Hormuz Closure Pushed LNG Up 60% and Is Making AI Compute More Expensive

Brent crude oil crossed $100 per barrel on March 8, 2026 — the first time in four years — and peaked at $126 before partial de-escalation pulled it back slightly. European LNG prices rose more than 60 percent. The Strait of Hormuz has been effectively blocked since February 28, when joint US-Israel strikes on Iran triggered retaliatory IRGC warnings prohibiting vessel passage. Twenty percent of global LNG supply moves through that strait every day.

This is not an oil story with a tech footnote. It is a direct cost shock to the AI industry, which runs on power, and increasingly on gas-fired power. And it arrives at exactly the moment when AI data center energy demand is at its highest point in history.

Why the Strait of Hormuz Matters More for LNG Than for Oil

The Strait of Hormuz is 21 miles wide at its narrowest point. It is the only maritime route from the Persian Gulf to the open ocean. Roughly 20 percent of global oil and 20 percent of global LNG passes through it daily. But oil and LNG do not respond to disruption the same way.

Oil has strategic reserves. The US Strategic Petroleum Reserve holds roughly 350 million barrels. The IEA coordinates member country reserve releases during supply shocks. When Hormuz disruptions happened in 2019, reserve releases cushioned price spikes within weeks.

LNG has no equivalent reserve infrastructure. You cannot store LNG the way you store crude oil. Liquefied natural gas must be kept at minus 162 degrees Celsius in specialized cryogenic tanks. The global LNG storage capacity is small relative to the volume of gas that moves through Hormuz daily. When the flow stops, the buffer runs out fast.

This is why CNBC reported that LNG markets may be hit harder than oil by the Hormuz closure. European LNG prices surging 60 percent is the early signal. If the disruption extends beyond 30 days, the curve gets steeper.

The AI Data Center Energy Dependency

New large-scale AI data centers — the kind that run GPT-5, Gemini Ultra, Grok 3, and Claude training runs — consume between 100 and 500 megawatts of continuous power each. Most are in the United States, where the power grid relies heavily on natural gas. Gas-fired power plants generate approximately 43 percent of US electricity.

Tom's Hardware reported directly: "LNG has been used heavily to run new AI data centers such as those run by xAI and OpenAI." When LNG import prices rise, US gas prices follow with a lag — typically 4 to 8 weeks. The Hormuz disruption started February 28. The full cost impact on US data center energy bills will materialize in April and May.

The IEA projects that global electricity demand from AI, computing, and cryptocurrency could reach 1,050 TWh by 2026 — roughly equivalent to the entire electricity consumption of Japan. Running that demand on a grid where fuel input prices have risen 60 percent is arithmetically significant.

Three Shortages Developers Have Not Heard About Yet

The Hormuz disruption creates three supply chain problems beyond energy prices that directly affect hardware procurement:

Aluminum shortage: Aluminum is a primary material in server chassis, cooling systems, and data center rack infrastructure. A significant share of global aluminum production uses energy-intensive smelting powered by Gulf gas. Tom's Hardware flagged aluminum as a Hormuz-exposed material specifically because Gulf aluminum producers — particularly in Bahrain and the UAE — rely on cheap local LNG for smelting. Higher gas prices squeeze margins; prolonged disruption threatens output.

Helium shortage: Qatar is the world's second-largest helium producer after the US, and Qatari helium exports move through Hormuz. Helium is not optional in semiconductor manufacturing — it is used in the growth of silicon wafers, in leak detection, and as a carrier gas in chip fabrication processes. A sustained Qatari helium supply disruption would pressure TSMC, Samsung, and SK Hynix simultaneously. Tom's Hardware identified this as potentially the most underreported consequence of the Hormuz crisis for the chip industry.

Rare earth and chemical feedstocks: Several chemical precursors used in photolithography and chip packaging are refined in Gulf facilities. Disruption to these supply chains has longer lead times than energy price spikes — factories can absorb 30 days of supply disruption from existing inventory but face production changes if the disruption extends to 60 or 90 days.

What the Cost Impact Looks Like for AI Inference

The chain of impact on AI inference costs runs like this:

LNG disruption raises European gas prices by 60 percent. US gas prices follow in 4 to 8 weeks. Gas-fired power becomes more expensive. Data center electricity costs rise. Hyperscalers pass a portion of those costs through to customers in compute pricing — though typically with a 1 to 2 quarter lag.

