Oil is the original API

On March 4, 2026, Iran declared the Strait of Hormuz closed. Within weeks, Brent crude surged past $126 a barrel, the highest in years. US gasoline prices jumped 35%, topping $4 a gallon. Analysts started whispering about $200 oil. And somewhere in a data centre in Virginia, an AI model kept running inference, burning through megawatts of electricity generated by natural gas that suddenly got a lot more expensive. The Strait of Hormuz crisis is a reminder that the physical layer always wins. Every abstraction we build, from cloud computing to large language models, ultimately depends on something tangible flowing through a chokepoint somewhere. Oil might be the oldest commodity on Earth, but it behaves a lot like a modern API: standardised, globally depended upon, and catastrophically disruptive when it goes down.

One chokepoint, global consequences

The Strait of Hormuz is a narrow waterway between Iran and Oman. Nearly 20 million barrels of oil pass through it every day, roughly 20% of global supply. When the US and Israel launched strikes on Iran on February 28, 2026, the strait became a war zone. Iran responded by shutting it down, threatening and attacking ships attempting transit. The consequences were immediate. Brent crude spiked from around $70 pre-war to $126 at its peak. The International Energy Agency called it the largest disruption to energy supply since the 1970s oil crisis. Goldman Sachs estimated that 16 million barrels per day of oil flows were at risk, with only 4.2 million barrels per day able to be rerouted through existing pipelines. Nobel Prize-winning economist Paul Krugman put it bluntly: "The scary scenarios are, unfortunately, extremely plausible. It's not at all hard to tell a $150 story, and it's not crazy to go to $200." This isn't just a geopolitical story. It's an infrastructure story. One physical chokepoint disrupted the entire global digital economy.

Oil as an API

If you squint, oil looks a lot like an API. It has a standardised interface. A barrel of Brent crude is a barrel of Brent crude, whether it's extracted in Saudi Arabia or the North Sea. It's fungible, globally traded, and priced in real time. Every downstream system, from power grids to petrochemical plants to data centres, depends on it. It has global dependency. Nearly every country on Earth imports or exports oil. The entire modern economy is a distributed system built on the assumption that oil keeps flowing. Remove it and the cascading failures are immediate. It has single points of failure. The Strait of Hormuz, the Suez Canal, the Strait of Malacca. These are the load balancers of the physical world, and they have no failover. When Hormuz goes down, there's no automatic rerouting to a backup endpoint. And it has rate limiting. OPEC has been rate-limiting oil supply for decades, adjusting production quotas to manage price and demand. The Hormuz closure is just the most extreme version of this: rate limited to zero. The analogy breaks down in one important way, though. When a software API goes down, you get a 503 error and maybe a frustrated tweet. When the oil API goes down, economies collapse, wars escalate, and people can't afford to heat their homes.

The irony at the heart of AI

Here's the part that should make every AI company uncomfortable: the industry racing fastest toward "the future" is completely dependent on the oldest commodity. Data centres consumed an estimated 415 terawatt-hours of electricity globally in 2024, roughly 1.5% of worldwide demand. By 2030, the IEA projects that figure will more than double to 945 TWh. A single hyperscale AI data centre can demand 300 to 500 megawatts, comparable to the power consumption of a mid-sized city. And where does that power come from? Fossil fuels provide roughly 60% of electricity to data centres globally, according to the IEA. In the US, natural gas alone supplies over 40% of data centre electricity. Coal covers another 15%. When oil and gas prices spike, electricity prices follow. US electricity prices jumped 6.9% in 2025, more than double the headline inflation rate, and Goldman Sachs projects they'll keep rising through the end of the decade as data centres drive 40% of electricity demand growth. This means the Hormuz crisis isn't just an energy story. It's an AI cost story. When gas prices rise, electricity prices rise. When electricity prices rise, cloud costs rise. When cloud costs rise, inference pricing rises. The ripple from a naval blockade in the Persian Gulf reaches all the way to the price of a ChatGPT subscription.

Nuclear as the new cloud

The tech industry knows it has a problem. Microsoft, Amazon, Google, and Meta are all investing billions in nuclear power. Microsoft struck a deal to restart Three Mile Island to power its data centres. Amazon invested $650 million in a campus next to Pennsylvania's Susquehanna nuclear plant, with plans to expand it into a $20 billion AI-ready facility. Google partnered with Kairos Power to build small modular reactors, targeting 500 megawatts by 2030. The logic is straightforward. AI needs baseload power, the kind that runs 24/7 without interruption. Solar and wind are intermittent. Natural gas is exposed to the very geopolitical risks we're watching unfold. Nuclear is the only carbon-free source that can deliver reliable, massive-scale power. But nuclear takes time. Google's first Kairos reactors aren't expected until 2030. The small modular reactor designs that everyone is excited about are still largely unproven at scale. Meanwhile, the data centres keep getting built, and they keep plugging into natural gas. The energy problem isn't solved. It's getting worse. Every new data centre announced is a bet that the grid can handle it, and right now, the grid is powered by the same fossil fuels that a single military conflict can disrupt overnight.

The Singapore lens

If you want to understand what energy vulnerability looks like at a national scale, look at Singapore. A March 2026 report by Energy World Mag ranked Singapore as the most vulnerable country in the world if global energy supplies were disrupted, scoring 85.2 out of 100 on their vulnerability index. The country imports more than 95% of its energy. Natural gas accounts for 93.1% of its electricity fuel mix. Nearly all of that gas is imported. When Qatar's LNG production was disrupted by the Hormuz crisis, Singapore faced immediate pressure on power bills. Wholesale electricity prices, already ranging between SGD 100 and SGD 200 per MWh in 2025, became even more volatile. Singapore is responding. The government has approved plans to import up to 6 GW of low-carbon electricity from neighbouring countries by 2035, including an ambitious project to import solar power from Australia via a 4,300-kilometre subsea cable. But these projects are years away from completion. For now, Singapore remains what it has always been: an energy-dependent island nation in a world where energy supply chains are fragile. It's a reminder that supply chain resilience isn't just a logistics problem. It's an existential one.

Every abstraction hides a physical dependency

This is the lesson that matters most for builders. We talk about "the cloud" as if it's ethereal. We talk about AI as if it runs on math alone. We build abstractions on top of abstractions, and after enough layers, we forget what's at the bottom. At the bottom is energy. And energy, for now, means fossil fuels. Fossil fuels mean tankers, pipelines, and narrow straits controlled by governments with their own agendas. The entire stack, from a GPT-4 response to a React component rendered in a browser, depends on molecules of natural gas being extracted, shipped, and burned. The Hormuz crisis didn't break the internet. But it showed how thin the margin is. A sustained closure, $200 oil, rolling blackouts in energy-dependent regions, these aren't science fiction. They're scenarios that energy analysts are modelling right now. If you're building software, you probably don't think about where your electricity comes from. You should. Not because you can fix it, but because understanding your dependencies, all of them, including the physical ones, is what separates resilient systems from fragile ones. Oil is the original API. It's globally deployed, universally consumed, and when it returns a 503, there's no retry logic that can save you.