“AI is driving the wider adoption of liquid cooling and at LiquidStack, we believe a majority of products and services that companies and consumers purchase will incorporate real-time AI in the coming years.”
Please tell us a little bit about your role in the company.
As CEO and co-founder of LiquidStack, I have spent most of my time these last 4 years charting the company’s strategy, driving industry leadership and differentiation with disruptive technology, providing the pathway for funding as well as positioning the company for sustained growth in multiple commercial verticals and business applications. I am particularly proud of the leadership team we have built, our retention of key talent and the superhuman commitment all of our employees demonstrate to exceed customer expectations.
How did you arrive at LiquidStack?
My career in this industry started in 1992, at American Power Conversion (APC), a NASDAQ mover and shaker popularized during that decade. Schneider Electric later acquired APC and I helped globalize and scale their cooling business from the ground up during a 19-year tenure. I’ve always found the start-up world thrilling, having held key leadership roles with several emerging power and cooling technology companies. Bitfury, LiquidStack’s majority shareholder, reached out to me in 2019 with an opportunity to lead their liquid cooling business, previously known as Allied Control Limited. Despite a great run with Schneider Electric, I was intrigued by the opportunity to contribute more directly, in a hands-on way, to accelerate data center sustainability and ESG. So I joined Bitfury, executed a spin-out of ACL, re-branded the company as LiquidStack and achieved Series A and Series B fundraising milestones. I guess as they say, “the rest is history!”
Read More: CIO Influence Interview with Pete Lilley, Vice President and GM at Instaclustr
What is LiquidStack? What are your core offerings?
LiquidStack has the world’s largest install base of liquid cooling for data centers globally. The company’s immersion liquid cooling works by submerging computers in dielectric heat-transfer fluid to passively reject IT heat without the use of fans or mechanical air-conditioning. Our solutions use far less energy compared to air cooling and other forms of liquid cooling. We also massively reduced the amount of water used for powering and cooling data centers. In fact, we are a zero-water technology. Our solutions are used in hyperscale, cloud, enterprise, edge and high-performance computing applications, including crypto-mining.
Our core offerings include the DataTank 48U, a 2-phase immersion cooling solution for full-scale data centers; the DataTank Modular and MicroModular, which are prefabricated and modular turnkey data centers integrated with 2-phase immersion cooling up to 1.5MW; and the DataTank 4U, which provides 2-phase immersion cooling for edge and micro data centers. We also design custom solutions.
Our systems can reduce a data center’s carbon footprint by over 1,500 tons of carbon dioxide equivalents per megawatt (MW) and with wider adoption, could save over 300 billion liters of water per year in the United States alone! They also reduce energy use by 41% and cost by 32% while providing more than 28x the heat rejection capacity than air cooling.
How important is immersion cooling technology for modern cloud data centers? What kind of infrastructure does a data center operator need to benefit from liquid cooling for data centers?
Liquid immersion cooling is one of the few ways that semiconductor chips with a TDP of higher than 270 watts can be cooled. Given that TDPs are projected to eclipse 1,000 watts by 2025, I would say immersion cooling is pretty important for modern cloud data centers!
Data center operators using LiquidStack’s solutions deploy DataTanks instead of conventional IT racks. DataTanks simply occupy the same space where the IT racks would normally sit. Because 2-phase immersion cooling removes 98% of the IT heat load, only a minimal amount of comfort cooling is required in the whitespace. The heat dissipated by the DataTanks is transferred to the primary liquid loop which exits the whitespace and is rejected to ambient via a chiller or fluid cooler. The result of these changes is data centers which can be 69% smaller, use 32% less land and be built 20% faster than air-cooled facilities.
The role of data centers in climate change is a hot topic of debate among all stakeholders in the cloud industry. Could you explain how data center operations contribute to climate change?
The more energy and water that data centers use, the greater their carbon footprint, the greater their contribution to climate change.
Processors are increasing transistor counts, and more businesses are using data-heavy technologies such as artificial intelligence (AI), machine learning, edge computing and internet of things (IoT). Air cooling can draw up to 40% of a typical data center’s energy consumption, and that amount can grow only up to a point. Processors are increasing transistor counts, and more businesses are using data-heavy technologies such as artificial intelligence (AI), machine learning, edge computing and internet of things (IoT). The reason I say that amount will grow up to a point is due to physics. At greater than 60kW per rack, air and mechanical cooling takes heroic engineering, becomes extremely costly and inefficient. At these higher densities, the impact of air cooling to carbon footprint can be directly tied to the massive increase in fan power, which is a cube to the fan speed.
But it’s not just about carbon footprint impact. In the past several years, many data centers started spraying water on chiller evaporator coils, called “adiabatic assist.” While this improves chiller efficiency and capacity, it increases water consumption. It’s estimated that data centers in the USA consumed more than 660 billion liters of water in 2020 alone, and that the average data center uses enough water to fill an olympic-sized swimming pool every two days.
