Oracle is transporting compressed natural gas by truck to help keep its artificial intelligence data centre projects moving while it waits for pipeline infrastructure to become available. The company has used truck-delivered gas at a site near Salt Lake City, Utah, for more than a year and is also using the approach during early work at a data centre campus being developed for OpenAI in Texas. Oracle is now considering a similar workaround for Project Jupiter, its major AI data centre development in New Mexico, where a delayed pipeline threatens the planned power supply.
The unusual strategy highlights a growing constraint on the AI industry: building computing infrastructure is not enough if developers cannot secure electricity and the fuel needed to generate it. Trucking compressed natural gas can help bridge infrastructure delays, but it comes with substantially higher costs and logistical demands than conventional pipeline delivery. Bloomberg reported on October 8, 2026, that the approach is being used at existing sites and is under consideration for the New Mexico project.
Key takeaways
- Oracle has used truck-delivered compressed natural gas at a data centre near Salt Lake City for more than a year.
- The company is also using trucked gas for early work at a data centre campus being built for OpenAI in Shackelford County, Texas.
- Oracle is considering the approach for Project Jupiter in New Mexico as a pipeline delay threatens its power infrastructure schedule.
- An energy analyst cited by Bloomberg estimates that delivered compressed natural gas can cost roughly four times the price of gas at a major pipeline hub.
- Project Jupiter is designed around up to 2.45 gigawatts of installed Bloom Energy fuel-cell capacity.
- The case illustrates how electricity access, fuel supply and permitting can become bottlenecks for AI infrastructure expansion.
Why Oracle Is Moving Natural Gas by Truck
Large AI data centres require substantial and reliable electricity to operate servers, networking equipment and cooling systems. Developers commonly depend on the electricity grid, dedicated power infrastructure or a combination of both. When grid connections or fuel pipelines are delayed, the facilities may struggle to meet their planned construction and operational schedules.
Oracle is using truck-delivered compressed natural gas as a temporary solution at selected locations. Under this arrangement, natural gas is taken from an existing pipeline network, compressed into specialised trailers and transported by road to the data centre site. The gas is then decompressed and supplied to equipment that uses it to generate power.
The arrangement is often described as a virtual pipeline because trucks provide the link between the existing gas network and a site that does not yet have its permanent pipeline connection.
According to Bloomberg, deliveries helped keep a data centre on the outskirts of Salt Lake City moving for more than a year while Oracle waited for a gas pipeline to be built and integrated with its equipment. The company is also using the approach for initial work at a campus being developed for OpenAI in Shackelford County, Texas.
The distinction between initial construction activity and full-scale operation is important. Trucked gas can help a site make progress before permanent infrastructure is ready, but that does not necessarily mean the arrangement can economically support a large campus at full capacity.
Oracle’s strategy reflects the urgency surrounding its AI infrastructure commitments. Bringing facilities online on schedule is important for cloud providers that have committed computing capacity to customers. Delays can postpone revenue opportunities and leave expensive infrastructure waiting for power.
Project Jupiter in New Mexico Faces a Pipeline Delay
Project Jupiter is one of Oracle’s major planned AI data centre campuses in the United States. Located in Doña Ana County, New Mexico, the development is intended to support large-scale AI computing capacity associated with the broader Stargate initiative.
Oracle announced in April 2026 that the project would use up to 2.45 gigawatts of installed Bloom Energy fuel-cell capacity for an on-site microgrid. The updated design replaced an earlier plan involving gas turbines and diesel generators.
The power system depends on a reliable supply of natural gas. However, the pipeline expected to serve the campus has encountered delays linked to regulatory and routing challenges.
According to an August report by Bloomberg Law, Energy Transfer’s Green Chile pipeline project was then expected to enter service on February 1, 2027, later than its previous target of August 15, 2026. The revised timeline complicated the power arrangements for Project Jupiter.
Bloomberg reported on October 8 that Oracle is considering trucking compressed natural gas to the site to allow early stages of the development to move forward before the pipeline becomes operational.
That proposal is not the same as confirmation that Oracle has begun truck deliveries to New Mexico. The company is already using the strategy in Utah and Texas, while the New Mexico application remains under consideration according to the reporting.
Oracle Has Already Issued a Force Majeure Notice
The pipeline issue is part of a broader set of challenges surrounding Project Jupiter. In September, Oracle issued a force majeure notice to the project’s developer, according to reporting by Reuters and Bloomberg.
A force majeure notice invokes contractual provisions that may provide relief when specified events outside a party’s control interfere with its obligations. In this case, Reuters reported that Oracle was seeking the ability to delay certain payments if the data centre failed to meet its planned 2028 operational target.
