AI / HPC capacity · Bizerte, Tunisia

Grid-ready AI capacity, one subsea hop from Marseille

Octantis is developing a 40 MW AI/HPC data center campus in Bizerte, at the landing point of the Medusa subsea cable to Marseille. Power and land are secured. The campus is designed for high-density GPU deployments, at a fraction of European power and build costs.

Discuss capacity

Artist's impression

40 MWgrid capacity attested by the national utility, 225 kV

~28–31 MWIT capacity at full build-out

< 30 msto London, < 15 ms to Marseille

~1.3annual average design PUE · liquid-cooled, zero water for cooling

Site & infrastructure

Power is the constraint on AI capacity in Europe. In Bizerte, it is already in place.

Power: 40 MW attested at 225 kV, with a dedicated connection to the national grid.
Land: secured in the Bizerte area, sized for the full 40 MW footprint.
Energy: renewable supply through a solar power purchase agreement, under negotiation.

Density: designed for high-density GPU racks with direct-to-chip liquid cooling.
Cooling: closed-loop liquid cooling with no evaporative towers, so no water is consumed for cooling.
Resilience: Tier III design, N+1 power and cooling, on-site generators.

Connectivity & jurisdiction

A neutral location at the southern edge of Europe, built for international workloads.

Map of subsea connectivity from Bizerte to Marseille, Paris, London and New York

< 15 msMarseille

< 20 msParis

< 30 msLondon

< 90 msNew York

< 120 msSingapore

Indicative round-trip latency from Tunisia; final figures depend on the routes contracted. Map lines are illustrative, not actual cable routes.

Subsea

Bizerte is a landing point for several submarine cables, including Medusa, in service since 2026, linking directly to Marseille, Southern Europe's main interconnection hub.

Jurisdiction

Tunisia holds formal ties with the United States, the European Union and China: neutral ground for international customers.

Workloads

Suited to workloads that do not require EU data residency: training, fine-tuning, batch inference and overflow capacity.

Engagement models

Capacity delivered in phases, from a first modular block to a dedicated campus.

Modular deployments

Powered, cooled and connected positions for prefabricated GPU blocks, starting at 500 kW to 1 MW per block and scaling to multi-MW deployments.

Wholesale capacity

Dedicated halls or MW blocks delivered in phases, with power, cooling and facility operated by Octantis.

Build-to-suit

A design adapted to a single anchor tenant's specification, with capacity reserved ahead of construction.

Let's talk

We are working with anchor tenants, technology partners and investors. To discuss capacity or partnership, contact Firas Mhedhebi, Founder & CEO.

fm@octantisai.com