Castellan is a European company developing and operating sovereign-grade AI infrastructure — headquartered in Switzerland, with operations across Europe and the United States. It is built on the operating DNA of the Bitkern Group — an established power & compute infrastructure operator — with access to +5 GW of interconnected power.
Castellan builds and operates the infrastructure your AI workloads run on — powered land, tenant-ready campuses, and sovereign GPU capacity under Swiss and European jurisdiction.
It is capacity the market barely offers: AI compute is becoming the most consequential infrastructure of the century, yet it sits concentrated in the hands of a few hyperscalers — a strategic exposure for every institution whose workloads cannot travel. Data sovereignty is not a slogan.
Castellan exists to close that gap — with the physical layer for sovereign-compliant AI, built and operated from Switzerland. Read the full vision →
Castellan does not start from zero. It is built on the operating DNA of the Bitkern Group — an established operator of large-scale power & compute infrastructure, with a decade of track record across energy and powered land.
Castellan has access to +5 GW of power, available as demand requires, alongside established energy-supply relationships — a position most newcomers simply cannot reach.
Most entrants start from zero on power. Castellan starts with the hardest asset already in hand. Our heritage in full →
Land secured on long-term tenure, with grid interconnection and access to power — the scarcest input in AI. For operators who want to build on their own terms.
From design to ready-for-rack — electrical, cooling, and construction planned, managed, and delivered end-to-end. We run the project; you own the result.
Land, power, and built infrastructure delivered tenant-ready on a long-tenor triple-net lease. Our core offering — everything in one, fast time-to-power.
Blackwell-class GPU systems, networking fabric, and storage — sourced through established OEM and distribution relationships. Pure GPU trading — from single cards to complete, integrated racks — with allocation access, price discipline, and delivery logistics handled end-to-end.
Bare-metal GPU clusters and reserved capacity blocks under Swiss and European jurisdiction — for training, fine-tuning, and inference. Compute you can contract, without building anything.
Your AI hardware runs below capacity? Castellan brings contracted offtakers and turns idle GPU-hours into revenue — for fleets, data-centre capacity, power, and sites.
Colocation in data centres across Europe and beyond — high-density racks for H100 to GB300 systems, DLC-ready cooling, 24/7 operations and remote hands. Your machines, our uptime — with optional monetization of idle time.
Every country needs to have its own sovereign AI infrastructure.
First site — Phase I is a minimum 50 MW greenfield campus in the USA or EU (site TBD); construction 2027, commissioning 2028 — Castellan's first triple-net, sovereign-aligned campus.
Six operators across power, engineering, commercial, legal, finance, and platform IT — a decade of running energy-intensive compute together at the Bitkern Group.
Meet the teamA note on the structural exposure of regulated workloads to hyperscaler concentration, and what neutral, sovereign-aligned infrastructure looks like.
Power-zone economics, interconnection queues, and why pre-secured power is the real unlock for AI capacity.
What the real-estate logic of AI campuses gives tenants: predictable costs, control, and security of tenure.
Tell us what you need. Every enquiry reaches our team and is routed to the right person.
Whether you're a prospective tenant, an enterprise looking for GPU compute, hardware, or hosting, or a land or construction partner — we're glad to talk.
Prefer email? info@castellan-intelligence.com
Press press@ · Careers careers@