In partnership with
This month, the International Maritime Organization warned that no safe passage exists through the Strait of Hormuz. Traffic is recovering since the April ceasefire, but the security rating stays critical. The 2026 Israel-Iran conflict still grips the chokepoint. About 20% of the world's liquefied natural gas (LNG) normally passes through it, almost all from Qatar. When the strait is unsafe, a fifth of global supply is at risk.
That is the fragile base beneath a confident forecast. The world is building toward an LNG supply peak around 2030. Yet the power grids that artificial intelligence (AI) needs must be ready by 2027. Supply peaks three years after demand binds.
Qatar bets on 142 MTPA by 2030. Its only exit is one strait
Qatar is the swing supplier. QatarEnergy is raising capacity to 142 million tonnes a year (MTPA) by 2030, an increase of almost 85%. The North Field expansion is the largest LNG project on Earth.
But geography is fixed. About 93% of Qatar's LNG exports transit the Strait of Hormuz. There is no pipeline bypass to global markets. Strikes on Ras Laffan in 2026 took about 17% of Qatar's capacity offline, roughly 12.8 MTPA. A bigger peak does not remove the chokepoint. It enlarges the exposure.
North America's LNG doubles to 28 Bcf/d. The molecules still need wires
North America answers with volume. US exporters plan to add 13.9 billion cubic feet a day (Bcf/d) by 2029. North American capacity will more than double, to 28.7 Bcf/d. The additions concentrate on the US Gulf Coast. Golden Pass, Plaquemines, and Rio Grande are under construction.
Gas is only half the chain. Exported LNG earns abroad. Gas burned at home must reach data centres through the grid. The molecules are arriving. The wires are not.
ERCOT is buried under 225 grid requests. Transmission is the wall
Texas shows the bottleneck. ERCOT received 225 large-load interconnection requests in one year, against a system built for 40 to 50. Most come from data centres. The operator now studies them in batches because one-by-one review collapsed.
The binding constraint is not generation. It is transmission. ERCOT can add gas plants faster than it can build the lines to move their power. Gas without wires is stranded. The grid, not the fuel, sets the pace.
Singapore's data centres hit 1.4 GW. New ones must go green
Singapore made the limit explicit. It hosts 1.4 gigawatts (GW) of data-centre capacity across more than 70 sites. New capacity comes only under a green roadmap. It adds at least 300 megawatts (MW), with 200 MW more reserved for green operators. Power, not land, is the binding constraint.
Singapore imports almost all of its gas. So it diversifies LNG suppliers rather than relying on one. A single-supplier strait is a single point of failure. Diversification is the hedge.
Britain's data-centre demand more than doubles to 6% by 2030
The United Kingdom faces the same maths. NESO, the grid operator, projects data centres near 6% of UK electricity by 2030, up from about 2.5% today. London is already Europe's largest data-centre hub.
Developers are not waiting for the grid. Across markets, some turn to combined-cycle gas turbines (CCGT) for primary power. Gas becomes the bridge because the grid cannot connect them in time. The pattern repeats across three continents.
Which grid constraint will bind first in your market?
The bridge fuel arrives in 2030. The bridge is needed in 2027
Here is the timing problem. The fuel meant to bridge the gap arrives after the gap opens. Demand binds in 2027, three years before the supply built to serve it peaks. The bridge is late by design.
Gas fills the first years. Peaker plants and CCGTs can typically be built in 24 to 36 months. They are the fastest firm power available. But they burn fuel that moves through contested water. The bridge is fast, and it is exposed.
Delfin's £3.7bn bet assumes the gas flows by 2030
Project finance carries the risk. In June 2026, the Delfin floating LNG project off Louisiana reached a £3.7bn ($5bn, €4.3bn) final investment decision. Long-term sales to Centrica, SEFE, Vitol, and Gunvor back it. First production is targeted for 2030. Those offtake contracts assume delivery on time. When supply stalls, as Ras Laffan did, that capital is exposed. The three-year gap between contract and power is where projects strand.
Gas peakers cover 2027. Nuclear baseload carries 2028 to 2035
A two-stage bridge is forming. Gas peakers cover the immediate gap, from now to about 2028. They start fast and absorb the first wave of AI demand. They are a hedge, not a foundation.
Nuclear baseload is the second stage, from roughly 2028 to 2035. It is slow to build but firm and domestic. It does not transit a strait. System operators, not algorithms, decide which mix connects and when. The intelligence informs them. The engineers choose.
LNG export, power grids, and AI compute now run on one timeline. None plans for the other two. The exporter builds for 2030. The grid binds in 2027. The data centre signs power contracts in between.
The 2026 disruption was a warning, not an exception. Firm, domestic power is the only supply no strait can close. The winners will build the bridge before the gap opens, not after. Three sectors, one deadline, no coordinator. That is the gap to close.
Next week: Construction reality capture works. The operators discover it 18 months late.
No theory. No slides. Just pipeline.
Most founders know their product. Few know how to get it in front of the right people. In this hands-on session, Clay + HubSpot for Startups walk you through ICP definition, prospect list enrichment, and AI-personalized outreach. You launch your first sequence before the session ends. June 18. 11am ET / 4pm GMT.