Immersion Server Demonstrated Reliability at Datacenter Scale Deployment (OCP EMEA Summit 2025)
Open Compute Project Foundation · Austin Hipes
A field report presented at the OCP EMEA Summit 2025 tracks the reliability of about 3,600 single-phase immersion-cooled servers in a tier-1 data centre in Asia-Pacific from 2023 to early 2025.
On this page
Overview
The deck was presented by Austin Hipes in the Immersion Cooling track of the OCP EMEA Summit 2025 in Dublin. It reports the field reliability of about 3,600 single-phase immersion-cooled Xeon servers in a tier-1 data centre in Asia-Pacific, tracked quarterly from 2023 to early 2025. The deployment consists of 3,590 servers: 2,932 2nd-gen and 658 3rd-gen Xeon. They sit in single-phase immersion tanks with PAO4 dielectric, run 24x7 and were deployed between 2H 2022 and 2H 2023. The facility hosts air-cooled, direct-to-chip and immersion servers side by side.
Failures are broken down by component, with failures caused by the air-to-immersion conversion and by cable choice separated out. All reported failures number 202 on 2,932 2nd-gen servers (about 0.9% average per quarter) and 76 on 658 3rd-gen servers (about 1.9%). Excluding conversion-related and DAC cable failures leaves 133 failures (0.6%) for 2nd gen, of which memory is 36% and HDD 23%.
Downtime is converted into availability: fleet uptime is above 99.999%, with about 443 million power-on hours and 1,112 h total downtime. Memory and drives dominate failures. Limit: it covers one facility and one hardware family, yet it is one of the few published fleet-scale reliability datasets for immersion.
Key points
- 3,590 servers (2,932 2nd-gen and 658 3rd-gen Xeon) in single-phase immersion tanks with PAO4 dielectric, 24x7 duty, deployed 2H 2022 to 2H 2023.
- All reported failures: 202 on 2,932 2nd-gen servers (about 0.9% average per quarter) and 76 on 658 3rd-gen servers (about 1.9%).
- Excluding conversion-related and DAC cable failures: 133 failures (0.6%) for 2nd gen; memory 36% and HDD 23%.
- Fleet uptime above 99.999%: about 443 million power-on hours with 1,112 h total downtime.
- The facility hosts air-cooled, direct-to-chip and immersion servers side by side.
Using it in design
- Give immersion tank and server catalogue items a field-reliability reference (AFR), and use component-level AFRs in the reliability module rather than assuming immersion degrades availability.
- Flag conversion of air-cooled servers (cables, PSU firmware) as a reliability risk in design notes.
- Include memory and drives in the spares plan as the dominant failure sources in immersion fleets.
- Treat the results as single-facility data; do not generalise directly to other fluids or hardware.
Monthly newsletter · free
Keep your reading list current
Once a month: new library sources, changes to Turkish regulations and upcoming data centre events.
SoSo CAD
Check cooling capacity alongside your design
Review zone heat loads, available capacity and the capacity balance after the largest unit is lost in SoSo CAD.
Request a demo
Who should read it
Designers evaluating immersion cooling, and operations and reliability engineers.
Related standards
- OCP Immersion Requirements
Publication details
- Authors
- Austin Hipes
- Publisher
- Open Compute Project Foundation
- Year
- 2025
- Document number
- OCP EMEA Summit 2025, Immersion Cooling track
- Event
- OCP EMEA Summit 2025, Dublin
- Document language
- English
- Source type
- Industry body
- Access
- Free to read
- Licence
- Copyright retained; citation only — the publisher’s terms apply to the original document
Citation
Austin Hipes (2025). Immersion Server Demonstrated Reliability at Datacenter Scale Deployment (OCP EMEA Summit 2025). Open Compute Project Foundation. OCP EMEA Summit 2025, Immersion Cooling track. OCP EMEA Summit 2025, Dublin. https://drive.google.com/file/d/168h6CsA4weoNITyBx9lMdPiypScZLv8L/view
Events on this subject
Browse the events calendarAll content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.
Related sources
All sources on this topic
Open Compute Project Foundation 2022
Leak Detection and Intervention, Revision 1.0 (OCP Advanced Cooling Solutions white paper)
Run a probability-and-impact leak risk analysis at project start, specify BSPP/O-ring joints for serviceable connections and check the O-ring compound against the coolant chemistry using the full wetted-materials list.
- A UQD2 dry-break quick disconnect is specified to drip less than 0.063 cc when disconnected at 200 psi.
Schneider Electric 2025
Direct Liquid Cooling System Challenges in Data Centers
Use the eight challenges as a DLC readiness checklist in the commissioning and design-review module.
- Heat capture ratio defines the air share: 80% liquid capture leaves 20% for air cooling.
Huawei Digital Power 2024
Top 10 Trends of Data Center Facility 2024 White Paper
For lithium BESS/UPS rooms, require detection and segregation sized for a 30-60 s runaway escalation.
- Rack power expected to rise from 6-8 kW to 30-40 kW.
Open Compute Project Foundation 2026
Single-Phase Direct-to-Chip Cold Plate Test Specification, Version 1.0.0 (OCP Cooling Environments: Cold Plate Sub-Project, effective September 3, 2026)
Reference the specification in a cold plate purchase requirement and state which tests, limits and sampling the customer requires, since many limits are examples.
- Flow rate testing goes up to velocities below 1.5 m/s to ensure no erosion or leaks.
Open Compute Project Foundation 2026
Whitepaper: Thermal Boundaries in AI Infrastructure - Matching cooling technologies to compute through the power inflection, Version 1.00 (October 2026)
Choose cooling architecture against target rack power and each method's usable envelope: a 45 kW air ceiling (60-80 kW with rear doors), 120-250 kW class cold-plate racks, and the paper's anchor values for announced 600 kW-class racks.
- Practical air-cooled rack ceiling is taken as 45 kW, extendable to about 60-80 kW with rear-door heat exchangers.
Open Compute Project Foundation 2025
Modular Technology Cooling System for "Cloud Scale": Design & Delivery of Liquid to Rack Distribution Systems, Version 1.0 (25 August 2025)
When sizing TCS capacity against electrical capacity, state the cold-plate capture fraction (70-80%) and target delta-T (10 or 20 C) explicitly, and use the roughly doubled delivery at 20 C (4,845 versus 2,422 kW) as design headroom.
- Most IT equipment is designed for a 10 C coolant delta-T, with some designs considering 20 C.