South Korea's state-run postal service is looking beyond mail for returns. With delivery losses rising, Korea Post is putting part of its $104 billion savings and insurance portfolio into overseas real estate secondaries tied to AI data centres, logistics and rental housing.
Korea Post is not treating AI infrastructure like a tech fad. It is treating it like income-producing real estate, and that distinction matters. The 142-year-old postal service is under pressure from a shrinking mail business, but it also manages a large pool of savings and insurance assets for retail customers, many of them retirees. That makes the investment strategy both cautious and urgent.
According to a Reuters interview with Korea Post president In-hwan Park published on May 22, the state-run group manages 157 trillion won, or about $104.28 billion, in savings and insurance funds. Park said Korea Post is looking at secondary funds focused on data centres, logistics facilities and multi-family housing in North America and Europe after a sharp correction in developed-market real estate. The postal business lost 311.6 billion won in 2025, and losses are forecast to reach 340 billion won in 2026.
The move is not a sudden pivot into high-risk assets. Korea Post has selected Blackstone and Madison International Realty as preferred bidders to run a $230 million overseas property secondaries fund. Park said secondary investments offer a margin of safety because buyers can enter existing assets at discounts to their underlying values. That is the attraction. Korea Post can seek higher returns without taking the full risk of building, leasing and operating new properties from scratch.
Why AI data centres fit a conservative portfolio
The important detail is that Korea Post still plans to keep around 70 percent of its portfolio in safe-haven assets such as bonds. That anchor is not going away. The postal service has a legal obligation to guarantee principal and interest on its savings and insurance products, so it cannot behave like a venture investor chasing the next cycle.
The remaining allocation is where the search for yield becomes more interesting. Private debt, mezzanine finance and real estate secondaries can help lift returns when bond yields alone are not enough to cover the pressure from mail losses and customer guarantees. In that context, AI data centres look less like a speculative bet and more like a property asset with long-term tenants, heavy infrastructure demand and lease structures that institutional investors understand.
That does not make the trade risk-free. Data centres need power, land, cooling and reliable connectivity. They are also exposed to the capital spending cycles of the largest cloud and AI companies. If hyperscalers slow their buildouts, landlords can feel it. But compared with office buildings, where hybrid work continues to cloud demand, AI infrastructure has a clearer demand story. The world keeps asking for more compute, and someone has to finance the buildings that house it.
The secondary market is doing the heavy lifting
Secondary funds buy existing stakes in private assets from investors that want liquidity. For a buyer like Korea Post, that can mean buying into operating assets rather than backing new developments. It can also mean better pricing when sellers are under pressure. That is why Park's margin of safety language matters. The discount is part of the investment case, not a footnote.
The market has grown quickly. Reuters cited Preqin data showing assets under management in real estate secondaries reached $45.1 billion as of September 2025, up from $16.1 billion in 2016. In November 2025, Singapore-based Aquilius Investment Partners raised $1.1 billion for its second Asia-Pacific real estate secondaries fund, the largest of its kind in the region. Korea Post is not alone in seeing an opportunity.
That growth creates its own problem. When more conservative capital chases the same pool of assets, discounts can narrow and expected returns can fall. AI data centres may have strong demand, but price still matters. A good asset bought too expensively can become an average investment very quickly.
What this signals for the AI economy
For startups, the story is not only about Korea Post. It is about who now owns the rails of the AI buildout. Institutional investors that once leaned heavily on office, retail and industrial property are increasingly looking at data centres as infrastructure-like assets. That changes the economics of compute. More capital can bring more capacity, but it can also keep rents firm if landlords capture the value created by scarce power and prime locations.
Korea Post's approach shows how far AI infrastructure has moved from Silicon Valley balance sheets into mainstream institutional portfolios. The next test is performance. If Blackstone and Madison can deliver stable returns from these secondaries, more Asian institutions are likely to follow. If discounts vanish or demand cools, the shift will slow. Either way, AI data centres are no longer just a technology spending story. They are becoming part of the income strategy for some of the world's most conservative pools of capital.
