Cooling Data Center Podcast

What if the only way to know your liquid cooling infrastructure will hold up is to make it fail on purpose, before a single rack goes into the white space?

In this episode, Demetrio Velasco of Liquidpulse joins Robert to talk about something the industry has needed for a while: a two-megawatt liquid cooled load simulator capable of stress testing critical cooling infrastructure in a live data centre under real test conditions.

From starting out as a controls engineer building pump and hydronics systems, to launching Liquidpulse out of Dallas and putting liquid cooled load banks into the market at scale, Demetrio brings a manufacturer's view to a commissioning problem that keeps getting bigger.

Together, they explore:

· Why the jump from a 500kW machine to a two megawatt prototype happened faster than anyone planned

· How modular, networkable units let teams run tests at 28 megawatts of simultaneous load

· The origin story: a client with an urgent need, a friend's phone call, and a full witness test where the scope changed at the last minute

· Why load bank cleanliness matters, and the standard Liquidpulse is pushing at OCP to protect flushed technical cooling systems

· The air separation problem nobody predicted, and the auto air vent fix that became a second generation feature

· What integrated testing really proves: the cooling, the medium voltage generation, the power failure recovery modes, and the people commissioning it all

· The Texas site that skipped integrated commissioning and spent a year working through double digit outages

· Why a hybrid unit generating 165kW liquid load and 60kW air load exists alongside the 500kW machines

If you have ever wondered how anyone signs off on infrastructure built for loads that do not exist yet, Demetrio explains the engineering and the commercial reality behind it.

His take? "It's either you live test or you load test." There is no third option.

What if the biggest barrier to maximizing cooling capacity and efficiency isn't the hardware in your data hall, but the static set points and manual control strategies you've been running since day one?

In this episode, Clifford Federspiel, founder of Vigilent and former Johnson Controls and UC Berkeley engineer, joins Robert for a wide-ranging conversation about what happens when you hand cooling control to a machine learning system that can predict what's coming next, and why the data centre industry's dependence on fixed, conservative control is leaving real savings untouched.

From building sensor-driven efficiency solutions for "lost and forgotten" buildings in 2004, to deploying predictive cooling optimisation across over a thousand sites worldwide, Cliff brings two decades of practical AI and controls experience to an industry still debating whether to trust automation.

Together, they explore:

• Why Vigilent's system can learn a facility's cooling dynamics in as little as a few hours
• How predictive "look ahead" control differs from the basic PID and fuzzy logic embedded in most OEM equipment
• The self-driving car analogy: why operators can always take control back, and why they rarely need to
• How saving cooling power in AI data centres directly translates to more compute capacity and revenue
• Why the common practice of "even wear" on equipment actually maximises the risk of simultaneous failure
• The growing complexity of dual cooling plants as liquid and air systems coexist at the rack
• Why fixed set points become exponentially more expensive as rack densities climb from 10kW to 100kW+
• How Vigilent's operational data helps teams make informed decisions about when hardware investment is truly needed

If you've ever wondered whether your cooling plant could be doing more with what it already has, or wanted someone to explain AI-driven optimisation without the hype, Cliff delivers straight talk backed by over a thousand deployments. His take? The industry is driven by fear, but the proof is already there, and a system watching every minute and adjusting dynamically will always outperform static setpoints and manual controls.

What if the biggest barrier to scaling liquid cooling isn't the engineering, but an industry-wide habit of keeping things secret that don't need to be kept secret?

In this episode, Bill Kosik of HED and chair of The Green Grid's AI Impacts working group joins Robert for a genuinely insightful conversation about the long view of data centre design, and why the lessons from IBM's water-cooled mainframes still apply to today's 300kW racks.

From his time at Hewlett Packard's facilities design group working on liquid-cooled supercomputing, to leading mission critical projects at HED today, Bill brings three decades of pragmatism to an industry that often confuses novelty with progress.

Together, they explore:

• Why liquid cooling isn't new technology and what the 1980s mainframe era can still teach us
• How the manufacturers stepping up with full liquid-cooled product suites unlocked the current scale
• The inflexion points reshaping cooling design, from 5kW cabinets a decade ago to 300kW racks today
• Why the industry's culture of secrecy is fuelling public misunderstanding around water use, noise and vibration
• How global end users, not just regulators, are driving sustainability standards in the United States
• The retrofit opportunity hiding in Tier IV enterprise data centres being repurposed for small-scale AI
• Why ASHRAE 90.4 becoming part of energy code means you can't design around it anymore
• The case for designing data centres that don't look like distribution warehouses

If you've ever wondered whether 300kW racks will one day feel as routine as 30kW racks do now, or wanted someone to put today's AI build-out in proper historical context, Bill delivers the kind of measured engineering wisdom that's increasingly rare in a hype-driven market.

