Even the worst Middle Eastern conflict in decades isn’t slowing things down. This week, Microsoft promised to invest $5.5 billion in cloud and AI infrastructure in Singapore, and an additional $1 billion into Thailand over the next few years.

But the Iran war may ultimately force Asia to revisit its AI playbook, following a surge in energy prices and shortages of the key inputs needed to build AI infrastructure.

“The scaling laws that have driven the AI boom are fundamentally peacetime constructs, which were discovered in an era of abundant energy and expanding chip supply, and operate on an implicit assumption: that energy elasticity is unbounded,” Wei Lu, a professor at the College of Computing and Data Science at Singapore’s Nanyang Technological University (NTU), explains. That’s led to what he deems a “brute force aesthetic,” where larger and more capable models are developed even as the energy per unit of compute keeps rising.

That’s tolerable when times are good; it’s less so when supplies are constrained. “The current conflict is repricing that bet,” Lu says.

Asia’s AI boom

Asia has become the center of the world AI boom, with Nomura estimating that the region contributed nearly two-thirds of global AI trade growth in the first half of 2025. 

Different regions have specialized in different parts of the AI trade. East Asian economies like South Korea and Taiwan have won big owing to their semiconductor manufacturing, supplying the AI capital expenditure boom in markets like the U.S. In Southeast Asia, investment has focused more on assembly, precision manufacturing, and data storage. 

But with oil, LNG, and helium prices surging in the wake of the Iran war, experts warn the region’s AI operations could grow more costly. 

“The main impact on Asia’s AI boom would be higher costs for AI infrastructure development,” says Bo An, a computer science professor at NTU. “Chipmakers may face higher energy, raw material, shipping, and insurance costs. Data center operators could face higher power and cooling costs.”

He also predicts that higher costs and supply disruptions in Asia will inevitably spill over to tech firms elsewhere, given the region’s central role in the global chip supply chain. 

TSMC, for example, is the lead supplier of advanced chips to giants like Nvidia and Apple. Yet TSMC’s base of Taiwan relies on imported energy for much of its power supply, potentially setting up a difficult choice for the island’s government if the Iran crisis continues. Oxford Economics estimates that Taiwan’s industrial production might fall by 0.7% below the baseline if shortages persist for six months. 

“We are already seeing panic procurement and logistics paralysis,” says Lu of NTU, noting that the global supply chain is now “a series of single points of failure.”

Efficiency-first design

In the short term, the AI trade is strong enough to overcome worries over the Iran conflict. South Korea’s chip exports hit a record high of $32.8 billion in March, jumping more than 150% year on year, according to government data released on April 1. 

“We do not expect the energy shock to materially derail South Korea’s AI‑led growth trajectory this year, particularly as the current [semiconductor] cycle appears stronger than previously anticipated,” noted Bank of America’s analysts in an April 2 research note. 

There may even be an upside for Asia in the long term. Iran has attacked data centers in the Middle East, highlighting how server racks are now possible military targets. 

After investing heavily in the Middle East, “AI companies are starting to look at Southeast Asia and India,” Sandeep Sethi, who oversees the APAC data center business for real estate company JLL, tells Fortune. 

But when it comes to East Asia, data center operators may face the longer-term challenge of limited power availability, especially in places like Japan, where it can take up to 10 years to connect a new data center to the grid. 

Lu argues that AI businesses need to start pursuing “efficiency-first” design, reducing the energy and raw materials needed to foster artificial intelligence. 

“The most valuable form of intelligence is the kind that knows how to do more with less,” he says.