Oracle has turned to the road to keep its ambitious artificial intelligence expansion on track. The company trucks compressed natural gas directly to server sites where pipelines have yet to arrive. This workaround, costly and logistically complex, buys time in a market where every month of delay threatens competitive standing.

One site outside Salt Lake City operated on trucked fuel for more than a year. Deliveries from suppliers such as Bloomberg reported, citing people familiar with the matter, allowed construction and early operations to continue while crews completed permanent infrastructure. The same approach now supports initial phases at a campus Oracle is building for OpenAI in Shackelford County, Texas. Suppliers including Certarus, a unit of Superior Plus, and VoltaGrid have handled the logistics.

But the practice comes at a steep price. Trucked gas runs roughly four times the cost of gas delivered through a major hub once labor, equipment and the trucks’ own fuel consumption factor in. Analysts at East Daley Analytics shared that assessment with Bloomberg. Powering just 100 megawatts would require each large trailer to supply only about 40 minutes of electricity, according to calculations from SemiAnalysis. Scale that across hundreds of megawatts and the operation resembles a constant fuel convoy.

Now Oracle eyes the same solution for Project Jupiter. That $165 billion AI campus in Doña Ana County, New Mexico, could eventually draw as much as 2.45 gigawatts. A dedicated pipeline known as the Green Chile Project was supposed to deliver the fuel. Its original in-service date of August 2026 slipped to February 2027 after the New Mexico State Land Office rejected the proposed route twice. Crypto Briefing noted the permitting setbacks forced the change. Oracle responded last month with a force majeure notice to the project’s developer, shielding itself from certain contractual penalties if timelines stretch further.

Publicly the company maintains Project Jupiter stays on course for a 2028 completion. Privately the contingency planning reveals pressure. AI training clusters demand uninterrupted power. When grid connections or gas lines lag, hyperscalers improvise. Oracle’s move mirrors broader industry strain. Data center operators face queues stretching years in parts of the United States and Europe alike.

The TNW article drew direct parallels across the Atlantic. In Europe similar bottlenecks plague new facilities. Britain’s Nscale project, backed by Microsoft, cannot secure 90 megawatts in time for its planned 2027 opening. Some grid connection waits approach a decade. Denmark’s Energinet halted all new agreements this year. The country confronts a 60-gigawatt queue against peak demand of only 7 gigawatts. Rather than trucking fuel, Spain proposes strict rules. A draft decree would block grid access for any data center above 1 megawatt that fails to source 80 percent of its hourly demand from renewables. Projects already in permitting would have six months to comply or lose approval without compensation. The Next Web detailed these contrasting regulatory responses.

Oracle’s reliance on mobile gas highlights a temporary bridge many expect to see repeated. Fuel cells from Bloom Energy form another piece of the strategy. The company has contracted for up to 2.8 gigawatts of such systems, with 1.2 gigawatts already committed. These units burn the delivered gas on site to generate electricity close to the servers. The setup avoids some transmission losses but still depends on steady fuel supply. A single disruption in the truck schedule could idle expensive hardware.

Recent market reaction proved telling. Oracle shares dropped about 5.5 percent following the initial reports on October 8. Investors weighed the higher operating costs against the risk of stalled capacity. The company has locked in substantial revenue from AI cloud contracts. Activating that capacity quickly matters more than short-term margin pressure for some analysts.

Yet the practice raises longer-term questions. Trucked deliveries increase road traffic, emissions from diesel trucks and operational complexity. They also expose operators to fuel price swings and supply chain risks. One Kentucky incident recently trimmed U.S. gas production by nearly 2 percent. An approaching hurricane threatened further disruption in the Gulf. Such events turn a stopgap measure into a vulnerability.

Oracle isn’t alone. Multiple hyperscalers explore similar stopgaps as they race to deploy graphics processing units by the hundreds of thousands. The Salt Lake City precedent shows the method can work. Over a year of reliable deliveries kept that project moving until the pipeline arrived. Texas operations for OpenAI demonstrate it scales to early-stage AI campuses. Success there may encourage wider adoption even as permanent infrastructure catches up.

Europe’s path diverges. Regulators there push harder on renewables and efficiency. Spain’s proposed hourly matching requirement would force data centers to pair with massive battery storage or dedicated clean generation. That raises capital costs but aligns with climate targets. The continent’s tighter grid leaves less room for gas-based improvisation. Oracle’s European facilities, while not yet reported to use trucked gas, face the same interconnection delays that prompted the U.S. workaround.

The episode underscores how infrastructure has become the binding constraint on artificial intelligence growth. Compute is available. Talent exists. Capital flows freely. But electrons and molecules lag. Pipeline builders, grid operators and permitting offices now hold unexpected sway over technology roadmaps. Oracle’s decision to accept fourfold fuel costs signals confidence that AI revenue will justify the expense.

And the trucks keep rolling. For now they represent both a clever adaptation and a symptom of deeper bottlenecks. How long operators tolerate the premium will determine whether these virtual pipelines become standard practice or fade once steel pipes finally arrive. Recent coverage from Hoodline on October 9 reinforced the New Mexico focus, noting the 2.5-gigawatt scale and Bloom Energy partnership. The story evolves daily. Oracle’s bet on speed at any cost continues.