Business news |
By Brian Tristam Williams
NVIDIA has backed the latest US regulatory move on grid access, after the Federal Energy Regulatory Commission issued show-cause orders aimed at speeding the connection of data centres, manufacturing plants and other large electricity users.
The FERC large-load interconnection action directs six regional grid operators under FERC jurisdiction to justify their existing tariff rules or file changes that address large users more directly. The order is aimed at AI data centres and other power-intensive facilities that are now arriving in grid queues at a scale utilities were not built to process quickly.
NVIDIA welcomed the action, arguing that AI factories, semiconductor support systems and advanced manufacturing facilities need faster routes to electricity supply while also helping to fund grid upgrades, add generation and operate more flexibly. The company’s framing is predictably pro-AI, but the regulatory problem behind it is real: large computing loads are starting to collide with long transmission planning cycles.
FERC’s orders cover PJM, MISO, SPP, CAISO, ISO-NE and NYISO. They do not cover ERCOT in Texas, which sits largely outside federal jurisdiction. Each operator and its transmission owners now has 60 days to explain why existing rules remain just and reasonable, or to propose tariff revisions.
The five areas under review include transmission service application and study processes, transparency around grid upgrade costs, co-location agreements and behind-the-meter generation, flexible transmission services for large loads, and studies for generation facilities serving nearby or co-located loads. A separate FERC docket describes large loads generally as demand above 20 MW.
FERC has also ordered each grid operator and its transmission owners to file an informational report within 30 days explaining how they intend to ensure enough generation is available for existing and new large loads.
NVIDIA says large customers should be treated less as passive demand and more as participants in the power system. In practice, that means paying for required network upgrades, bringing additional generation to the grid, and offering flexible load that can be shifted or curtailed when grid conditions are tight.
The argument connects with NVIDIA’s separate work with Emerald AI on flexible AI factories. The companies have said commercial-scale deployment of their DSX Flex approach is expected later this year at the NVIDIA AI Factory Research Center in Virginia, using NVIDIA Vera Rubin infrastructure. The pitch is that some AI workloads can respond to grid signals without simply behaving like fixed industrial load.
As we reported previously when AI data centres drove power trends at PCIM, accelerator-heavy computing is already pushing interest in higher-voltage power architectures, energy storage, solid-state transformers and more efficient conversion stages. FERC’s action sits one level above that hardware story: before power can be converted inside the facility, the facility has to get connected.
The FERC move follows an October 2025 direction from US Energy Secretary Chris Wright to begin rulemaking around faster connection of large loads, including data centres. It also follows FERC’s separate approval of PJM’s expedited interconnection track for selected capacity projects, which is due to take effect on July 31, 2026.
The policy tension is straightforward enough. AI infrastructure developers want speed-to-power. Utilities and regulators want reliability, cost transparency and protection for existing customers. The FERC large-load interconnection orders do not settle that balance, but they force the main regional grid operators to put their rules on the table.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。