ABB joins NVIDIA DSX blueprint as AI factories near $60B each

ABB is integrating its power distribution equipment as digital twins into NVIDIA's DSX Blueprint, letting AI factory operators validate electrical and cooling systems before breaking ground. The Zurich-based electrification company will contribute SimReady 3D models of medium-voltage switchgear, electrical distribution equipment and uninterruptible power supply platforms to NVIDIA's Omniverse-based design environment.

"Through this collaboration, ABB provides the electrification technology leadership that AI demands," Jorge Lis, Global Data Center Segment Leader at ABB Electrification, said. "DSX integration gives our customers a proven path from design to deployment, backed by ABB's global manufacturing reach and local engineering expertise."

The integration covers source-to-rack power distribution, protection schemes and energy efficiency modeling at campus scale. Engineers can validate electrical, thermal and compute systems together in simulation before ordering prefabricated modules, compressing design cycles for AI factories where compute racks already exceed 1 megawatt per rack. The expanded collaboration builds on an October 2025 agreement to develop 800 VDC power architecture for advanced AI data centers.

The partnership comes as NVIDIA's Vera Rubin platform enters full production, with Microsoft, Dell and CoreWeave already operating engineering racks of the NVL72 system, according to Jensen Huang's GTC Taipei keynote. Huang described AI factories as the largest infrastructure buildout in history, with single-site capital costs reaching $50 billion to $60 billion per gigawatt and projected to hit $80 billion to $100 billion. "Tokens are now profitable units of revenues," Huang said, arguing that throughput per watt directly translates to revenue for AI infrastructure operators. He said the supply chain for Vera Rubin is twice as large as the one built for Grace Blackwell, with rack assembly time dropping from two hours to five minutes per rack. NVIDIA also introduced the Vera CPU, an 88-core processor built for agentic workloads that delivers 1.8 times the agentic sandbox performance of x86 CPUs, according to the company.

Taiwan's manufacturers adopt DSX

Taiwan's manufacturing partners are already applying the DSX framework to their own operations. Foxconn is building a $1.4 billion AI cloud supercomputing center powered by 10,000 NVIDIA GPUs and using the DSX-inspired Factory Operations Blueprint to build its MoMClaw manufacturing agent. The company estimates an 80% speedup in root-cause analysis, a 15% increase in labor productivity and a 10% decrease in machine failure rates. Foxconn is also applying NVIDIA's Cosmos and Metropolis platforms for vision AI, boosting first-pass yield by 3%.

Wistron is using the Omniverse DSX Blueprint alongside NVIDIA's PhysicsNeMo framework and Cadence Reality DC Design to simulate burn-in environments for stress-testing across global manufacturing sites. The workflow speeds layout analysis by as much as 70% and cuts facility power demand by 20% through dynamic rack optimization, according to NVIDIA. Pegatron is adopting the same blueprint to connect design data, thermal simulation and digital twins, reducing AI visual inspection deployment time by 67% and operational effort by 10%. Quanta Cloud Technology is using Omniverse-based digital twins to accelerate factory planning, giving engineering, operations and logistics teams shared access to design data for faster layout feedback. TSMC is applying NVIDIA CUDA-X libraries and AI models across computational lithography, transistor simulation and process control, with cuLitho improving cost-effectiveness by 20% to 50% over CPU-based methods.

What it means for investors

The ABB-NVIDIA collaboration positions ABB to capture a larger share of data center electrification spending as AI infrastructure scales. ABB competes with Siemens and Schneider Electric for medium-voltage and power distribution contracts that are growing in both size and complexity as racks demand DC power, solid-state protection and liquid cooling integrated from day one. For NVIDIA, the DSX platform deepens its moat in AI infrastructure by embedding its software stack into the design phase of every new AI factory, creating switching costs for operators who standardize on the blueprint. NVIDIA shares trade at about 35 times forward earnings, reflecting the market's expectation that AI infrastructure spending will sustain its current trajectory through the Vera Rubin cycle and beyond.

This article is for informational purposes only and does not constitute investment advice.