Oracle has said Project Jupiter remains on schedule. Blue Owl Capital, whose unit is involved in the development, also said the notice did not change its financial commitments to the multi-year project.
The notice therefore should not be interpreted as confirmation that the project has been cancelled or that Oracle has abandoned its planned opening date. Instead, it underscores the importance of power infrastructure and construction timing to the project’s delivery.
If trucked gas is adopted in New Mexico, it could provide another way to manage the timing gap between data centre construction and the availability of permanent fuel infrastructure. Whether it can support meaningful computing capacity at an acceptable cost remains a separate question.
Trucked Gas Can Cost Four Times More Than Pipeline Supply
The biggest disadvantage of moving natural gas by road is cost.
Jack Weixel, senior director of energy analysis at East Daley Analytics, told Bloomberg that delivered compressed natural gas can cost roughly four times as much as gas purchased at a major pipeline hub once labour, specialised equipment and fuel for the trucks are included.
The additional costs arise at several stages. Gas must be collected from an existing pipeline, compressed, loaded into specialised trailers and transported to the destination. Once there, it must be unloaded or decompressed before it can be used by the site’s power equipment.
The process also requires a steady supply of vehicles and trained personnel. Long travel distances can increase delivery times and reduce the amount of fuel each truck can move over a given period.
| Factor | Pipeline supply | Truck-delivered compressed natural gas |
|---|---|---|
| Delivery method | Continuous flow through fixed infrastructure | Gas transported in specialised trailers |
| Infrastructure requirement | Pipeline connection and associated equipment | Compression facilities, trailers and on-site handling systems |
| Cost profile | Generally lower delivery costs at scale | Can be substantially more expensive |
| Main limitation | Construction, permitting and connection delays | Transport capacity, logistics and repeated deliveries |
| Best potential role | Long-term supply for large facilities | Temporary or specialised supply while permanent connections are unavailable |
Comparison based on the delivery methods described in Bloomberg’s reporting; actual costs depend on location, contract terms and infrastructure.
For a large AI data centre, the volume of fuel required can make the logistical challenge considerable. A site that consumes gas continuously cannot depend on occasional deliveries. Trucks must arrive at a frequency sufficient to prevent interruptions to the power supply.
Bloomberg cited Ellie Holbrook, an energy analyst at SemiAnalysis, who estimated that if Project Jupiter used trucked gas to generate 100 megawatts of electricity, each large trailer would provide enough fuel for approximately 40 minutes of power generation. That estimate illustrates the scale of the delivery challenge; it is not a statement of Oracle’s actual truck schedule or planned operating load.
Even a temporary supply arrangement would therefore need careful coordination between fuel deliveries, storage, decompression equipment and power generation.
Why the AI Data Centre Industry Is Facing Power Bottlenecks
The growth of generative AI has increased demand for specialised computing infrastructure. Training and operating advanced AI models requires large clusters of servers, accelerators and networking equipment, often concentrated in data centres designed to handle significant electricity loads.
Developers are consequently racing to secure land, computing equipment, grid connections and power-generation capacity. These requirements do not always progress at the same speed.
A data centre building can be substantially complete while waiting for a grid connection or fuel supply. Conversely, a power project can be planned but face delays because of environmental reviews, route approvals, equipment availability or construction requirements.
Natural-gas pipelines are particularly relevant to facilities that use gas-powered generation or fuel cells. Although gas may be available elsewhere in the region, the final connection to a specific site still requires infrastructure and approvals.
Oracle’s experience shows how a delay in one part of the energy system can affect the schedule of a much larger technology investment.
The workaround is not unique in concept to data centres. Compressed natural gas has been transported by truck for industrial customers that are not connected to conventional pipelines. What is unusual is the pressure to use this model for AI infrastructure that may ultimately require hundreds of megawatts or more of continuous power.
The Role of VoltaGrid and Certarus
Specialised energy providers are helping Oracle bridge the gap between existing gas infrastructure and data centre sites.
Bloomberg reported that Certarus, a unit of Superior Plus, supplied trucked gas at the Utah location. VoltaGrid provided deliveries for early work at the Texas campus being developed for OpenAI.
These companies operate in a market where data centre developers are looking for rapidly deployable energy solutions while waiting for permanent infrastructure.
Certarus said in a statement to Bloomberg that longer timelines for securing pipeline and grid connections had created a growing need for scalable energy solutions that can be deployed quickly. The company did not comment on specific deals.
