Nscale and partners complete digital-twin pilot at Portugal AI data center

Nscale can now model changes to its Portuguese AI infrastructure before making them on-site, according to an October 27, 2026 announcement from Schneider Electric, Nscale and Start Campus. The companies said they successfully piloted a digital twin—a virtual representation of the facility—at SIN01, connecting computing requirements with power, cooling and operational models.

One model for computing and the building

Called Immersive Twin, the pilot uses NVIDIA’s Omniverse DSX Blueprint, part of NVIDIA DSX Sim. It combines Schneider Electric, AVEVA, ETAP and third-party technologies in an OpenUSD-based digital model. The deployment focuses on Nscale’s operations inside SIN01, rather than representing a completed rollout across the entire SINES Data Campus.

The integration brings engineering and operational data together with power and cooling simulations and AI workload requirements. Its purpose is to show how the data center performs as a connected system, giving Nscale a way to identify potential improvements before committing to physical changes. Computing needs and facility constraints sit inside the same model.

Each participant supplies a different part of that picture. Nscale contributes workload and computing-infrastructure requirements. Start Campus supplies the physical facility data, engineering expertise and operating environment. Schneider Electric’s digital-twin technology, powered by AVEVA software, connects those perspectives with the simulation tools. The collaboration is intended to give operators and customers a shared basis for infrastructure decisions.

By using digital twin technology to simulate how a jump in rack density moves through the power train and the liquid cooling loop before it becomes a physical reality, we can test new hardware generations, agree the configuration with our customer and build it once.

Robert Dunn, chief executive officer, Start Campus

Checking an upgrade before installation

The companies describe rapid what-if testing across electrical power, cooling and process behavior. Those tests are intended to validate design decisions, assess the effects of upgrades and reduce procurement risk. That extends the model’s role beyond displaying equipment: it is meant to help teams evaluate a proposed configuration before buying or installing it.

The intended scope runs from design through operations. During operation, the twin links computing data with data from physical equipment and control systems to support continuing performance improvements. The project also includes simulation-ready asset files intended to make equipment models easier to integrate into digital twins, an interoperability effort alongside the site-specific pilot.

A successful pilot, not a quantified efficiency result

Nscale expects the twin to shorten the path to producing AI output, improve token efficiency, strengthen operations and provide more flexibility when changing its technology stack. It also expects lower risk from those changes. The announcement calls the pilot successful but gives no numerical results for deployment speed, efficiency gains or risk reduction. Those benefits remain expectations rather than demonstrated improvements.

The physical setting gives the experiment a substantial future workload to accommodate. SIN01 sits within Start Campus’s 1.2-gigawatt SINES Data Campus in Portugal. The announcement says the campus will become home to more than 66,000 NVIDIA Rubin GPUs. That is a planned deployment, not a count of processors already installed or evidence that the pilot has modeled that full fleet.