are in US dollars and exclude VAT. VAT is applied according to your billing entity and location.

When can HGX B300 clusters be delivered?

Reservations are open now. First deliveries are scheduled from December 2026; the delivery date for your cluster is confirmed in the contract.

When is RTX PRO 6000 capacity available?

The current capacity is fully reserved through 2026. Reservations for 2027 are open; tell us your start month and we confirm availability in the quote.

Where do the clusters run?

In Bratislava, Slovakia, on systems operated by Sapience. Customer data is processed and stored there; see the Security page for the controls.

Get a quote

Tell us what you need.

Node type, cluster size, term and start month. We reply with a written quote: rate, term, delivery date and confirmed availability. If you would rather talk it through first, contact our team.

RTX PRO 6000: terms from three months; 2027 start dates. HGX B300: multi-year terms; start from December 2026. Clusters of 8, 24 or 32 servers. Rates in USD, excluding VAT.
Used only to answer your request — see the privacy policy. We reply from sales@sapienceai.eu. No newsletter, no automated sequences.