MetLife's Record Quarter Comes With an AI Footnote

A 22% EPS jump was amplified by private equity, but the new disclosure that matters is a $4.1 billion AI-infrastructure exposure—small, investment-grade, and suddenly quantified.

MET · Earnings Call · 2026-11-05

MetLife's Record Quarter Comes With an AI Footnote

MetLife's third quarter was, by the headline, a blowout: adjusted EPS of $2.85 excluding notable items, up 22%; adjusted earnings of $1.8 billion, up 17%; adjusted ROE of 19.6%; and a direct expense ratio of 11.9%, below the 12.1% full-year target. The stock has noticed. It is up 14.4% over the last 90 trading days and sits only 3.7% below its October 22 peak of $103.57. But record quarters in insurance are rarely clean, and the most interesting thing MetLife disclosed was not the EPS number. It was a footnote about AI infrastructure that management chose to put on page 10.

The Beat Is Real, but the Mix Is the Story

The quarter leaned on variable investment income: $720 million pretax versus an implied quarterly guide of $400 million, with venture capital returns above 8% and private equity returns above 4%. Management knew exactly how that would read. “While VII amplified the quarter, it did not define it.” — Michel Khalaf, President and Chief Executive Officer · 2026-11-05 That sentence is the whole debate. The underlying insurance engine was genuinely strong—group life mortality of 79.5%, a non-medical health ratio of 72.3%, RIS total spread of 130 basis points with core spread at 99—but VII was nearly twice the normal run rate, and the stock is near a high. The more durable pitch is capital efficiency. John McCallion said the company “deployed $3.8 billion of capital and generated $2.8 billion of value of new business” — John McCallion, Chief Financial Officer · 2026-11-05, with a 19% internal rate of return and a five-year payback. That is a different story from “we beat on private equity.” It is also why the recurring expense ratio discipline matters: AI and technology productivity are helping hold the ratio down even as MetLife absorbs PineBridge, a structurally higher-expense business. Michel Khalaf called AI “a force multiplier for us” — Michel Khalaf, President and Chief Executive Officer · 2026-11-05. The market may treat that as boilerplate, but the expense math is real.

The AI Disclosure Is the Actual News

For the first time, MetLife put numbers around its AI-infrastructure exposure. Hyperscaler exposure totals $1.6 billion, mostly public fixed income. Data-center exposure totals $2.5 billion, split between private fixed income, public fixed income, and a small real-estate slice. Combined, that is $4.1 billion, less than 1% of the general account. The company says over 95% of the fixed income is investment grade, with no delinquencies or impairments.

Our hyperscaler exposure totals $1.6 billion. Most of it, $1.4 billion, is public fixed income and the other $200 million is private fixed income. The entire hyperscaler exposure is investment grade with an average rating of Aa3 and a weighted average life of approximately 17 years.
John McCallion, Chief Financial Officer · 2026-11-05

This matters because the global tape is already voting on AI credit. AI data centers are down roughly 30% over the last 30 days in the Know Trend theme tape, with dozens of negative tickers across semiconductors, power, and data-center infrastructure. MetLife's new hyperscaler exposure is a company-specific disclosure, not sector boilerplate. It sits mostly in public and private fixed incomes, which is exactly where an insurance company's credit risk would live. A quarter ago, John McCallion was only qualitative on the topic: “I think in the industry, we're probably not that exposed just given our approach to kind of ingress investment-grade oriented investing.” — John McCallion, Chief Financial Officer · 2026-02-05 Now the company has quantified it. The risk did not change; the disclosure did. That is usually the right kind of change.

What MetLife Stopped Talking About

The missing catalyst is pension risk transfer. In the prior quarter, Ramy Tadros said “we are seeing a stronger pipeline in the second half of the year. And so we see more opportunities emerging for Q3 and Q4.” — Ramy Tadros, President, U.S. Business · 2026-08-06 On this call, the tone shifted: U.S. PRT activity has been lighter, and management framed it as “timing, not a structural change.” “We do have a back-ended pipeline this year. We've got a few more deals coming in the quarter.” — Ramy Tadros, President, U.S. Business · 2026-11-05 That is not a collapse—RIS still grew adjusted earnings 12% on U.K. longevity reinsurance and retained liability growth—but the PRT market was a major swing factor in prior calls, and it is now a wait-and-see item. The stock is near a high anyway, which tells you investors are giving credit for breadth elsewhere. That breadth claim deserves scrutiny. Management repeated that “no single business carried the quarter,” and technically that is true: Group Benefits, RIS, Asia, Latin America, and EMEA all contributed. But the quarter's biggest single lever was VII, and the prior call's private-equity message was more cautious. Wilma Burdis asked about trimming PE, and McCallion said “the direction of travel is a modest decline over time on PE” — John McCallion, Chief Financial Officer · 2026-08-06. A quarter later, venture capital produced an 8%-plus return and helped drive a record EPS. That is not a contradiction—PE is a long-duration portfolio—but it is a reminder that the earnings stream is not as smooth as the “quality compounder” label implies.

Valuation and the Quality Test

The latest filed quarter, before this call, showed total revenue of $19.2 billion, up 10% year over year. Operating margin was 8.4%, and the stock trades at a price-to-net-income multiple of 14.7x, up 18% year over year. On the company's preferred adjusted lens, that is not demanding. On GAAP earnings, which include the mark-to-market noise, it is less clean. The balance sheet still supports the story: $3.4 billion of holding-company cash and liquid assets, roughly $3.5 billion returned to shareholders year-to-date, and a Japan ESR of 192%, just above the top end of the previously discussed range. The AI footnote helps here. If an investor was worried that MetLife's record quarter came with hidden data-center credit risk, the company just showed that the number is small and high-grade. If an investor was hoping MetLife was an AI infrastructure play, the same disclosure says it is not. MetLife is an AI user and a modest AI lender, not an AI operator. That is probably the right exposure for a life insurer—but it means the stock's next leg depends less on AI enthusiasm and more on whether VII normalizes gently, PRT closes, and the expense ratio keeps beating. The quarter was strong enough to justify the near-high stock price. The AI disclosure was smart enough to remove a tail risk. The unresolved question is the one management cannot fully control: how much of the 22% EPS growth was the insurance engine, and how much was a private-equity mark that flattered the print.