What operational challenges arise from liquid cooling in AI infrastructure during the decommissioning process?
Answered at 12:08 in The Circular Supply Chain for AI Hardware · The Dynamo Show: a Dynamo Ventures Podcast
What operational challenges arise from liquid cooling in AI infrastructure during the decommissioning process?
GPU systems with liquid cooling are heavier and more valuable than air-cooled CPU systems, requiring additional handling steps to ensure proper management. Once equipment reaches Iron Mountain facilities, the core processes remain the same, but the company needs significantly more power and cooling distribution units to handle the cooling element of systems. As racks scale from 12-15 kilowatts to 100 kilowatts and beyond, only a limited number of companies have the infrastructure capability to test such systems, which Iron Mountain plans to build out worldwide.
Full answer and detailsRead the whole answer, key facts and every other question from this episode
In short
GPU systems with liquid cooling are heavier and more valuable than air-cooled CPU systems, requiring additional handling steps to ensure proper management.
Tom Delahunty, VP of Operations for Asset Lifecycle Management at Iron Mountain · The Circular Supply Chain for AI Hardware · The Dynamo Show: a Dynamo Ventures Podcast · answer at 12:08
Key takeaways
- GPU systems with liquid cooling are heavier and more valuable than air-cooled CPU systems, requiring additional handling steps to ensure proper management.
Quote to remember
“I see it as a supply chain opportunity. I would look, my lens says those assets have a future in life.”
— Tom Delahunty
The full answer · about 2 min
GPU systems with liquid cooling are heavier and more valuable than air-cooled CPU systems, requiring additional handling steps to ensure proper management. Once equipment reaches Iron Mountain facilities, the core processes remain the same, but the company needs significantly more power and cooling distribution units to handle the cooling element of systems. As racks scale from 12-15 kilowatts to 100 kilowatts and beyond, only a limited number of companies have the infrastructure capability to test such systems, which Iron Mountain plans to build out worldwide.
Key facts
- Iron Mountain serves almost 225,000 organizations with data storage and records management
- Q4 2025: approximately 70% year-over-year growth in asset lifecycle management
- Q2 2025: plus 88% year-over-year growth and plus 82% organic growth
- One international customer's decommissioning project spans five months and touches seven countries
- AI infrastructure racks scale from 12-15 kilowatts up to 100 kilowatts and beyond
People, companies and topics mentioned
More answers from this episode
- 4:00What growth metrics did Iron Mountain's asset lifecycle division achieve in Q2 and Q4 2025?
- 6:02What is the decommissioning process for data center equipment at Iron Mountain, from de-racking through disposition?
- 10:30How does Iron Mountain's decommissioning workflow differ between CPU server deployments and GPU-dense AI infrastructure?
- 19:48What drives increasing customer acceptance of used and refurbished GPU and AI equipment in secondary markets?
- 23:48What infrastructure and R&D investments does Iron Mountain believe the ITAD industry needs to handle future AI hardware volumes?