SARION Monarch
An eight-GPU NVIDIA HGX B300 architecture with ~2.3 TB HBM3e for development teams and frontier-scale models with trillions of parameters. On your premises, or in Monaco. From €1,750,000.
Institutional scale
SARION Monarch brings the complete SARION stack to an eight-GPU NVIDIA HGX B300 architecture on a single, unified NVLink fabric. Designed for engineering teams, institutional research groups, and dozens of simultaneous concurrent users, Monarch runs our standard sovereign model line-up and unlocks frontier-scale models with parameters counted in the trillions, available as options. Fully owned, deployed on your premises or in a dedicated Monaco facility.
Parameters counted in the trillions, available as options, running on dedicated hardware.
NVIDIA HGX B300 baseboard with ~2.3 TB HBM3e unified across a high-speed fabric.
Engineered for development groups and high-concurrency enterprise workloads.
Agentic reach, email, calendar, tasks, text-to-image, and text-to-video generation.
No cloud dependency for inference, no subscription, no telemetry, complete ownership.
Our full standard model portfolio included, with larger frontier models available as options.
Flexible deployment
SARION Monarch is an enterprise-class rack-mounted appliance drawing approximately 14 kW under peak sustained load. For clients with suitable power, cooling, and facility infrastructure, our founding team installs Monarch directly on your premises. For clients who cannot host server-grade power, cooling, weight, or acoustic footprint on-site, Volta Sovereign AI provides deployment in a secure, state-of-the-art local data center in Monaco. In both cases, the hardware is 100% dedicated to you, owned by you as a capital asset, with zero shared resources and no third-party access. Hosting terms are agreed during confidential consultation.
What's included
Three years, hardware and labor, covering the complete server architecture.
Three years of direct assistance from the founding engineering team, no intermediary.
Personal installation by the founding team, on your premises or in the Monaco data center.
Common questions
Everything in Sovereign at institutional team scale, built around an eight-GPU NVIDIA HGX B300 system with ~2.3 TB HBM3e on a unified NVLink fabric. It supports dozens of concurrent users and runs frontier-scale models with parameters counted in the trillions, available as options.
From €1,750,000. Acquisition begins with a direct, confidential conversation with the founding team, the same process as SARION Regent and SARION Sovereign.
Yes. While Monarch is an on-premise appliance by design, clients who cannot host its power (~14 kW), cooling, weight, or acoustic requirements on-premise can have it deployed in a secure, dedicated local data center in Monaco. The hardware remains 100% owned by and dedicated to you, with zero shared infrastructure. Hosting terms are agreed during consultation.
No. Unlike Sovereign which networks two separate workstation nodes, Monarch is a unified single-chassis 8-GPU server engineered for maximum NVLink interconnect bandwidth and massive GPU memory pooling across all eight GPUs.
Delivery currently takes approximately six months from order to installation, driven primarily by current component lead times. Component pricing and delivery timelines are both subject to ongoing volatility in the global semiconductor and memory markets.
Next step
Every SARION Monarch acquisition begins with a conversation, in person or by video call. Leave your details below and Alexandre Ouaknine will respond to you directly.
Attorneys, family offices, and wealth managers: if a client's situation may call for this kind of conversation, reach out directly and confidentially.