Three things I took away from Reinsurance 2026 in Zurich
I was in Zurich last week for the Reinsurance 2026 conference. A few themes came through strongly across the sessions, and they’re worth sharing because they connect directly to questions I hear from clients.
The capital gap is a pricing gap
There was a lot of discussion about coverage gaps, and one thread ran consistently through the answers: there is enough capital in the market. The problem is that customers are not willing to pay what reinsurers calculate is needed to cover certain risks. So capital exists, but it is not being deployed, and gaps emerge as a result.
Several speakers reinforced this from different angles. One CEO made the point that only around a third of a reinsurer’s capital is actually available for property and casualty events like hurricanes; the rest is committed to life, investments and other obligations. Another was blunt about the current softening market: if you are growing into lower prices and higher risk, you should be asking yourself how comfortable you really are with that trade.
The message was clear. Capital is not the constraint. Pricing discipline is.
Diversification is not what it used to be
A second theme was whether the traditional reinsurance model still works for today’s risks. The classic approach depends on diversification: spreading exposure across geographies and perils that behave independently. But as one panellist pointed out, global consolidation and interconnected risks are eroding that independence. When shocks are correlated, diversification offers less protection than the models assume.
The practical implication raised in the session was the need for more scenario modelling alongside probabilistic approaches. If diversification is weakening, reinsurers need better tools to detect and stress-test for correlated shocks before they materialise, not after.
Is AI the new asbestos?
The sharpest question of the conference asked whether AI poses a systemic risk to insurance comparable to asbestos: a latent, compounding exposure that the industry does not yet fully understand.
The responses were nuanced. On one side, AI used internally by insurers and reinsurers is already delivering real value, particularly in underwriting, where richer data sources such as real-time monitoring of insured assets are giving underwriters a level of detail they have never had before. On the other side, AI embedded in clients’ operations and products introduces risks that are difficult to quantify, hard to isolate, and potentially correlated across the portfolio. One speaker drew the parallel directly: the hidden accumulation of AI-related exposure could behave much like asbestos did, growing quietly until the claims arrive.
Others were more cautious, simply noting that it is too early to know. Which, given the history of asbestos, is not entirely reassuring.
Worth watching
These are not abstract questions. Pricing discipline, correlated risk and latent AI exposure all come back to the same thing: whether the data and models underpinning decisions are good enough for the world as it is now, not the world as it was when they were built.
