Hard Market
Notes on the insurance capital markets from Charlie Evans
Over 10 years deploying capital in the insurance capital markets I've found that most coverage rarely goes beyond surface-level headlines, outcomes, and broad narratives.
This newsletter takes a different approach - a perspective from inside the market - the pricing cycles, the capital flows, what's moving markets, and what's behind the headlines.
Edge in public markets is harder and harder to find - too many teams chasing the same signals. Citadel, Point72, and plenty of others have handed billions back to investors rather than grow.
Meanwhile, insurance is an industry that takes two years to launch a product.
Just hire a bunch of quants out of the proven shops, give each team a slice of the balance sheet and its own P&L, launch products in months, and cut the teams that lose - easy, no?
What makes a pod shop work isn't the talent. It's the control system that sits around the talent. In insurance that control system runs on a lag.
What a pod shop actually is.
Millennium is a fund backing around 300 hedge funds. Each "pod" is a small team with its own strategy and its own P&L. The centre gives them capital, finances them and watches them with a few golden rules:
Every pod has to be market neutral. No directional view on the market. Strip out the market exposure and what's left is hundreds of independent bets which can then (in theory) be safely levered.¹
Every pod is marked every night. The P&L is valued by market every night at the close. No delay, no internal marks.
Every pod has a tripwire. At Millennium, a team that loses 5% of its capital has its book cut in half. At 7% they're taken off risk.² Staff turnover runs 15-20% a year.
When a pod is cut, the risk leaves with it. Positions are sold back to the market within days.
At the top firms the pods keep about a quarter of the profits they make, and investors fund that through "pass-through" expenses of 4.5-6% of assets a year, before the performance fee. Citadel's own bond prospectus showed investors keeping $30bn of $56.8bn of gains.
Why do pod shops charge these fees? Because they can. Millennium's had one down year since 1989 as the sector's grown to about $430bn.
Insurance already has the ingredients.
This pitch isn't new to insurance.
The pods exist. They're called MGAs (managing general agents). Underwriting teams that write business on someone else's balance sheet for a commission. US MGAs wrote about $128bn of premium in 2025, growing at 12% a year against 5% for the market.
The platforms exist. They're called fronting carriers. Licensed balance sheets that rent their paper to MGAs for roughly 5-8% of premium and cede most of the risk to reinsurers. They carried more than $22bn of premium last year. Accelerant, a self-described "risk exchange" for 280 MGAs, went from $3m of premium in 2018 to $4.2bn in 2025.
The performance fee exists. It's called a profit commission. Most carriers pay it to MGAs on a sliding scale tied to the loss ratio.
The whole thing exists in one building. Lloyd's of London is effectively 100+ pods run by 50+ independent managing agents, sharing a central fund and a management that approves every syndicate's business plan and can cut its risk.
Lloyd's solvency report puts the diversification benefit of running these 'pods' under one capital framework at £29bn against £57bn of undiversified risk.³
Lloyd's CEO Patrick Tiernan calls the market "one of the world's best-kept financial secrets hiding in plain sight".
So the question isn't whether insurance can be organised into pods - it has been for decades. It's why none of these structures behaves like Millennium…
Why doesn't the Millennium for insurance exist?
At Millennium a portfolio manager's P&L is marked to market every evening.
An underwriter's P&L is estimated for years before crystallising.
The premium is known the day the policy is written. But claims arrive over anything from a few months to 10+ years depending on the risk. Even after the policy period ends the "profit" is a best estimate - a number set by actuaries, built on pricing assumptions.
How late does the truth arrive? In calendar 2024 alone, US insurers added $15.8bn to casualty reserves for prior years. The accident years that moved most were 2015 to 2019 - re-pricing of losses almost a decade after the business was written.⁴
In the fronting market, where the 'MGA pods' live, initial accident-year loss ratios have developed adversely in each of the past seven years.⁵
What happens when you run the control system on an insurance 'lag'?
The tripwire can't fire. A 5% drawdown is invisible when the loss is a reserve estimate the pod helped to set. Lloyd's last soft market ran from 2013 to 2017 - while a £2bn loss was booked in 2017, the remediation began in 2018 and the repricing showed up in 2019-21 (Issue 11). That loop took 5 years. Millennium's takes a week.
