Joe Capes
CEO and Co-founder at LiquidStack
The US continues to invest heavily into AI data centres. With US President Donald Trump eager to become the world’s most dominant AI infrastructure player – in addition to clamping down on US tariffs – the country could critically impact the data centre industry in the rest of the world.
Currently, the country has the highest concentration of data centres in the world, with investment only continuing to surge on account of AI technologies.
With this in mind, Data Centre Magazine recently spoke with Joe Capes, CEO and Co-founder at LiquidStack, about how dramatic US data centre growth could reshape the global AI landscape and create more competition – and therefore more demand – worldwide.
“Surges in liquid cooling adoption could accelerate AI adoption globally by removing what has traditionally been a constraint to increasing compute power: thermal management,” he says.
“Advancements in AI-specific infrastructure, coupled with economies of scale, will reduce the cost of AI resources and enable cloud providers to offer more robust AI services worldwide.”
The impact of Stargate
As the US remains dominant in AI infrastructure, Joe explains how the country has also raised concerns over talent migration, data sovereignty and potential technological dependencies for countries lacking comparable capabilities.
He says: “Data sovereignty, in particular, creates complex compliance challenges and potential pressures for data localisation with differing national laws on data protection and access requiring international cooperation to balance innovation with national interests.”
One of the biggest industry stories to come out of the US is Stargate, an AI project that promises to dramatically disrupt the data centre industry.
“The Stargate AI project will drastically reshape the data centre industry by driving unprecedented demand for infrastructure, including power and cooling, bringing more resources into manufacturing, but also potentially straining supply chains,” Joe says.
“It will likely increase public visibility into AI, creating scrutiny over the resources data centres consume, such as water and electricity. However, it will create numerous jobs, and lead to standardisation that is lacking in some infrastructure areas, such as liquid cooling.”
Stargate faces no shortage of challenges, as Joe explains. For one thing, power remains a significant issue for the wider data centre industry, including the amount of power and time to power when considering sustainable generation capacity, transmission and distribution, grid interconnects and new substations.
“AI data centres need considerably more power than non-AI data centres,” Joe explains. “The International Energy Agency (IEA) estimates a conventional data centre demands ~10MW of power, while an AI data centre can demand over 100MW.
“The average time it takes for a new data centre to receive this power is somewhere around 3-5 years. A lot can happen in that timeframe.”
Confronting AI workload pressures
High-performance cooling requirements for AI systems are likely to be another challenge. To confront this, Joe suggests that Stargate will need to design its data centres for a “new frontier” in cooling.
“Homogenous AI workloads will require liquid cooling and white space while heterogeneous workloads will require both air cooling and liquid cooling,” he says.
“Liquid cooling will take various forms, such as direct-to-chip and immersion cooling, or hybrids of both technologies.
“Planning for cooling flexibility will be key, given that power and cooling infrastructure typically last 20 years. The decisions made today for Stargate will have impacts for decades to come.”
Joe explains that questions over finance, navigating supply chain pressures for critical components and addressing unique technology issues are inevitable.
He notes: “The project must contend with competition for resources like prime locations and skilled workers, adapt to evolving regulatory landscapes, and mitigate sustainability concerns related to high energy consumption. This requires careful coordination, creative solutions and broad collaboration to succeed.”
No data centre progress without cooling
Stargate demonstrates how urgent the need for infrastructure is if technology companies are going to reach the full potential of AI.
Cooling and sustainability guidelines will need to be an essential part of that plan, particularly as data centres are currently consuming roughly 1.5% of global electricity consumption.
On this, Joe touts liquid cooling solutions as vital for data centre owners and operators aiming to achieve net zero.
“Liquid cooling drastically improves energy efficiency and significantly reduces carbon emissions,” he says.
“It not only enables energy savings of up to 50% (versus legacy air cooling), but it also brings space savings, while offering water conservation benefits and opportunities for heat reuse, all of which further minimises environmental impact and reduces carbon footprint.”
He adds: “By supporting HPC demands more efficiently, liquid cooling can allow data centres to scale their operations and meet growing computational needs without proportional increases in energy and water consumption.
“This offers a flexible and scalable path toward a more sustainable and environmentally responsible industry.”
LiquidStack technology drives data centre sustainability evolution
Eager to improve energy efficiency and sustainability for AI and HPC workloads, LiquidStack’s CDU-1MW technology in particular delivers 1.35MW of heat rejection capacity.
This is while significantly reducing energy consumption and data centre space when compared to air cooling, Joe explains.
“This results in an impressive improvement in Total Cost of Ownership while supporting chips up to 2,000W TDP,” he says.
“With its universal compatibility with various direct-to-chip cooling solutions, the CDU-1MW helps data centres effectively manage the growing energy demands of high-density workloads, minimise environmental impact and enhance operational efficiency through superior cooling capacity and reduced resource consumption.”
Likewise, the company says its liquid cooling technology significantly outperforms traditional air cooling methods – particularly when it comes to energy efficiency and sustainability for AI and HPC workloads.
“It substantially reduces energy consumption and achieves ultra-low pPUE of 1.02-1.03,” Joe adds. “LiquidStack immersion cooling supports chips greater than 2,000W TDP, eliminates data centre water consumption in almost all cases and offers potential for waste heat reuse.
“By enabling more efficient cooling for power-intensive workloads, LiquidStack's technology helps data centres manage growing energy demands while significantly enhancing sustainability and operational efficiency.”
When it comes to reducing data centre emissions moving forward, Joe can’t disclose any details about ongoing projects. However, he does share his excitement about the company gaining US$20m in a Series B funding round extension, led by Tiger Global, at the end of 2024.
“We are investing significantly to ensure our liquid cooling solutions meet customer’s total cooling needs,” Joe says.
“As the world’s only true full-service liquid cooling company, LiquidStack is in a unique position to leverage its deep knowledge and R&D resources to bring new high capacity direct-to-chip, immersion and hybrid solutions and services to the market.”
To read the full article in the magazine, click HERE.
Explore the latest edition of Data Centre Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. Discover all our upcoming events and secure your tickets today
Data Centre Magazine is a BizClik brand