01
HPC and AI hosting
Purpose-built halls for high power density, direct-to-chip and immersion cooling, and latency-optimised peering for training and inference workloads at industrial scale.
Institutional datacenter infrastructure
Sovereign datacenters, engineered for the compute that matters.
Ymir designs, builds, and operates institutional-grade capacity for AI training, inference, and high-performance computing — sited where energy is abundant, governed where jurisdiction is clear, capitalised on horizons that match the asset.
100 MW
Operational capacity
4
Active jurisdictions
99.99 %
Uptime SLA
1.20
Target PUE
Capabilities
One operating discipline, three commercial modes.
From dense AI training halls to long-tenor build-to-suit campuses, every Ymir asset is underwritten, designed, and operated by the same team.
Full capabilities register 
02
Enterprise colocation
Multi-tenant suites with sovereign-grade security, redundant power paths, and uptime commitments engineered for regulated institutional counterparties.
03
Build-to-suit delivery
Bespoke datacenter buildouts on long-term partnership economics — from substation to white space, delivered to specification under owner-side management.
Anatomy
A datacenter, delivered as one engineered object.
The Ymir container compresses substation-to-rack engineering into a single shippable unit. Four systems, one accountable owner, one commissioning event.
- POWER
- Industrial
- Up to 2.5 MW per container, long-tenor PPAs.
- COOLING
- Direct-to-chip
- Liquid + immersion supported beyond 100 kW per rack.
- COMPUTE
- Dense fabric
- Pre-engineered for the next two density generations.
- NETWORK
- Carrier-diverse
- Redundant spine-leaf, low-latency cross-connects.
Footprint • Q2 2026
Live capacity, audited every quarter.
100 MW
Operational capacity
Live across the operating portfolio.
4
Active jurisdictions
Gulf, Levant, Europe, and North Africa.
99.99 %
Uptime SLA
Tier III+ design and operating posture.
8760 h/yr
Resident operations
NOC and field engineering on every site.
Insights
Notes from the operating floor.
Essays on compute infrastructure, energy markets, and the capital cycle that funds them — written by the team that builds and operates.
All notesEssay
The compute capital cycle is here.
Why infrastructure spend is reshaping power and capital markets — and why the unit of digital infrastructure is no longer the rack, but the megawatt block.
Read note
Perspective
Why sovereign capacity matters in 2026.
On compute that does not cross borders, and why jurisdictional clarity has become a hard requirement for the institutions training frontier models.
Read note
Technical
Notes on cooling at the edge of physics.
Direct-to-chip, immersion, and what comes after — a field perspective on what is changing at 100 kW per rack and why thermal envelopes now dictate site selection.
Read note
Engage
Capacity requests are reviewed by the operating team.
Send a workload profile, sovereignty envelope, and target energise date. A feasibility note follows within five business days.