His take? In five years we'll be sitting around laughing at how we didn't know how to cool a 300kW rack. The history of this industry is one of inflexion points, and the people who recognise them early are the ones who shape what comes next.

What if the biggest barrier to liquid cooling efficiency isn't the technology itself, but the legacy thinking that's keeping temperatures unnecessarily low?

In this episode, Jake Roberts of Excool joins Robert for a refreshingly honest conversation about what it really takes to deliver cooling solutions across global markets, and why the data centre industry's conservative approach might be leaving serious efficiency gains on the table.

From retrofitting existing facilities without ripping out infrastructure to challenging assumptions about water use in cooling, Jake brings practical engineering insight to an industry still debating whether to raise chip temperatures above 20 degrees Celsius.

Together, they explore:

· Why retrofitting for liquid cooling doesn't have to mean reinventing the wheel or running chilled water everywhere

· The real efficiency cost of running entire chiller plants at 18 degrees just to handle 20-30% of airside load

· How indirect cooling units can decouple airside and liquid cooling temperatures for better PUE

· Why NVIDIA's push toward 45 degree facility water loops could eliminate the need for water chillers entirely

· The nuanced truth about water use: why a little adiabatic cooling on peak days beats moving the problem to power stations

· How West Midlands manufacturing heritage enables rapid, bespoke delivery to global markets

· What speed to market really means when customers need solutions in expedited timelines

If you've ever wondered whether your cooling design philosophy is leaving stranded power on the table, or wanted someone to explain retrofit options without assuming you're starting from scratch, Jake delivers straight talk with engineering credibility. His insight? Pushing temperatures up has such a positive effect on the industry, from sustainability through to cost efficiency, and the ROI on reduced peak PUE is immediate.

What if the biggest barrier to liquid cooling adoption isn't the technology - it's just protecting against what happens when things go wrong?

In this episode, Cassandra Zentner of Actnano joins Robert for a refreshingly practical conversation that strips away the idealism and gets to the heart of why protection matters just as much as prevention - and why automotive-grade coating technology is already more relevant than most people realise.

From her decade protecting Tesla and Rivian autonomous driving systems to bringing 15-20 year warranty-grade protection to data centers, Cassandra brings grounded realism to an industry still assuming leaks and condensation can be completely prevented.

Together, they explore:

· Why protection is the missing piece in the prevent-detect-protect trilogy

· The reality behind leak assumptions - catastrophic failures AND slow PG25 coolant buildup both damage expensive chips

· Why coating technology (not just tighter controls) enables operational flexibility

· How Actnano's collaboration across chip makers, hyperscalers, and ODMs creates supply chain-wide protection

· Why condensation concerns and narrow HVAC windows mean data centers are over-engineered for prevention

· The insurance policy mindset - protecting hundred-thousand-pound investments when infrastructure inevitably fails

If you've ever wondered whether direct liquid cooling protection is essential or wanted someone to explain it without assuming perfect conditions, Cassandra delivers the straight talk with automotive credibility. Her insight? "The data center is the easy environment" - because if it protects against salt, frost, and corrosive gases in EVs, your server boards are already covered.

What if the biggest barrier to immersion cooling isn't technical - it's just understanding what it actually is?

In this episode, Allison Boen of Alcatex and the Open Compute Project joins Robert for a warmly authentic conversation that strips away the jargon and gets to the heart of why single-phase immersion cooling matters - and why it's already more deployed than most people realise.

From her 30-year journey as a self-described "CRAC dealer" (computer room air conditioner specialist) to becoming Shell's Brand Ambassador and OCP's immersion community champion, Allison brings infectious enthusiasm to a technology she discovered in 2014 and hasn't stopped evangelising since.

Together, they explore:

• Why Intel's recent warranty certification with Shell and Supermicro changes everything
• The reality behind deployment numbers - Shell's Houston facility runs half its 7MW workload on just 2MW of immersion
• Why hybrid data centres (not rip-and-replace) are the practical path forward
• How OCP's collaboration breaks down ego barriers and creates agnostic TCO tools anyone can use
• Why protest signs in tiny Irish villages and Mission Impossible films mean data centres have officially gone mainstream

If you've ever wondered whether immersion cooling is still just "proof of concept" or wanted someone to explain it without the sales pitch, Allison delivers the straight talk with genuine passion. Her advice? "Don't spill it" - because if you need floor drains, you've engineered it wrong.