The risk doesn't leave with the pod. Fire a portfolio manager and the positions are sold within days. Fire an underwriting team and you keep every policy it wrote until the last claim is paid. Markel stopped writing a group of casualty programs through its fronting carrier, State National, in 2021. In the summer of 2026 it took a $205m charge on them because the losses had "got ahead of the collateral" posted by the reinsurer behind the programs.⁶ The platform was still paying out 5 years after the 'pod' was cut.
Fees are paid on estimates. Profit commissions are paid annually on booked results (Issue 8 covered why the fund manager fee contract can't fix this). Fidelis is the cleanest public example: in 2023 it split into an underwriting partnership that earns the fee and a balance sheet that holds the risk. In 2025 the partnership earned $325m of ceding commission at an EBITDA margin above 60%, while the balance sheet earned an 8.5% return on equity and bought back its own shares at about 0.7 times book - the market valued a dollar of its equity at seventy cents.⁷
The pods aren't neutral. A multi-manager forces every pod to be market-neutral, so the platform's diversification is real. Every underwriting pod is long the same thing: the pricing cycle. A soft market takes margin off every line at once (Issue 11 covered where losses hide when prices fall), while inflation and court rulings hit every casualty book regardless of who wrote it. Lloyd's saw what happens when pods share one factor exposure in 1988-92, when its syndicates reinsured each other's excess-of-loss books until the same catastrophe losses had gone round the market several times, and the market took an £8bn loss. The diversification the platform depends on is smaller than it looks.
The last time hedge funds tried.
This has been attempted before…in reverse.
Between 2012 and 2016 Third Point, SAC, Paulson, Highbridge, BlackRock and Blackstone each put capital behind a reinsurer. The plan was the opposite of what we've articulated so far; alpha on the asset side, with the underwriting meant to do little more than break even on low-volatility business and hand the float to the hedge fund.⁸
The underwriting never broke even. Third Point Re ran a combined ratio above 100% (claims and costs exceeding premium) in each of its nine years, and AM Best eventually confirmed that it hadn't reported an underwriting profit since inception.⁹
Watford, which had Arch doing its underwriting, never made an underwriting profit in 6 full years and was taken private at about 0.74 times book.¹⁰
Third Point Re merged into SiriusPoint in 2021, and SiriusPoint only began producing combined ratios in the 80s after cutting the hedge fund allocation by about 90%.¹¹
Why wasn't breakeven available? Because a new balance sheet with no renewal book is shown the business the market doesn't want (see Issue 10 and Issue 11).
Today's pitch flips the sides: alpha on the liability side with plain assets.
The issue remains - a pod with no renewal book sees the declined risks.
Where it could work.
To be fair, the control system can work where the P&L arrives fast.
Property catastrophe tells you within a season, as do parametric and short-tail specialty lines. The trouble is that fast-feedback lines are either beta (cat has the most capital crowding in the market) or commodity (everyone already has the data).
So the pods go looking for products that are fast, new and uncrowded. In practice that means covers built around a financial variable: a residual value, a revenue shortfall, a default.
They're quant-shaped products, and the market is full of them right now.
Nvidia has been talking to insurers, through Howden Re, about cover for lenders if the chips pledged against neocloud loans can't be resold for enough.
Munich Re is guaranteeing GPU valuations behind a crypto lending protocol.
Nvidia will backstop residual value on up to a quarter of a deal, while the banks depreciate the chips over three to four years.
The history of this shape of product is pretty clear:
In the 1970s Lloyd's underwriters insured computer lessors against their customers handing back IBM mainframes early. IBM launched the cheaper 4300 series in 1979 and every insured machine became obsolete at once. Lloyd's puts the market's loss at $400m.¹²
In the late 1990s London insured banks against films failing to earn back their budgets. The bill: $1.4-1.5bn of losses and more than fifty lawsuits.¹³
In 2001 eleven insurers had written surety bonds on Enron's forward oil and gas deliveries. One bankruptcy triggered every bond at once, and they paid $654m to settle.¹⁴
It's a similar pattern: fast feedback, real novelty, and triggers that look independent until they turn out to be the same event - correlation to a single factor that an insurance platform can't diversify away.
As Nassim Taleb would tell us, that correlation will come from a factor the pod's model isn't tracking, because the model is built from the years when that correlation didn't happen.
