Life insurers step up AI infrastructure investment: Exposure is manageable for now, but growth is changing the risk profile
Barclays estimates that digital infrastructure accounts for about 1.4% of the sampled life insurers’ bond portfolios, but purchases accelerated in the first half of 2026. Higher yields and new financing supply could drive allocations further upward, while bringing concentration, refinancing and collateral-value risks.
Original source：BarclaysFirst published on Unhidea：Original source date：
Key takeaway
Barclays believes life insurers’ AI-related investments currently account for a small share of their portfolios and have strong credit quality, but rapid expansion and changes in financing structures will make risks more concentrated and complex.
How large is life insurers’ AI exposure?
Using insurers’ statutory filings, Barclays estimates that digital infrastructure accounts for about 1.4% of the sampled companies’ bond portfolios. As of the second quarter of 2026, related holdings amounted to about 24% of statutory surplus; after an adjustment estimated using funds-withheld reinsurance assets, this falls to about 15%.
Broadly defined digital infrastructure includes financing for data center hardware, securitized financing for data center properties, and debt issued by companies such as cloud computing giants, chipmakers and data center operators. Of the exposure covered by the analysis, 99% is investment grade and 68% is NAIC Class one, roughly equivalent to A- or above.
In funds-withheld reinsurance, assets may remain on the ceding company’s statutory financial statements while their economic performance is transferred to the reinsurer. Barclays made a rough adjustment based on the proportion of bonds represented by funds-withheld assets; meanwhile, investments through channels such as commercial mortgages and limited partnerships are not fully captured, potentially understating total exposure.
Which companies have more concentrated exposure, and why might allocations keep growing?
After adjusting for reinsurance, FG, CRBG and Athene’s digital infrastructure exposures amounted to 39%, 33% and 28% of statutory surplus, respectively, above the peer average of about 15%. Purchases in the first and second quarters of 2026 have already pushed holdings noticeably higher.
Yield is one attraction: as of the end of June 2026, A-rated data center investments held by US life insurers had a weighted average book yield of 5.5%, compared with 4.6% for all their A-rated debt investments over the same period.
Barclays estimates that, as of the end of June, life insurers held about 15% of the outstanding public and 144A data center credit market. Even if their share of holdings remains unchanged, expanding new financing supply over the next few years will increase the size of their allocations. Moody’s estimates that industry holdings could rise by 50 billion to 100 billion dollars by the end of 2027 from 85 billion dollars in the second quarter of 2026; Barclays considers this estimate reasonable.
How do the risks behind high ratings differ?
In the data center asset-backed securities transactions reviewed by Barclays, loan-to-value ratios were typically 40% to 60%. Securitization structures also allocate cash flow and value to senior creditors first, helping create highly rated bonds.
Different financing instruments face risks at different times: data center corporate bonds typically fund construction, with risk more concentrated during the construction phase; asset-backed securities and commercial mortgage-backed securities typically finance mature assets after leases take effect, but have higher initial leverage and generally do not amortize principal, leaving greater refinancing and re-leasing risk at maturity.
Some AI transactions are held almost entirely by US insurers. As allocations increase, concentration in individual transactions and the industry could weaken portfolio diversification and increase the risk that asset values move in tandem.
How will risks change in the next phase?
Nvidia and Broadcom have said that together they expect nearly 1 trillion dollars in chip-related financing to enter the market over the next few years. Barclays expects insurers to participate substantially; but rapid technological change and an uncertain chip depreciation path make the residual value of financing collateral uncertain.
Spreads on investment-grade AI ecosystem corporate bonds tracked by Barclays widened by about 20 basis points in the two weeks before the report. The bank believes life insurers’ capital and liquidity buffers help them withstand market-value fluctuations, but new supply may continue to disrupt spreads, leaving existing holdings priced lower relative to new issuance.
Barclays believes that if the credit standing of large cloud computing companies deteriorates significantly, the more important impact on life insurers may come from broad fixed-income stress accompanying an economic downturn, rather than losses caused by direct digital infrastructure exposure alone.
