Next-Generation Data Center Cooling
Liquid Cooling Systems for the AI Era
Engineered in Oregon for high-density CPU, GPU, AI clusters, and crypto mining operations. Complete turnkey solutions from immersion tanks to full facility thermal management.
The Challenge
Why Can't Air Cooling Handle Modern AI Data Centers?
Air cooling fails at rack densities above 40–50 kW because it cannot remove heat fast enough from processors with 700W+ thermal design power. Liquid cooling delivers up to 1,000x greater heat transfer efficiency, making it the only viable solution for modern AI GPU clusters.
Modern AI accelerators like NVIDIA's H100 and B200 push thermal design power past 700W per chip. At rack densities exceeding 100 kW, fan walls grow louder, CRAC units multiply, and PUE climbs above 1.5 — consuming electricity that should be powering inference.
Direct-to-chip and full immersion cooling deliver the thermal headroom that next-generation silicon demands, cutting cooling-related energy by 30–50% compared to air-based systems.

Engineered Solutions
What Liquid Cooling Solutions Does MRX Offer?
MRX provides three tiers of data center cooling — direct-to-chip cold plates for individual processors, full immersion tanks for entire server boards, and complete facility thermal infrastructure — all designed and fabricated in Canby, Oregon.
Every system is ASME-coded, UL-Listed, and delivered turnkey with commissioning support.
Direct-to-Chip (D2C) Cooling
For High-Performance CPUs & GPUs
Precision-machined cold plates mounted directly on processors transfer heat to liquid coolant with unmatched efficiency. Our D2C systems include custom manifolds, coolant distribution units (CDUs), and leak-detection arrays — all engineered and assembled in Canby, OR.
- Custom cold plate design for any GPU/CPU socket
- Integrated manifolds with redundant flow paths
- Factory-tested CDUs with intelligent monitoring
- Compatible with H100, B200, MI300X, and custom ASICs
Full Immersion Cooling Tanks
For AI Clusters & Crypto Miners
Complete single-phase and two-phase immersion cooling systems that submerge entire server boards in dielectric fluid. Achieve the lowest possible PUE while eliminating fans, dust, and hot-spot failures entirely.
- Custom ASME-coded immersion tanks
- Single-phase and two-phase fluid support
- Integrated heat rejection and fluid management
- Designed for 24/7 unattended mining and AI training

Facility-Scale Thermal Management
Complete Data Center Infrastructure
End-to-end facility cooling infrastructure including cooling towers, chillers, plate-and-frame heat exchangers, and power integration. MRX delivers complete thermal plants as turnkey packages — engineered, fabricated, and commissioned by a single team.
- Cooling towers and dry coolers
- Industrial chillers and heat exchangers
- Power distribution and UPS integration
- Full commissioning and performance validation
Direct-to-Chip vs Immersion Cooling: Which Is Right for Your Facility?
Direct-to-chip cooling is best for retrofitting existing air-cooled racks, while immersion cooling delivers maximum density and efficiency for greenfield AI training deployments. Both are ASME-coded and UL-Listed when built by MRX.
| Feature | Direct-to-Chip (D2C) | Full Immersion |
|---|---|---|
| Max Rack Density | 80–120 kW | 150–250+ kW |
| Target PUE | 1.10–1.15 | 1.02–1.06 |
| Retrofit Friendly | Yes — fits existing racks | No — requires custom tanks |
| Component Coverage | CPU/GPU only | All components submerged |
| Maintenance Access | Standard rack access | Requires fluid draining |
| Best For | Mixed workloads, retrofit | AI training, max density |
| ASME Compliance | CDU vessels coded | Full tank coded |
Written by the MRX Engineering Team — Canby, Oregon
The MRX Difference
Why Choose MRX for Data Center Cooling?
Not every cooling vendor can match ASME fabrication quality with UL-Listed controls expertise. MRX does both — under one roof in Canby, Oregon.
ASME 2026 Compliant Vessels & Tanks
Every immersion tank and pressure vessel is built to current ASME BPVC standards — Oregon engineered, U-stamped, and ready for inspection.
UL-Listed Intrinsically Safe Controls
All control panels, monitoring systems, and safety interlocks are designed and UL-Listed in-house. No third-party panel shops.
Full Turnkey Integration
From thermodynamic modeling through fabrication, controls, power distribution, and on-site commissioning — one team, one responsibility.
Proven in Regulated Industries
Over a decade of success in pharmaceutical, food, biotech, and other high-safety sectors. We bring that rigor to every data center project.
Key Takeaways — Data Center Liquid Cooling
- Liquid cooling delivers up to 1,000x greater heat transfer than air, enabling 100+ kW rack densities for AI GPU clusters.
- Direct-to-chip (D2C) cooling retrofits existing racks; full immersion cooling eliminates fans entirely and achieves PUE below 1.05.
- Every MRX cooling vessel is ASME BPVC 2026 compliant with U-stamp certification and UL-Listed intrinsically safe controls.
- Turnkey scope: thermodynamic modeling through commissioning — one team, one facility, one point of accountability.
- MRX supports NVIDIA H100/B200, AMD MI300X, Intel Gaudi, and custom ASIC platforms with precision-machined cold plates.
Common Questions
Data Center Cooling FAQ
Answers to the most common questions about liquid cooling for AI, HPC, and cryptocurrency mining data centers.
Know your GPU platform and rack count? Get an engineering-grounded cooling proposal.
Tell us your heat load, fluid preferences, facility constraints, and timeline. No generic spec sheets — we respond with real engineering.
Data Center Liquid Cooling Cluster
Technical Resources in This Cluster
Technical Guide
Liquid Cooling Strategies for High-Density AI Compute (100kW+ Racks)
Technical Guide
Direct-to-Chip vs. Immersion Cooling for High-Density GPU AI Clusters
Technical Guide
Pump Skid Engineering Guide
Technical Guide
Heat Exchanger Thermal Management Guide
Oregon Advantage
Supporting Oregon's Semiconductor & AI Growth
Case Study
AI Compute Liquid Cooling Infrastructure — High-Density Deployment
Ready to Cool Your Next-Generation Infrastructure?
Whether you're scaling an AI training cluster, deploying a new GPU farm, or upgrading existing cooling infrastructure, MRX engineers turnkey solutions from our facility in Canby, Oregon.
Oregon Engineered • ASME BPVC 2026 Compliant • UL-Listed Intrinsically Safe Controls