More than 1,000 provided heat-resilience solutions spanning cooling, sensors, worker protection, analytics and heat-reducing materials. “A large amount of the adaptation resilience that needs to take place is already starting to happen inside companies, but much of it never appears under a climate label,” she said.

The challenge for investors may be less about whether those opportunities exist than how to access them. One investor described the bottleneck as translating a fragmented real-economy opportunity into something investors can actually own.

At the same time, the economic case for adaptation at individual assets may be highly concentrated. MSCI research examined roughly 435,000 corporate assets with higher exposure and lower preparedness and found that just 2.2%, or 9,612 assets, had a positive modeled savings-on-investment case under present-day conditions. The top 1% accounted for 99% of the modeled benefit. For investors, that means finding both the companies supplying resilience and the individual assets where resilience spending may generate the greatest economic benefit.

 

5. AI isn’t a single trade and its digital expectations depend on the physical world

AI exposure reaches far beyond technology stocks. MSCI classifies 37% of the world’s listed companies as part of the AI value chain, spanning physical infrastructure, digital infrastructure and applications. A hypothetical MSCI scenario involving a relatively modest repricing of hyperscaler debt produced an approximately 11% modeled loss across a multi-asset portfolio, particularly through public fixed income, private credit and equities. “All of the expectations of the buildout of AI are baked into markets and exposures today,” said Laura Nishikawa, MSCI’s head of emerging risks research and development. Those expectations, she added, are now “colliding with the physical world.”

The buildout needed to meet those expectations is running into physical constraints. Our research finds that median energy capacity of newly built U.S. data centers rose from 11 megawatts (MW) in 2016 to 130 MW by June 2026, while average waits for power connections in data-center regions exceed five years. Seventy percent of global data-center capacity under construction is in the U.S., and more than half (52%) of U.S. developments are in counties where previous projects have been withdrawn or data-center-specific rules enacted.

Those constraints can turn water availability, grid capacity, physical hazards, insurance gaps, permitting and community opposition into investment questions. And because data centers are long-lived assets, investors need to understand how these conditions may evolve. “There is no one-size-fits-all answer,” noted Jeremy Porter. “The analysis needs to reflect the asset and the vulnerabilities of the particular market in which it sits.”

Investors are also looking beyond the hyperscalers to the second-order effects of the buildout. One described examining power-intensive manufacturers to identify where higher electricity prices could put greater pressure on earnings — and where hedging or providing flexibility to the grid could differentiate companies. AI may be a digital technology, but much of its investment story is increasingly physical.