Choose your
Sovere system.
Two sizes of on-site AI. Explore what each system can run, the cost to get started and the managed AI service that keeps it improving. Your first service year is included.
Sovere S2: $159,000 initial purchase, including the first service year; $120,000 annual managed AI service from year two.
Sovere S2
S2 is our smallest AI server, suited to coding help, document checks and staff questions.
- HardwareOne-time payment
- $50,000
- Software, model setup & first-year serviceIncluded in your initial purchase
- $109,000
- Total initial purchaseComplete system + first service year
- $159,000
- Annual managed AI serviceModel improvements, optimisation & support
- $120,000per year, from year 2
Compare prices.
S2 and S4, side by side.
| Price |  Sovere S22 GPUs · 20–150 people |  Sovere S44 GPUs · 300–800 people |
|---|---|---|
| Hardware | $50,000 | $90,000 |
| Software, setup and 12 months of service | $109,000 | $135,000 |
| Year 1 total | $159,000 | $225,000 |
| Managed AI service, from year 2 | $120,000 a year | $144,000 a year |
| Per-token charges | $0 | $0 |
What your price includes.
Year by year.
Year 1 is one purchase, with everything included. From year 2, you pay only for the managed AI service, if you want it.
Compare with cloud AI.
Use your own numbers.
You save about $681,000 in three years. Cloud costs more than Sovere S2 from month 6.
Three years in totalCloud AI $1,080,000Sovere S2 $399,000
Pricing questions.
A complete system: hardware, Sovere software, initial model training and setup, agreed data connections, user access and deployment configuration. The software line also includes your first 12 months of managed AI service. The total is the hardware line plus this bundle; there is no additional first-year service fee.
Your first 12 months of managed AI service begin at system acceptance and are included in the initial purchase. The annual fee applies from year two. You are renewing the ongoing service, not buying the hardware and initial setup again.
Model training and upgrades, knowledge-graph improvements, inference and GPU performance optimisation, monitoring, security updates, testing on your tasks and engineering support. The standard scope covers one agreed application in one business domain, with four planned training and improvement cycles per year. Business-hours engineering support and on-call help for critical incidents are included; response targets and allowances are agreed in your proposal.
The initial price is a complete-system package that includes software, setup and the first service year. The renewal funds the continuing engineering work: improving models, training, optimisation, system care and support. Both prices are shown before you buy so you can plan your budget.
S4 is intended for larger workloads, such as voice or shared team use, with more performance tuning and operational work. The annual fee reflects the agreed service workload, not just the number of GPUs. We confirm the scope in your quote.
Yes, when the combined models and workload fit its capacity. The standard price includes one agreed application and business domain. Additional applications or domains, major retraining projects and extra capacity are quoted separately.
Included models running locally have no per-token charges. Additional hardware or services are quoted separately. External APIs or cloud compute you choose to use carry their own charges.
Prices are in USD, excluding tax. Delivery, site alterations, power and cooling, paid third-party licences, external APIs or cloud compute, and added hardware are separate. Your written quote confirms the configuration, service scope and any required extras.
Let’s choose your system.
Tell us what your team needs to run.