Organisations turn to integrated simulation to tackle 300bn equipment backlog and 200GW AI power queue

Artificial intelligence expansion is currently hindered by severe power infrastructure bottlenecks, causing significant delays for organisations attempting to deploy new facilities. According to research published on the NVIDIA blog, this constraint has become the primary obstacle in modern infrastructure development, affecting global deployment timelines and capital expenditure plans.

The scale of the disruption is substantial, with more than $300 billion in equipment sitting in backlog and over 200 gigawatts of projects stalled in United States interconnection queues. These figures highlight the acute mismatch between the rapid demand for AI compute capacity and the existing capabilities of electrical grids and facility power systems.

What we know

The root cause of these delays is identified as siloed system design, where the electrical grid, the physical facility and the internal compute systems have historically been planned independently. This separation creates conditions for operational risks, stranded capacity and prolonged construction timelines. To address this fragmentation, NVIDIA introduced the DSX AI Factory Platform, which allows engineers to simulate and operate the grid and AI factory as a single, cohesive system from the substation down to the server rack.

Several major energy and technology partners are integrating their solutions with this platform to validate infrastructure before construction begins. GE Vernova is deploying its grid-to-facility expertise to help customers transform power constraints into competitive advantages. Hitachi Energy has qualified its battery energy storage systems as NVIDIA DSX Ready, ensuring they meet reference architecture requirements for resilient and scalable AI factories. Schneider Electric, in collaboration with AVEVA and ETAP, is advancing the platform with validated power and cooling reference architectures and digital twin capabilities.

Siemens Energy is developing a reference design for grid-aware power infrastructure aligned with NVIDIA DSX Vera Rubin AI factory requirements. The company is also exploring the use of NVIDIA technologies for its Noedra digital twin platform to enhance grid monitoring and prevent unplanned outages. Phaidra has integrated DSX components into its self-learning agents, reportedly delivering about 10 per cent more compute within the same power envelope by optimising cooling and power allocation.

Emerald AI integrated its Conductor platform into DSX Flex to help utilities access underused grid capacity. This approach is modelled to unlock up to 100 gigawatts on the existing United States grid, potentially alleviating some of the pressure on interconnection queues. These collaborations aim to turn AI factories from static loads into flexible grid assets, synchronising demand with grid conditions and onsite generation.

Defender's checklist

• Verify that power infrastructure designs account for grid interconnection lead times.
• Ensure facility cooling and power systems are modelled together, not in isolation.
• Check for compatibility with industry-standard digital twin simulation platforms.
• Assess whether current power contracts allow for flexible load management.

The background in plain language

Teams usually plan electrical grids, building facilities and computing hardware as separate projects, which often leads to mismatches in capacity and timing. When these systems are designed in silos, engineers may discover late in the process that the power supply cannot support the compute load, or that the building’s cooling cannot handle the heat generated. Integrated simulation allows organisations to test how these three elements interact before breaking ground, reducing the risk of costly delays and ensuring that power, cooling and compute are balanced from the outset.

What to do and how to stay safe: NVIDIA

• Review current infrastructure plans to identify dependencies between grid connections and facility builds.
• Engage with energy providers early to understand interconnection queue lengths and capacity constraints.
• Evaluate digital simulation tools that can model the interaction between power supply and compute demand.
• Monitor vendor announcements for integrated platform solutions that unify grid and facility management.

Step-by-step guide: AI Governance for Small Teams: Practical Controls and Cost

General security guidance from the TechWire Beat newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Why are AI projects facing power bottlenecks?

The bottleneck arises because the electrical grid, the physical facility and the compute systems have traditionally been designed independently, creating delays and stranded capacity.

What is the scale of the current backlog?

More than $300 billion in equipment is backlogged, and over 200 gigawatts of projects are waiting in US interconnection queues.

How does NVIDIA DSX address this issue?

The platform allows engineers to simulate and run the grid, substation and AI factory as one system, validating components before construction and syncing demand with grid conditions.