Who We're Built For

Workloads That Demand More Than Standard Colocation

AI training, HPC research, and enterprise compute have outgrown what conventional datacenters were designed for. Thermae Computatio is purpose-built for the density, power, and thermal requirements these workloads actually need.

Primary Workload Categories

Three workload categories drive the facility's design requirements. Each has distinct infrastructure needs — all of which Thermae Computatio is planned to support.

AI Training

Large-Scale Model Training

Training frontier AI models demands sustained high-density power, fast GPU-to-GPU interconnects, and thermal infrastructure that can handle continuous full-load operation. Thermae Computatio is designed with these requirements as first-class constraints.

Infrastructure requirements

  • High per-rack power density (up to 50kW+ target)
  • Low-latency, high-bandwidth GPU fabric (up to 400G+ planned)
  • Direct liquid cooling for sustained thermal load
  • Stable, uninterrupted power with N+1 redundancy target

Example workloads

  • Foundation model pre-training runs
  • Fine-tuning and RLHF pipelines
  • Multimodal model training (vision, audio, text)
  • Distributed training across multi-node GPU clusters
HPC Research

High-Performance Computing Workloads

Scientific computing, simulation, and research workloads require dense compute with predictable performance and low-carbon energy sourcing — increasingly important for grant compliance and institutional sustainability commitments.

Infrastructure requirements

  • High-throughput compute with consistent performance
  • Low-carbon grid for research sustainability reporting
  • Flexible rack configurations for diverse HPC hardware
  • High-bandwidth storage interconnects

Example workloads

  • Molecular dynamics and protein folding simulations
  • Climate and atmospheric modeling
  • Computational fluid dynamics (CFD)
  • Genomics and bioinformatics pipelines
Private Colocation

Dedicated Enterprise Colocation

Organizations requiring dedicated, private infrastructure — with physical security, compliance-aligned controls, and direct access to their hardware — can deploy in a purpose-built environment designed for high-density workloads.

Infrastructure requirements

  • Dedicated cage or suite options planned
  • Physical access control and audit logging
  • Remote hands and eyes service
  • Per-rack power metering for accurate billing

Example workloads

  • Private AI inference clusters
  • Regulated-industry compute (healthcare, finance)
  • Sovereign or data-residency-sensitive workloads
  • Enterprise disaster recovery and backup infrastructure

Why Thermae Computatio

The combination of infrastructure design, grid carbon profile, and waste heat economics creates advantages that are structural — not dependent on incentives or offsets.

Power Density

Up to 50kW+ per rack target supports the GPU configurations that AI training actually requires — not the 8–12kW average of legacy facilities.

Liquid Cooling

Direct liquid cooling as the baseline design — not a retrofit. Enables sustained full-load operation without thermal throttling.

Network Throughput

Up to 400G+ optical networking planned for intra-cluster traffic. Designed for the east-west bandwidth patterns of distributed AI workloads.

Low-Carbon Grid

~70% of Washington State's grid from renewables. Structural low-carbon advantage for Scope 2 reporting — no RECs required.

Tenant Profiles

We're actively engaging with organizations across these profiles. If your requirements don't fit neatly into a category, reach out — we're in the design phase and can accommodate specific needs.

AI Labs & Startups

Early-stage and growth-stage AI companies that need dense GPU infrastructure without the capital commitment of building their own facility. Access to high-density power and liquid cooling from day one.

Good fit for

  • GPU cluster colocation
  • Burst capacity for training runs
  • Low-carbon compute for ESG reporting

Research Institutions

Universities, national labs, and research consortia requiring HPC infrastructure with verifiable low-carbon energy sourcing — increasingly required for federal grant compliance and institutional sustainability targets.

Good fit for

  • NSF / DOE grant-compliant low-carbon compute
  • Dedicated HPC cluster deployments
  • Long-term research infrastructure partnerships

Enterprise AI Teams

Large enterprises building internal AI capabilities who need dedicated, private infrastructure with the security controls and compliance posture required by their legal and IT teams.

Good fit for

  • Private AI training and inference infrastructure
  • Compliance-aligned physical controls
  • Dedicated power and cooling capacity

Regulated Industries

Healthcare, financial services, and government-adjacent organizations that require data residency, physical security, and audit-ready infrastructure for sensitive workloads.

Good fit for

  • HIPAA-compatible physical controls (planned)
  • SOC 2 audit readiness (planned)
  • Data residency in Pacific Northwest

Inference Operators

Companies running large-scale AI inference who need reliable, high-density power and cooling with low-carbon sourcing — and who benefit from the operational cost advantages of Washington State's grid.

Good fit for

  • High-density GPU inference racks
  • Low-carbon Scope 2 emissions profile
  • Stable, long-term capacity planning

Development Partners

Strategic partners interested in co-developing the facility, contributing capital, or participating in the heat reuse and hospitality asset ecosystem adjacent to the datacenter.

Good fit for

  • Anchor tenant partnerships
  • Capital and development co-investment
  • Heat Purchase Agreement participation

Colocation Options

Three deployment formats are planned, from individual rack colocation to fully private suites. Final configurations will be confirmed through the design and development process.

Dedicated Cage

Private, locked cage within the shared datacenter floor. Physical separation with your own access controls.

  • Dedicated physical perimeter
  • Your own access credentials and audit log
  • Scalable from a few racks to a full cage
  • Shared facility services (power, cooling, network)

Private Suite

Fully enclosed, dedicated room with independent access control — the highest level of physical isolation available.

  • Floor-to-ceiling separation from shared space
  • Independent HVAC and power distribution options
  • Suitable for regulated-industry and sovereign workloads
  • Custom fit-out options under evaluation

Rack Colocation

Individual rack or partial-rack deployments within a shared, secured area. Ideal for smaller deployments or initial evaluations.

  • Single rack to multi-rack configurations
  • Shared physical security perimeter
  • Full access to facility power and cooling infrastructure
  • Flexible term options planned

Planning-stage disclosure: Colocation formats, pricing, and availability are not yet finalized. Thermae Computatio is in active development. We're engaging with prospective tenants now to shape final specifications — early conversations directly influence facility design.

Tell Us About Your Workload

We're in the design phase and actively shaping the facility around tenant requirements. Early conversations matter — reach out to discuss your compute needs.