The Signal Index

Innovation Signal 9 (2026) | 24 March 2026

Short note, about 5 minutes to read

What happened

Data centers are beginning to adopt cooling systems that use little to no water, replacing traditional water-intensive cooling methods. Companies like Microsoft have tested data centers that rely on closed-loop or air-based cooling systems, significantly reducing water consumption while maintaining performance.

Why this matters

As demand for computing grows, especially with AI workloads, data centers are expanding rapidly. Cooling is one of their biggest constraints, and many current systems consume large amounts of water. In regions facing water scarcity, this is becoming a serious limitation. Reducing or eliminating water usage addresses a hidden but critical infrastructure bottleneck.

Who should care

  • Cloud providers and data center operators
  • Policymakers focused on resource sustainability
  • Anyone tracking the physical limits of AI infrastructure

What could change over the next 2 to 3 years

We may see wider adoption of alternative cooling technologies such as liquid immersion and advanced air cooling systems that minimize or eliminate water use. This could enable data center expansion in regions where water availability is constrained.

What might block this

New cooling systems may involve higher upfront costs and require redesigning data center infrastructure. Efficiency trade-offs and reliability at scale are still being tested, especially under heavy workloads.

Why I am sharing this

Because discussions around AI often focus on models and software, but the real constraints are increasingly physical. Cooling, energy, and water are becoming key limiting factors, and solutions like this highlight how infrastructure is evolving to support growing computational demand.

If this is useful, I can share more detail on how different cooling methods compare in terms of efficiency and cost.

And if you think I am missing an important limitation here, I would really like to hear that.

Archive Tags

Data Centres · Cooling · Water · AI Infrastructure