Oracle did not directly respond to Bloomberg’s request for comment on the reported strategy. However, the company posted on social media praising VoltaGrid for helping it provide Oracle Cloud Infrastructure capacity on schedule.
The partnerships illustrate how the AI infrastructure boom is creating opportunities beyond server manufacturers and cloud software companies. Fuel transport, on-site power systems, electrical equipment and energy services are also becoming important parts of the supply chain.
For energy providers, the opportunity is to supply customers that place a high value on speed and reliability. For data centre operators, the trade-off is whether the cost of temporary energy delivery is justified by the benefits of bringing facilities online sooner.
What the Strategy Means for Oracle’s Costs
Oracle has made substantial investments in AI infrastructure and has committed to delivering computing capacity for major customers, including OpenAI. Those commitments increase the importance of completing data centres on time.
A delayed facility can create financial pressure because developers may continue to incur construction, financing and other costs before the infrastructure starts generating the expected revenue. The economics can become more difficult if the facility needs an expensive temporary power arrangement as well.
Bloomberg reported that Oracle’s free cash flow was negative and was expected to remain under pressure until more of its AI data centres were completed. Trucking gas adds another operating cost, although the actual financial impact depends on the volume delivered, the duration of use and the amount of computing capacity brought online.
The company must therefore balance two competing considerations. Waiting for permanent pipeline infrastructure could delay the use of expensive data centre assets, while paying a premium for trucked gas could increase the cost of operating them.
A temporary solution may still make commercial sense if it enables a facility to begin generating revenue earlier than would otherwise be possible. But that decision depends on the expected revenue from the available capacity, the cost of delivered fuel and the time until permanent infrastructure becomes available.
The approach should not automatically be interpreted as evidence of a long-term shift away from pipelines. In the locations described by Bloomberg, trucking is being used or considered as a response to specific infrastructure constraints.
Environmental and Community Considerations
Natural gas can provide dispatchable power, meaning generation can be adjusted to meet demand when fuel and equipment are available. Fuel cells, such as those planned for Project Jupiter, generate electricity through an electrochemical process rather than conventional combustion.
Oracle has said the updated Project Jupiter design is expected to reduce nitrogen oxide emissions by approximately 92% compared with its earlier gas-turbine plan. That is a company projection comparing two designs, not evidence that the project has no environmental impact.
Transporting gas by truck also introduces additional fuel consumption, road traffic and logistical requirements. The overall environmental impact depends on factors such as gas production and leakage, transport distances, the efficiency of the power equipment and the amount of electricity generated.
Project Jupiter has faced regulatory scrutiny, including a separate review of the air-quality permit for its proposed microgrid. Oracle’s September statement distinguished the data centre campus from the proposed microgrid, which is subject to its own permitting process.
The project has also attracted attention over its potential impact on local resources and infrastructure. Oracle has said it will fund the project’s energy infrastructure and electricity costs rather than passing those costs to local residents through their electricity bills.
As AI data centre development expands, communities and regulators will continue to weigh economic benefits, energy reliability, environmental effects and local resource requirements.
The Bigger Picture
Oracle’s use of truck-delivered natural gas demonstrates how the AI infrastructure race is forcing developers to look for alternatives when conventional energy infrastructure cannot be delivered on time. The strategy can help a project move forward, but it is expensive and difficult to scale to the power requirements of a major AI campus. The wider lesson is that computing capacity depends not only on access to chips and servers, but also on reliable power, fuel logistics, construction and regulatory approvals.
For Oracle, the immediate objective is to keep its data centre projects moving while permanent supply arrangements are completed. The longer-term economics will depend on how quickly those connections become available, how much temporary fuel delivery is required and whether the revenue from earlier deployment outweighs the additional costs. The New Mexico proposal remains a possibility rather than a confirmed rollout, making pipeline progress and further project disclosures important developments to watch.
Looking Ahead
The Green Chile pipeline’s reported February 2027 target will be an important milestone for Project Jupiter’s fuel supply, although the final schedule will depend on construction and regulatory progress. Any use of trucked gas in New Mexico would need to be assessed against the project’s scale, the capacity that could be supported and the cost of repeated deliveries. Oracle’s statements about the campus remaining on schedule should also be distinguished from independently confirmed completion of the energy infrastructure.
More broadly, the AI industry is likely to continue exploring a mix of grid connections, on-site generation, fuel cells and other energy arrangements as developers compete to bring computing capacity online. Temporary power solutions can bridge gaps, but they cannot eliminate the need for dependable long-term energy infrastructure. Oracle’s experience offers a clear example of how energy supply has become a central factor in the timing and economics of AI expansion.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
