Iran tensions threaten ai boom's fragile finances

Donald Trump’s saber-rattling over Iran’s access to the Strait of Hormuz initially appears to be about gasoline prices—a predictable political calculation. But the potential fallout extends far beyond the pump, threatening to destabilize a sector already teetering on the edge: artificial intelligence.

The energy crunch and ai's insatiable appetite

The immediate impact will be higher energy costs globally, a reality analysts are bracing for even with a swift resolution. This isn't merely about filling up cars; it’s about powering the vast datacenters that underpin the AI revolution. For an industry whose very business model relies on ever-increasing computing power, and which is fueled by staggering amounts of debt, this presents a uniquely acute challenge.

Sam Altman, CEO of OpenAI, offered a peculiar analogy in February, comparing the energy demands of training AI models to the energy required to raise a human being. “It takes about 20 years of life – and all the food you consume during that time – before you become smart,” he remarked. A sentiment that, in retrospect, underscores the unsustainable nature of the current AI growth trajectory.

A financial house of cards?

The Bank of England, with characteristic foresight, flagged the potential link between energy costs and AI company valuations in a recent risk assessment. Even before Trump’s escalation, investors were growing wary of the sector’s mounting debt and questionable returns on investment. Now, the conflict adds another layer of uncertainty, particularly given the energy-intensive nature of AI’s supply chains.

The scale of the financial engineering at play is frankly alarming. Quinn Emanuel, a US law firm, recently highlighted the sector’s revenues of $60 billion against a capital expenditure of $400 billion – a ratio that would raise eyebrows in almost any other industry. The structure resembles the pre-2008 financial crisis, with off-balance sheet vehicles and asset-backed securities obscuring the true level of risk.

Datacentre operators, in particular, have been aggressively using these complex structures to borrow vast sums, often from private credit firms whose opacity makes it nearly impossible to track the total liabilities of these companies. As much as $120 billion in datacentre debt has been shifted off-balance sheet in just the last two years, creating a web of interconnected risk that, as Quinn Emanuel aptly put it, means “distress at any single node…can propagate across multiple counterparties and financing layers.”

Beyond the us – a global shockwave

While the United States can largely insulate itself from the immediate energy crisis thanks to its oil production, the ripples will be felt worldwide. Egypt faces curfews, Indonesia mandates work-from-home Fridays, and the Philippines has declared a national energy emergency. The WTO estimates that 70% of investment growth in the US last year was in AI-related goods. A prolonged period of high energy prices could, as WTO chief economist Robert Staiger warned, “crimp” investment in the sector.

The fundamental question remains: can the AI sector ever justify its current, stratospheric valuations? Even a modest increase in energy costs could trigger a reassessment—a cascade of consequences that could reverberate throughout US markets and beyond. Trump's actions may have unleashed forces far beyond his control, revealing the fragility of a boom built on increasingly precarious foundations.