Over just a handful of years, demand for AI has skyrocketed, and investment in AI-ready data centers has matched pace. The rise of these high-capacity, always-on facilities is rapidly redefining the data center – from how they’re designed and built, to how they operate.

While the effects of this shift are visible across the entire data center ecosystem, the extent of change is perhaps most pronounced in cooling. A new type of workload is driving heat outputs beyond what traditional air-cooled facilities were designed to handle. That’s where liquid cooling comes in.

“An AI data center equals liquid cooling,” states Peter Huang, global president of thermal management and data centers at BP Castrol. “When we move to higher power and rack densities, liquid cooling isn’t just an option – it’s a necessity.”

There’s no doubt that the industry is accelerating toward liquid cooling as standard practice, but as with any transition, the journey isn’t without its challenges. For smaller facilities and brownfield retrofits where air cooling remains the go-to, operators face multiple barriers to adoption.

Against this mixed backdrop, in a recent DCD>Talks episode, Huang shares the reality of current cooling trends, the key bottlenecks to wider adoption, and how liquid cooling is enabling high-density innovation.

The big picture

Huang sums up today’s cooling landscape: “We’re seeing hyperscalers, large cloud providers, and service providers adopting liquid cooling at pace. And we know firsthand that significant investments are planned over the next three to five years.

“On the other side, smaller enterprise and co-location customers are more likely to adopt hybrid solutions – but liquid cooling is definitely gaining traction. Right now, we’re seeing roughly a 70/30 split between air and liquid.”

And this shift isn’t being driven solely by higher rack densities. It’s also about efficiency: “Liquid cooling can reduce cooling-related opex by up to 30 or 40 percent, and it’s proven that you can achieve a lower power usage effectiveness (PUE) with liquid-based solutions,” continues Huang.

This reality is the catalyst behind BP Castrol’s focus on lowering the barriers to industry-wide adoption via its strategic partnerships, global footprint, and an end-to-end approach to deployment.

Breaking down barriers

The transition to liquid cooling is littered with hurdles – including the twin challenge of a complex design phase and supply chain constraints. Liquid cooling systems rely on highly specialized components – such as cold plates, pumps, quick disconnect couplings (QDCs), manifolds, and piping – many of which have long lead times.

With so many design variations and delivery models at play, the supply chain is struggling to keep up with both the complexity and speed at which operators need to deploy. It’s a challenge felt across the industry, and for Huang, the solution is clear:

“What’s ultimately required is standardized design. That will drive supply chain standardization and help address both these bottlenecks – but right now, there’s very limited standardization available.”

Regulatory uncertainty adds another layer of difficulty. “Questions around dielectric fluids, water-based liquids, local compliance, environmental regulations, and internal policies are significant,” says Huang. “And globally or regionally, these guidelines still aren’t well defined.”

The fourth major barrier is cost. But Huang is confident that this concern is often rooted in a knowledge gap: “It’s true that liquid cooling carries higher upfront capex, but the total cost of ownership compared to air cooling is lower over time.”

The opex benefits stem from improved efficiency, and as adoption increases and the technology matures, efficiency will continue to improve while the price of initial installation falls.

Introducing liquids into critical digital infrastructure naturally raises some more obvious concerns. As a result, liquid cooling demands careful monitoring and management, which can add cost and complexity. This is where BP Castrol plays a key role:

“We specialize in leak detection and sensing,” explains Huang. “At a basic level, this protects expensive servers by detecting leaks. But we’re doing things differently in offering both sensors and analyzers that monitor the fluid additives themselves. It’s the difference between assessing someone’s health just by looking at them, versus running a blood test. We’re getting right into the DNA of the fluid to ensure detailed oversight.”

Go with the flow

Addressing these challenges is key to securing deployment – but the real value lies in ensuring long-term flexibility comes built-in to the system. With rack densities only expected to rise, today’s cooling systems must be designed to keep up.

“Our approach starts with proof-of-concept designs,” says Huang. “We begin small to build confidence and ensure the system won’t disrupt existing loads – because uninterrupted data flow remains the top priority. From there, we build scalability into the design and adopt a modular approach once the concept is proven.”

In practice, this means deploying cooling, power, and IT blocks with standardized interfaces that make scaling up fast and straightforward when required.

Huang explains why ensuring fluid compatibility is another cornerstone of flexibility: “Chip and server generations – plus AI workloads – are evolving faster than ever. What once lasted three to five years is now upgraded every six months. That means planning not just for fluid compatibility across generations, but for all the surrounding components as well.”

Joining the dots

Removing adoption barriers and designing for flexibility requires a truly end-to-end view of the data center ecosystem. Liquid cooling doesn’t exist in isolation – its impact is felt across the entire facility.

At the rack level, systems must be liquid-ready – equipped with appropriate piping, manifolds, pumps, and QDCs. But racks connect to servers, servers connect to chips, and beyond that lies a broader web of supporting infrastructure. This interdependence adds complexity, and successful deployment begins with understanding the knock-on effects.

“That’s why we’re proud to be one of the most integrated liquid cooling providers across the full end-to-end chain,” says Huang. “We have expertise spanning from the chip and server level upward. That’s a fundamental advantage.”

This integrated approach ensures compatibility across the data center and delivers the flexibility modern operations require. With visibility over every critical component, anticipating future demand is no longer a stab in the dark.

BP Castrol’s strategy is further strengthened by its global footprint and collaborative mindset. “Our reference designs build customer confidence and ensure each solution is fit for purpose,” explains Huang. “Regional safety standards and regulations influence design decisions – meaning location matters. Our partnerships help us deliver that reach.”

There for the taking

Amid the complexity that comes with scaling infrastructure and navigating the available options, Huang emphasizes the importance of stepping back to define clear objectives:

“Go back to the drawing board and ask: What am I trying to achieve? Is it cloud computing? HPC? Inference or training? Are your customers enterprises or hyperscalers? The answer shapes everything.”

One thing remains clear: liquid cooling is central to unlocking greater efficiency and enabling high-density compute.

“The biggest trend – and where we’ll see the most OEM approvals – is cold plate cooling,” notes Huang. “It supports up to 100kW per rack, and even approaches 200kW. Two-phase cooling is also gaining attention. But today, 99 percent of deployments are direct-to-chip cold plate cooling, and there’s already strong standardization to support it.”

Huang concludes by highlighting three critical considerations for operators looking to specify the right liquid cooling solution: material compatibility; maintenance, operations, and services; and finally, scalability.

To hear more about how BP Castrol is supporting an industry-wide shift to liquid cooling, watch the full DCD>Talks episode with Peter Huang, here