Executive Summary

• Liquid cooling is becoming the baseline for data centres with high-density AI workloads, even with growing environmental concerns around water usage.
• Air cooling has a hard limit of 41.3kW per rack; it would be impossible for air cooling alone to cope with AI workloads of 100kW and beyond; the fans would have to move the air so fast it would reach speeds of a Category 2 hurricane.
• The liquid cooling market hit $5.52 billion last year, forecast to hit $15.75 billion by 2030, cementing the fact that liquid cooling is fundamentally shifting the industry.

AI workloads are expected to outstrip current capacity by 2027 – yes, next year. This is the consensus from Salute’s State of the Industry 2026 report, where 83% of respondents (made up of 200 senior IT and data centre professionals in the UK and US) believe AI workloads will exceed current capacity, with readiness an issue already this year.

A whopping 74% said they’re not fully prepared to support AI workloads right now, let alone when the demand exceeds capacity. In terms of workloads, we’re seeing the rise of ultra-high density AI workloads; the kW is old news, MW are in. Racks are routinely exceeding 50-100kW. Hyperscale clusters? Well, they’re reaching multi-megawatt deployments, and that means it’s getting hot hot hot in there.

Rock you like a hurricane

If a single rack can reach (or exceed) 100kW, air cooling won’t be able to hack it; air cooling has a hard limit of 41.3kW per rack. If it tried to handle past that threshold, the practical design makes it impossible for it to do so; to cool that much power using air cooling alone, the fans would have to move the air so fast it would reach speeds of a Category 2 hurricane (100mph roughly). Not to mention, it would create thermal chaos and deafening noise attempting to handle the workload.

Liquid cooling is the baseline for 2026

Liquid cooling is being implemented by around 22% of data centres and the market hit $5.52 billion last year, forecast to hit $15.75 billion by 2030, cementing the fact that liquid cooling is fundamentally shifting the industry. For data centres, hyperscalers and multi-facility campuses that specifically handle AI workloads, liquid cooling is now formulating the baseline. It’s not a nice-to-have for them; it’s an essential strategy so they can handle workloads and don’t hit a thermal wall.

While there are concerns around WUE (water usage effectiveness), water scarcity and climate change pledges, data centre operators are more environmentally-conscious than ever, deploying innovative solutions and tech to stay on track of climate change commitments. This year, we will see more shifts as the cooling markets evolve rapidly.