DPresident Donald Trump’s immediate concern as he calls for Iran to reopen the Strait of Hormuz may be rising U.S. gasoline prices, but if the conflict drags on, the rise in energy costs will be felt far beyond the pump.
Systematic electricity price increases and supply chain disruptions will put pressure on industries and consumers around the world. For the United States, one of the consequences could be that the AI boom threatens its fragile economy.
Many oil importing countries, especially those in the Global South, are having to consider a complete shortage of oil and its products. Shops in Egypt faced a curfew, Indonesia imposed a work-from-home order on Friday and the Philippines declared a national energy emergency.
As a wealthy oil exporter, the United States can largely avoid these concerns. However, as the rising cost of refueling cars in the United States shows, global increases in energy costs cannot be completely avoided. Many analysts now believe this rise will continue for months, even if the Strait reopens within days.
As a result, many companies will be looking anxiously at their cash flow forecasts. But the challenge can be particularly acute in this energy-intensive industry, where business models are not yet firmly established and investments are financed by large amounts of debt.
In February, OpenAI’s Sam Altman made a less-than-comforting comparison in an attempt to allay concerns about AI’s environmental impact ahead of its expected major market launch later this year.
“People talk about how much energy it takes to train an AI model, but it also takes a lot of energy to train a human,” he said. “It takes about 20 years of your life and all the food you consume during that time to become wiser.”
The Bank of England last week highlighted the potential link between energy costs and AI companies’ share prices in its regular review of the risks facing the UK financial system.
The central bank’s Monetary Policy Committee began by noting that investors were already asking questions about the sector before President Trump went to war. “Prior to the dispute, rising debt financing needs and concerns about whether expected returns on highly significant AI-related investments would be realized led to selling pressure,” the report said.
“Conflicts could exacerbate these concerns, especially given the energy-intensive nature of key component supply chains and data center operations.”
This was one aspect of a broader warning that a war with Iran could further exacerbate existing vulnerabilities in the market, given that it could “strain growth, raise inflation and tighten financial conditions.”
Robert Steiger, the World Trade Organization’s chief economist, also drew a link between AI and the effects of conflict, telling me last month that persistently high energy prices could “chill” investment in the sector. “Booms are very energy-intensive,” he says.
To underline the real-world impact of a possible retrenchment, the WTO calculated in its latest World Trade Outlook that 70% of US investment growth in the first three quarters of last year was in some form of AI-related products.
The sheer complexity of the financial engineering behind the AI investment megaboom was revealed in a forensic note published last month by US law firm Quinn Emanuel. The note began by pointing out that last year the sector had revenues of around $60bn (£45.3bn) and capital expenditure of $400bn.
For those of us who remember the 2008 global financial crisis, this article is somber. Off-balance sheet special purpose vehicles are getting a lot of attention, as are asset-backed securities.
Essentially, AI spearhead “hyperscalers” and infrastructure providers like CoreWeave are borrowing unimaginable amounts of money in a rush to build data centers (although a recent analysis by AI skeptic Ed Zitron suggests that real-world projects are far behind their promise).
Lenders are often private companies such as asset management companies, making it difficult for regulators and even investors to track the total amount of debt owed by each company.
There are separate but interconnected concerns about the activities of this burgeoning private credit sector, and regulators, including the Bank of England, have consistently warned of its opacity.
In some cases, technology companies have obediently issued corporate bonds. But a far more Byzantine arrangement, familiar in the run-up to a crash, is in fact taking place.
Data center operators are creating off-balance sheet special purpose vehicles that “own” vast data centers and future rental income, and borrow and lend them. In some cases, these liabilities are pooled, split up and resold to pension funds or investment managers.
As older readers may remember, this type of structure creates a false sense of security that risk is distributed rather than cumulative, and can make it very difficult to know exactly who owes what to whom.
Quinn Emanuel analysts believe that approximately $120 billion of data center debt has been taken off balance sheet over the past two years. And they say: “A deeply interconnected AI ecosystem means difficulties at a single node can propagate to multiple trading partners and funding tiers.”
While rising energy costs over a long period of time could be one trigger for such a “hardship,” changes in interest rates and expectations for weaker consumer demand, likely as a result of wars in the Middle East, are also unlikely to help.
The basic questions are familiar. Can the AI sector generate enough returns to justify its soaring valuation?
But even a small increase in energy costs could certainly cause a rethink. Given the financial wizardry at work, the impact could be felt not only in the U.S. market but also in markets overseas.
Could this be another way in which President Trump’s mindless onslaught against Iran has unleashed forces beyond his control?
