The short version: Digital twins for AI data centers moved from slideware to live sites in late October 2026. Schneider Electric, AVEVA, cloud provider Nscale and campus operator Start Campus finished a pilot at a working AI data center in Portugal, and Synopsys joined a Singapore testbed for cooling AI racks in tropical heat. The idea is to test rack upgrades, power and liquid cooling in software before touching the real building.
The Portugal pilot
On October 27, 2026, the four companies announced a completed pilot of an end-to-end digital twin at Nscale’s SIN01 facility on Start Campus’s Sines data campus. It was built with NVIDIA’s Omniverse DSX Blueprint and links Schneider Electric, ETAP and AVEVA software in one OpenUSD-based model, which the partners call the Immersive Twin.
- Workload to infrastructure. The twin connects what the AI workload needs to the electrical distribution and liquid-cooling models.
- Rehearse before you build. Operators can see how a jump in rack density moves through the power train and the cooling loop before new hardware arrives.
- Agree once, build once. Nscale’s CTO said the goal is to test new hardware generations, agree the setup with the customer and build it one time.
- Design through operations. The partners describe it as covering the full life of the site, not just the design phase.
This builds on the reference designs and AI operations tools we covered in Schneider, NVIDIA and AVEVA bring AI into data center design. The difference now is a live site.
Cooling is the hard part
The same week, Synopsys said it had joined a research group led by the National University of Singapore to study how to cool AI data centers in tropical climates. The work covers liquid cooling, computational fluid dynamics, AI-assisted thermal design and digital twins. Synopsys cited an industry estimate that cooling can use up to 40% of a tropical data center’s energy, though no savings from this project have been reported yet.
In Europe, cooling equipment maker LUVE joined COLLINS, a three-year project running from September 2026 to August 2029 to build a validated digital twin of the full liquid-cooling chain, from chip cold plates to heat rejection. Models will be calibrated against real equipment running realistic AI loads.
A healthy dose of skepticism
A digital twin does not save energy by itself. Its value depends on how closely its predictions match the real plant, and on whether it is checked against data that was not used to tune it. That is why projects like COLLINS focus on validation. Operators weighing vendor claims should ask how the model was validated and against what.
Who should pay attention
Data center developers and operators, colocation providers, design firms and the mechanical and electrical contractors building high-density AI campuses. With hundreds of projects in the pipeline across Texas, Georgia and the Carolinas (see our US data center map), the ability to test changes before a shutdown is worth real money.
What it means for hiring
Digital twins raise the bar for the engineers who build and run AI data centers. The model is only as good as the people who check it.
- Mechanical engineers with liquid cooling and CFD experience are in demand. They set up the cooling models and judge whether results make sense.
- Electrical engineers who know power studies software such as ETAP become more valuable as power models connect to live workloads.
- Commissioning and operations engineers who compare model predictions to real readings are the people who make a twin trustworthy.
Hiring for a data center project? See our data center recruiting page and the Data Center Workforce Report.
Interview questions for data center engineers
- How would you check whether a cooling model’s predictions match what the site is actually doing?
- Walk me through what changes in the power and cooling design when rack density doubles.
- What liquid cooling systems have you designed, commissioned or operated?
- Which power study tools have you used, and how did you validate the results?
Frequently asked questions
What is a data center digital twin?
A virtual model of a data center’s power, cooling and IT load, kept up to date with real data, so engineers can test changes in software before making them on site.
Do digital twins reduce data center energy use?
They can help find savings, but only if the model is validated against real conditions. The twin itself does not save energy; better decisions made with it can.
Will AI replace data center engineers?
No. These tools help engineers plan and test, but someone still designs, builds, commissions and runs the plant, and the AI build-out is short on exactly those people.
More stories like this are on our AI for Engineers hub.
Sources
- HPCwire: Schneider Electric and partners pilot digital twin for Nscale’s AI infrastructure
- MarketScale: Nvidia and Schneider test a digital twin to rehearse AI data center upgrades
- TNGlobal: Synopsys joins NUS tropical data-center testbed
- Refindustry: LUVE joins COLLINS project
- IEEE ComSoc Technology Blog: Can digital twins deliver real energy savings for AI data center cooling?