When you calculate what data centers consume: energy, water, land, plastic, steel, aluminum and other materials, it’s clear that the current data center model is unsustainable. That’s not even considering the cost of transportation and packing materials for everything that goes into building and equipping a data center.
Read More: CIO Influence Interview with Muhi S. Majzoub Chief Product Officer at OpenText
What are the major differences between single phase and two phase immersion cooling technologies? Which one should the CIO of an AI company adopt?
The greatest difference is that a two-phase immersion system uses latent heat r****** capacity and a single-phase immersion system does not. As a result, single-phase immersion liquid cooling has to constantly pump dielectric fluid around to reject the heat and is drastically affected by how cold the chiller leaving water temperature is. Increasing the chiller leaving water temperature by just a few degrees can have a huge negative impact on single-phase cooling capacity.
In a two-phase system, the electronics are immersed in a dielectric (non-conductive) fluid, which has a relatively low boiling point. The heat transfer occurs in phase 1 as the hot vapor rises. Then, in phase 2, the vapor condenses and falls back into the tank as droplets, and the cycle repeats. The process is entirely passive. No pumps or moving parts are required to cool IT equipment.
A single-phase system calls for the fluid to be circulated out of the system, where it can be cooled before it’s brought back in again.
Two-phase systems provide twice or or greater heat rejection capacity, which is the ability to cool, than a one-phase system. The high value waste heat that a two-phase system produces can be re-used for hot water, district heating or energy generation.
How are energy saving benchmarks linked with performance?
The partial PUE of a 2-phase immersion cooling system is 1.02 to 1.04 nominal and 1.05 to 1.10 for a 1-phase system.
What is the future of data center cooling stacks? Do you see a future with AI or automation-powered cooling technologies?
Broadly speaking, liquid cooling within the hyperscale data center space is already here; meaning it is approaching wider scale adoption for certain workloads, like AI for instance. In traditional enterprise applications, the adoption cycle is likely to be longer, as many of these customers will resist change just like we have witnessed resistance to new air cooling technologies in the past, like variable speed fans, rack/row based cooling and even the ‘abolishment’ of raised floor as a preferred means of distributing air.
AI is driving the wider adoption of liquid cooling and at LiquidStack, we believe a majority of products and services that companies and consumers purchase will incorporate real-time Artificial Intelligence in the coming years. AI is driving a rapid acceleration in semiconductor chip power and rack densities where air cooling becomes difficult, impractical, unsustainable and costly. Liquid cooling, such as immersion cooling, solves many challenges that air cannot. Physics, ESG and economics are driving this rapid change.
Read More: CIO Influence Interview with Antoine Jebara, Co-Founder and GM, MSP Products at JumpCloud
What are your plans for the next 2 years? What kind of products and resources do you plan to bring to the market?
LiquidStack just received substantial funding from Trane Technologies. We are using some of these funds to open a US manufacturing site, increase investment into R&D, as well as scale our commercial and services operations. In addition, building off our massive install base of product, LiquidStack expects to diversify its liquid cooling offer portfolio using advanced fluid technologies and harnessing the immense potential of waste heat re-use and repurposing. My vision for LiquidStack is to continue to pioneer and deliver high-performance, cost-effective, and reliable liquid cooling solutions that solve the thermal management needs of IT infrastructure, no matter if it’s in centralized and edge data centers.
Thank you, Joe! That was fun and we hope to see you back on cioinfluence.com soon.
[To participate in our interview series, please write to us at sghosh@martechseries.com]
Since 2019, Joe Capes has been the CEO of LiquidStack, a leading provider liquid cooling solutions for data centers, edge, and high-performance computing. Prior to this role, Capes served as Global Director of Cooling Offer Development for Schneider Electric’s cloud and service provider customer segment. He has held several senior-level management positions in sales, marketing, product management and business development for Schneider Electric, spanning 19 years of service with the company. Capes is an accomplished entrepreneur and start-up company executive with extensive fund-raising, mergers, and acquisitions experience. He was previously Vice President of Sales and Marketing for Premium Power, responsible for the commercialization of advanced battery power systems used in telecoms, utility, and renewable energy applications. Capes was also co-founder and General Manager of Coolcentric, a company that provides high efficiency liquid cooling solutions for data centers and high-performance computing applications. He holds a Bachelor of Science degree in Electrical Engineering from Lehigh University and is named as an inventor in many patents related to data center cooling technologies.[/vc_column_text
LiquidStack has the world’s largest install base of liquid cooling for data centers globally. LiquidStack’s immersion liquid cooling is the only proven, highly scalable, environmentally safe and sustainable solution to meet the growing thermal challenges of cloud, enterprise, edge, high performance computing and crypto-mining applications. Since pioneering 2-phase immersion cooling in 2012, LiquidStack has deployed advanced cooling solutions across the world’s most demanding compute environments, actively reducing energy and water use on a massive scale. Today, LiquidStack provides thermal management solutions to many of the world’s largest cloud services, semiconductor, manufacturing and IT hardware providers