What happens when you've been testing data centres since before "data centre" was even a term?

In this episode, Dave Wolfenden of Heat Load joins Robert to trace an extraordinary journey from COBOL programming at Heathrow Airport to becoming one of the UK's pioneering data centre commissioning specialists. Starting with Argos heaters and 500kW of rack-mounted units delivered in transit vans, Heat Load has evolved to deploy up to 120MW of testing equipment across 10 articulated lorries.

Dave shares candid insights from over three decades in the industry - from the early days of chaotic rack layouts and liquid CO2 cooling experiments to today's hyperscale pressure cooker where data halls are worth £1M per megawatt per day in lost revenue.

Together, they explore:

• How testing has evolved from basic USB sensors to real-time 3D modelling with wireless monitoring
• The logistics nightmare of juggling equipment across Europe while customers demand instant deployment
• Why immersion cooling adoption is slower than expected (and why Heat Load is ready anyway)
• The shift from owner-operators to hyperscalers - and what that means for the industry

If you want to understand how data centre commissioning really works - from the sharp end of delivery deadlines to the evolution of an entire industry - this conversation delivers the unvarnished truth from someone who's been there since day one.

What if we could decarbonise IT… by rethinking the liquid?

In this episode, Maarten Trautmann of Oleon joins Robert to unpack why plant-based dielectric fluids aren't just a greener alternative - they may be the key to unlocking waterless, sustainable cooling at the edge, at scale, and across Europe.

From his early fascination with Ireland to leading Oleon's charge into liquid cooling, Maarten shares how Oleon's biodegradable, non-toxic fluids are offering a credible path to closed-loop cooling, heat reuse, and decarbonisation - without compromising performance.

Together, they dive into:

• The surprising history of Oleon (yes, it started with candles)
• Why Europe is missing the mark on incentivising heat reuse
• How immersion cooling adoption is being throttled by warranty fears and retrofitting costs
• Why innovation in cooling is still in full swing - and where consolidation might come next

If you're interested in liquid cooling beyond just AI hype, this one's for you.

🔗 Connect with Maarten: linkedin.com/in/maarten-trautmann-2997a71a 🔗 Connect with Robert: linkedin.com/in/robertussher

Two-phase cooling isn't new - but it's never been more needed.

In this episode, Dr. Richard Bonner of Accelsius joins Robert to unpack how a decades-old thermal principle, once used in industrial processes, is being reimagined to meet the explosive demands of AI and high-power GPUs.

They explore Accelsius' patented two-phase cold plate technology - originally developed at Bell Labs - and how it's being commercialised to deliver unmatched reliability, redundancy, and sustainability. From N+1 pump systems to the role of two-phase in future data center architectures, this episode covers everything from thermal physics to practical deployment strategies.

If you're curious about the roadmap to next-gen cooling, the reality behind liquid vs. immersion, or what's actually going to enable the Rubin generation of chips… this episode is for you.

Connect with Robert: www.linkedin.com/in/robertussher Connect with Richard: https://www.linkedin.com/in/richbonnerphd/

Craig MacFadyen is back on the podcast - this time fresh from the floor of the OCP Global Summit in Dublin.

In this episode, Craig and Robert dig into the real conversations happening around liquid cooling, hyperscaler expectations, and whether the data center world is preparing for a future that hasn't quite arrived yet.

From CDUs and 500kW rack forecasts to the tension between innovation and affordability, they tackle the practical, the political, and the philosophical challenges facing data center cooling in 2025. Craig also shares his perspective on startup tech, smarter not newer design, and why collaboration still feels more talked about than practiced.

It's an honest, thought-provoking follow-up to Craig's first appearance - don't miss it.

🎧 Topics include:

• Key takeaways from OCP 2025
• The rise of rack density and the "500kW by 2030" challenge
• How hyperscalers are influencing the market—and when adoption will really happen
• The innovation bottleneck caused by NDAs and over-regulation
• What smart design actually looks like in cooling
• The role of power availability and why cloud still dominates over AI

Connect with Robert: www.linkedin.com/in/robertussher Connect with Craig: www.linkedin.com/in/craig-macfadyen-5b4b811a