What the next Miami hurricane will encounter

What it is

Last Friday marked a hundred years since the Great Miami Hurricane came ashore. Days before the centenary, the Swiss Re Institute published the number the insurance industry has spent this softening cycle trying not to think about: a Category 5 hurricane striking Miami or Tampa Bay today could drive more than $300 billion of insured losses across the global market. A repeat of the 1926 storm itself, a Category 4, lands at more than $200 billion. A repeat of 1992's Andrew comes in at close to $100 billion. Swiss Re's trend extrapolation points to roughly $148 billion of insured catastrophe losses in 2026 even without a major Florida hurricane, which means the extreme scenario would take the year beyond $450 billion.

The timing was pointed in the way these things always are. On Monday the National Hurricane Center was issuing advisories on Tropical Storm Fay, the first near-hurricane of the season, spinning harmlessly 385 miles southwest of the Azores. NOAA is maintaining its below-normal season forecast, and a strengthening El Niño should suppress activity through the season's remaining weeks. The storm is not the story. The story is what Swiss Re's Balz Grollimund said alongside the numbers: the question is not how powerful the next major hurricane will be. It is what it will encounter when it reaches shore.

That sentence is aimed at the industry's most quietly consequential function. Not underwriting. Not capital. Exposure management.

Why it matters now

This lands at the exact moment the London Market is putting numbers on 2027. The final tranche of syndicate business forecasts goes in on Monday. The capacity auction forms are due on Wednesday 30 September. S&P says rate reductions will continue into 2027. Moody's expects further softening at 1 January. And in May, the PRA ran its first dynamic stress test, DyGIST, in which a Gulf of Mexico hurricane arrived unannounced as one of five live scenarios – the same tail risk Swiss Re has just quantified, simulated against the London market's balance sheets. The regulator has already run this hurricane. The market is now pricing it.

The discipline debate has spent a month arguing about whether rates are adequate. The $300 billion scenario gives that debate a denominator. But it also changes the question firms should be asking themselves. The market's problem is no longer whether the tail is priced. It is whether any firm in the market can prove what it holds when the question is asked.

What firms should do

Know your answer time

The test is brutally simple. If the PRA, or the Corporation, or your own board asks on a Thursday afternoon for your Miami accumulation, when does the answer land? In a connected firm it is a query. In a fragmented one it is a project. The gap between those two answers is now the difference between defending a 2027 plan and apologising for one.

Treat exposure data as a designed asset, not a by-product

The exposure view has to travel portfolio to policy to location, and it has to carry its provenance: which model version, which rollup, which reconciliation, last touched when. That is an architecture decision. It cannot be improvised in the two weeks before the coming-into-line conversation.

Govern the models like the controls they are

Catastrophe models are assumptions dressed as software. Version them, validate them, document what changed and why, and make the link between model output and reported exposure traceable in both directions. The regulator's question will not be "do you use a model". It will be "show me".

Drill the response, not the report

The May stress test separated the market. Some firms cancelled leave and flew in experts. Others found the live element genuinely useful. The difference was not capital. It was whether the organisation had rehearsed responding, with the people, data and systems named in advance. Crisis response is an operating model. It either exists by design or it exists by heroism, and the second one does not scale.

Evidence the plan you are about to file

Every syndicate's 2027 forecast now has to trace its premium and loss assumptions to the processes that produced them. The same obligation applies to peak exposure. A plan that cannot show its accumulation work is asking the market to price a spreadsheet.

The opmodal perspective

Exposure architecture is an operating model artefact before it is a data problem. It is a chain of ownership: who holds the exposure data, who rolls it up, who reconciles it to the model output, who signs it, and how each of those steps is evidenced. When that chain is mapped and connected, the $300 billion question becomes an export. When it is not, the question becomes a fire drill – and the drill is exactly what May's cancelled leave looked like.

This is the quiet season's real lesson. A quiet season is not a holiday from the tail risk. It is the time when the tail risk gets re-quantified, the capital gets re-priced, and the regulator gets to test whether the answers exist. The firms that build the exposure chain now will walk into the coming-into-line conversations with evidence. The firms that wait for the storm to build the chain will be building it in the dark, in a hurry, while everyone else is trying to buy capacity.

The next Miami hurricane will encounter whatever we have built between now and then. Right now, for most of the market, the honest answer to that sentence is: it depends who is in the office.