A social media post from SemiAnalysis raised the question of whether an AI datacenter campus can operate on natural gas delivered by truck. The answer, based on comments from Ellie Holbrook, the research firm’s gas-market analyst for the AI buildout, is that trucking may work as a temporary measure but does not scale easily to the largest facilities.
That matters because the race to build AI infrastructure is often described as a contest for chips, land and electricity contracts. Fuel logistics may be just as important. A site can have generators installed and a nominal power allocation on paper, yet remain unable to support a large computing campus if the fuel supply is too slow, too local or too vulnerable to disruption.
The trailer problem
The most vivid illustration comes from Bloomberg’s reporting, republished by Yahoo Finance. Oracle has used truck-delivered compressed natural gas at a facility in the Utah area and at an OpenAI campus in Texas, the report said. Oracle has also considered the approach for Project Jupiter, its planned large-scale AI infrastructure project.
But Holbrook estimated that one large trailer would last only about 40 minutes when serving a facility drawing 100 megawatts. That calculation turns what sounds like a flexible workaround into a demanding industrial operation. A campus running continuously would need a steady stream of deliveries, along with equipment for unloading, storage and backup supply.
The challenge grows rapidly with the size of the site. AI campuses are being planned around clusters of power-hungry accelerators, rather than a handful of conventional servers. Every additional building increases the amount of fuel that must arrive without interruption. A delivery model that is useful during construction or while a permanent connection is delayed may become impractical once the campus reaches full operation.
The issue is not simply the number of trucks on the road. Local gas availability can limit how much fuel suppliers can provide. Weather, traffic, equipment failures and driver availability can all affect deliveries. The site would also need redundancy, since a missed shipment could reduce computing capacity or force operators to shut down part of the campus.
SemiAnalysis challenges the workaround
In its analysis, SemiAnalysis wrote that trucking compressed or liquefied natural gas is not viable for very large campuses. Holbrook and her colleagues said proven compressed natural gas delivery is roughly limited to 200 megawatts, while local supply and the availability of trucks may impose additional constraints.
That is the critical argument at the center of the debate. Trucked gas can make a project appear faster to power because it may avoid waiting for a new pipeline or a major grid upgrade. Yet the apparent speed comes from shifting the bottleneck somewhere else. Instead of waiting for infrastructure to be built, operators depend on a transportation system that must function continuously at an unusually large scale.
SemiAnalysis’s position does not mean trucked fuel has no role. It can help a smaller facility begin operating, provide temporary generation or keep a project moving while utilities complete more durable connections. The concern is that temporary arrangements may be treated as a long-term answer for campuses whose demand is measured in hundreds of megawatts or more.
That distinction is important for communities and investors evaluating proposed sites. A developer may describe a location as having access to power because generators can be installed quickly. But the practical question is whether those generators can receive enough fuel every day, including when demand expands beyond the original plan.
Oracle points to a different destination
Oracle’s own materials suggest that trucked gas is not necessarily the permanent solution for its biggest projects. On its data-center page, Oracle identifies VoltaGrid power infrastructure at its Texas sites. The same page describes its Doña Ana County, New Mexico, campus as designed around gigawatts of Bloom Energy fuel cells, rather than presenting trucked gas as the final power arrangement.
Oracle has also changed the proposed design for Project Jupiter. In an announcement opening public review of an updated power plan, the company said it replaced an earlier gas-turbine and diesel-generator design with a Bloom Energy fuel-cell microgrid. Oracle described the revised system as lower-emissions, water-efficient and intended to provide reliable on-site power.
That approach may reduce some of the difficulties associated with conventional generators, but it does not eliminate the underlying need for a dependable energy supply. Fuel cells still require fuel logistics, and large campuses still need systems that can operate through interruptions and changing demand. The technology changes the way power is produced, not the requirement that energy reach the site consistently.
The broader lesson is that “available power” is not a single figure. For an AI company, it means power that can be delivered in the right quantity, at the right time and with enough redundancy to support expensive computing equipment. Trucked gas may help bridge a gap, but Holbrook’s assessment suggests it is unlikely to bridge the entire distance for the largest AI campuses.
Daniel Reyes is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Daniel Reyes's usual length and in Daniel Reyes's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.