For developers running significant inference workloads on OpenAI, Anthropic, Google, or AWS, the practical question is whether API pricing adjustments follow if the energy cost increase is sustained. Historically, hyperscalers have absorbed energy cost volatility in margins rather than immediately repricing APIs. But a prolonged Hormuz disruption lasting 60 to 90 days is not the kind of cost shock that gets absorbed silently.

The more immediate impact is on self-hosted inference. Teams running on-premise GPU clusters or dedicated cloud capacity with fixed contracts are somewhat protected in the short term. Those on variable-rate compute or spot instances will see the energy cost signal faster.

Countries Most Exposed

CNBC identified the countries most directly exposed to a Hormuz closure based on LNG dependency:

Japan and South Korea are the most exposed among major AI-producing countries. TSMC's Japanese fabs, Samsung and SK Hynix's Korean facilities, and the dense concentration of AI data centers in both countries are all energy-intensive operations running in countries with high Gulf LNG dependency.

Historical Precedent: What the 2019 Tanker War Tells Us

The current Hormuz crisis is not without precedent. In 2019, a series of attacks on oil tankers in the Gulf of Oman — widely attributed to Iran — created a similar episode of shipping disruption. Six tankers were attacked between May and June 2019. Oil prices spiked 4–5 percent within 24 hours of each incident. The disruption lasted approximately 8 weeks before shipping insurers, military escorts, and diplomatic de-escalation stabilized the situation.

The key difference between 2019 and now is the AI data center energy dependency. In 2019, data centers were a growing but still modest share of global electricity consumption. By 2026, AI training and inference account for a meaningful fraction of the load growth on gas-dependent grids. The same energy price spike hits a much larger surface area than it did seven years ago.

The 2019 precedent suggests disruptions of this type typically resolve within 8 to 12 weeks under diplomatic pressure. If that pattern holds, the current Hormuz disruption would ease by late April or early May 2026 — before the full downstream impact on US data center energy bills fully materializes. If it does not hold, and the conflict becomes protracted, the energy market impact enters genuinely uncharted territory for the AI industry.

How Long Until Data Center Energy Bills Actually Reflect the Spike

Understanding the transmission lag matters for planning. The chain runs with these approximate timelines:

Weeks 1–2: LNG spot prices spike in European and Asian import markets. Qatar-linked contracts reprice on next delivery. European gas hub prices adjust immediately — the 60 percent TTF increase happened within days of the Feb 28 disruption.

Weeks 3–6: US domestic gas prices begin reflecting reduced LNG export availability as European buyers bid more aggressively for cargoes, pulling LNG away from US export terminals. Henry Hub futures begin shifting.

Weeks 4–8: US utility power purchase agreements with gas-fired generators reflect higher input costs at next pricing interval. Utilities with forward-hedged gas exposure are buffered; those on spot pricing are not.

Weeks 6–12: Data center operators on variable-rate utility contracts see electricity bill increases. Operators with long-term fixed-rate PPAs are protected until renewal.

Q2–Q3 2026: Hyperscalers with variable energy exposure model whether to reprice API compute costs or absorb the margin hit. Historically they absorb for one to two quarters before adjusting.

For teams running significant inference workloads on major AI APIs, the practical window to lock in current pricing through reserved capacity or committed use discounts is approximately now through end of March 2026. After that, the LNG cost signal will have had enough time to propagate through to data center operators.

Key Takeaways

• Brent crude hit $126/barrel peak after Feb 28 Hormuz disruption; European LNG prices up 60 percent
• 20 percent of global LNG moves through Hormuz daily — Qatar is the primary exporter and has no bypass route
• AI data centers running on gas-fired power will see energy cost increases materialize in April to May 2026 with a 4 to 8 week lag
• Three hidden shortages: aluminum (server hardware), helium (semiconductor manufacturing), chemical feedstocks (chip fabrication)
• Japan and South Korea are the most exposed major AI hardware-producing nations due to Gulf LNG dependency
• For developers: Self-hosted inference on variable-rate compute is most exposed to near-term cost increases. Fixed-rate contracts and reserved instances provide a short-term buffer. Monitor hyperscaler pricing announcements in Q2 2026.
• What to watch: Whether the disruption extends past 30 days — at that point, LNG storage buffers in Europe and Asia begin running thin and the price curve steepens sharply.