To deliver computing power, Oracle uses trucks to transport natural gas to power its data centers

In order to deliver its AI data centers on schedule, Oracle has adopted unconventional energy supply methods to cope with pipeline delays, yet high costs and scalability bottlenecks remain significant concerns.

On October 8, Bloomberg reported that Oracle is using trucks to transport compressed natural gas to maintain the construction and operation schedules of several AI data centers, seeking to avoid time losses caused by pipeline infrastructure delays.

This unorthodox strategy has already been implemented in Utah and Texas, and is under consideration for key projects in New Mexico. Oracle currently has negative free cash flow, and it is expected to remain under pressure until more AI data centers are completed and operational.

Oracle's stock price closed down 5.5% on Thursday at $135.69, its largest single-day drop since July 16.

Meanwhile, Bloom Energy—which has a fuel cell collaboration with Oracle—saw its share price plunge more than 6% on the same day, as market concerns mounted over Oracle’s data center expansion pace and the future of related partnerships.

This is not Oracle’s first attempt at such measures.

According to sources cited in reports, in the data center project in the suburbs of Salt Lake City, Oracle has sustained normal progress for over a year by having natural gas delivered by truck, with services provided by Certarus, a subsidiary of Superior Plus.

Certarus noted that timelines for connecting data centers to both pipelines and the power grid are extending, “creating rapidly increasing demand for energy solutions that can be rapidly deployed and scaled flexibly.”

At the OpenAI data center campus project in Shackelford County, Texas, Oracle adopted the same approach, with VoltaGrid LLC as the supplier. Oracle posted on social media:

Thanks to the VoltaGrid team for finding a cost-effective power supply solution, ensuring customers are able to use OCI (Oracle Cloud Infrastructure) computing power on schedule.

The highest-pressure bottleneck is in New Mexico.

Wallstreet Insight reported that the data center project code-named "Project Jupiter" is facing serious pipeline delays. Pipeline operator Energy Transfer LP had its original plan rejected by New Mexico regulators and must now redesign the route, pushing the expected operational launch from this summer to next year.

To keep the project on schedule, Oracle is considering trucking compressed natural gas to the data center, allowing early-stage operations to begin before the pipeline is fully connected. The urgency is apparent: Last month, Oracle issued a force majeure notice to the project developer. Should delays persist, this move may exempt the company from some payment obligations.

Trucking in compressed natural gas is not a cheap solution.

Jack Weixel, Senior Director of Energy Analysis at East Daley Analytics, estimates that—factoring in labor, specialized equipment, and vehicle fuel costs—the price of truck-delivered compressed natural gas is about four times that of natural gas at major pipeline hubs.

Natural gas must first be extracted from pipelines, compressed and loaded onto trucks, then transported for several hours before being decompressed and injected into generators. For continuously operating data centers, this means non-stop truck dispatches around the clock.

Scalability constraints are also significant.

SemiAnalysis analyst Ellie Holbrook calculates that if Oracle used compressed natural gas trucks to provide 100 megawatts of power for "Project Jupiter," this would represent only about 4% of the project’s planned 2.45-gigawatt total capacity, with each large truck delivering only about 40 minutes’ worth of power—which means dozens of truck trips would be required daily for continuous supply.

For this reason, compressed natural gas trucking has historically been confined to remote industrial settings such as mining or oil drilling that are far from grid connections. However, for Oracle—which has made commitments to clients including OpenAI and urgently needs to deliver computing power on schedule—this may be the most viable transitional solution under current conditions.