The artificial intelligence boom is emerging as the biggest potential risk to financial markets, even as large institutional investors continue to increase exposure to the technology, senior executives from Temasek, Mubadala and Granite Asia said.

Rohit Sipahimalani, Chief Investment Officer at Singapore state investor Temasek, said the market has remained resilient despite a series of shocks in recent years, largely because of strong earnings momentum among major AI companies.

“If you’re asking me what can go wrong, I mean, I think unwinding of the AI trade is the biggest risk,” Sipahimalani said during a panel discussion at the Milken Institute Asia Summit 2026. “We don’t see that as imminent. But will you have bumps in 2027? Possibly yes.”He said an unwinding could be triggered by AI safety concerns, tighter regulation or weaker-than-expected returns as companies and consumers begin to assess whether the technology is generating sufficient return on investment.

Sipahimalani also flagged inflation and interest rates as risks, particularly if higher energy prices, supply-chain disruptions or geopolitical conflicts keep inflation elevated. But he said the AI narrative remains strong and is supporting equity markets even as a large portion of the broader market trades below recent highs.

Temasek is seeking to remain exposed to AI while limiting concentration risk. Its portfolio spans data-centre energy infrastructure, chips, cloud platforms, models and applications, with different levels of risk across the value chain.

“We construct our portfolio to make sure that looking at risk and reward, we are balancing the size in each segment,” Sipahimalani said. About half of Temasek’s AI exposure is currently public, with the firm seeking to increase that to about 70-75% to retain flexibility to pivot as the technology develops.

Mubadala is taking a similar approach. Carlos Obeid, Chief Financial Officer of the Abu Dhabi-based investor, said AI valuations are based on expectations of rapid earnings growth, making disciplined deployment important.

“Valuations are what they are. They’re built on an estimate of how these earnings are going to grow. Things may change, so you need to be careful in how you deploy that capital, and not just focus on one segment of the value chain, but across all the elements of the value chain, whether it’s data centres, whether it’s applications, whether it’s the language models, and so on.”Mubadala has also sought to broaden its capacity to invest in the sector through partnerships. Abay said there was not enough capital available at Mubadala alone to invest across the opportunity, which was one reason the investor established MGX to expand the pool of capital available for AI investments.

Granite Asia, meanwhile, is focused on the business-model changes being driven by AI. Jenny Lee, senior managing partner at the venture investor, said startups are emerging across AI safety, testing and infrastructure, while established businesses are using the technology to dramatically reduce development times and costs.

Lee said Asia and China could also benefit from the next phase of AI development by leveraging their manufacturing bases, physical-world data and consumer markets, rather than competing directly with US frontier-model companies.