The two questions.
Which brings me back to where every issue of this newsletter ends up.
Who's being selected against? The pod with no renewal book. The business that reaches a new underwriting team first is the business someone else declined, and the quant can't see that in the data.
Does the return survive the cycle? Every pod is paid annually on an estimate, every pod is long the same market cycle, and the platform holds the tail for years after the pod is gone.
Here's what I'd want to see from anyone building an insurance pod shop:
Tripwires that run on leading indicators rather than results: Rate adequacy against plan, terms, the mix of submissions (roughly what Lloyd's performance management is meant to do).
Dollar aligned GPs. See Issue 8: Skin in the game.
None of that is a pod shop. It's an insurance company with a strong centre, which is what the good ones have always been.
Looking ahead.
The standing invitation holds. If you've been sitting on a question - about a product, a structure, an idea - send it over.
Until then.
Charlie Evans, CAIA
Chief Investment Officer, Zephyros Investments
———
¹ BIS Quarterly Review, "Hedge fund exposure to the carry trade", 16 September 2024 - multi-strategy funds' leverage "was 4:1 by traditional measures and 14:1 when accounting for synthetic leverage via derivatives", from SEC Private Fund Statistics. https://www.bis.org/publ/qtrpdf/r_qt2409x.htm
² Bloomberg, "Millennium Nears $100 Billion in New Era for Giant Hedge Funds", 7 September 2026 - "Lose 5% and your book gets cut in half; lose 7%, and you're probably looking for a new job"; one losing year since inception (2008). https://finance.yahoo.com/markets/stocks/articles/millennium-nears-100-billion-era-210034426.html
³ Lloyd's Solvency and Financial Condition Report 2025 - market-wide SCR of £27.6bn after a diversification benefit of £29.4bn on undiversified components of about £56.8bn (p.28). https://assets.lloyds.com/media-651c0e64-c1d0-4f97-90f7-883c69fe2ef2/dae999dc-d31c-45f8-b0d0-1fe85845afcc/Lloyd's%20of%20London%20-%20SFCR%202025%20.pdf
⁴ Milliman, "US casualty insurance 2024 financial results", 16 September 2025 - $15.8bn of adverse prior-year development on casualty lines in calendar 2024; the largest percentage increases on accident years 2015-2019. https://www.milliman.com/en/insight/us-casualty-insurance-2024-financial-results
⁵ Conning, US fronting market report, 27 August 2026 - "Initial gross accident-year loss ratios have developed adversely in each of the past seven accident years." https://www.conning.com/about-us/news/ir-pr---fronting-market-2026
⁶ Coverager, "Markel records a $205 million loss at State National", 31 July 2026, and Markel Group Form 10-Q for the quarter ended 30 June 2026 - habitational casualty programs written 2012-2021. https://coverager.com/markel-records-a-205-million-loss-at-state-national/ · https://www.sec.gov/Archives/edgar/data/0001096343/000109634326000064/mkl-20260630.htm
⁷ Fidelis Insurance Holdings, full-year 2025 results, 25 February 2026 - ceding commissions to The Fidelis Partnership $325.0m; operating return on average equity 8.5%; 15.2m shares repurchased for $261.4m (average $17.22) against diluted book value per share of $24.61.; Insurance Business, 28 September 2026 - The Fidelis Partnership's adjusted EBITDA margin of 60% in the first half of 2026, down from 64%. https://www.barchart.com/story/news/425085/fidelis-insurance-group-reports-outstanding-fourth-quarter-2025 · https://www.insurancebusinessmag.com/uk/news/breaking-news/fidelis-partnerships-ipo-filing-shows-how-much-still-rides-on-pelagos-591364.aspx
⁸ Third Point Reinsurance Form 10-K for 2015 - the float-generating strategy and 2015 premium mix; Business Insurance, 28 October 2012 - "The hedge-fund backers of the new companies are focused on the investment strategy, rather than the underwriting strategy." https://www.sec.gov/Archives/edgar/data/0001576018/000157601816000075/tpre-20151231x10k.htm · https://www.businessinsurance.com/hedge-fund-backed-reinsurers-manage-both-sides-of-business/
⁹ Third Point Re combined ratios 2012-2020 from company results (129.7%, 107.5%, 102.2%, 104.7%, 108.5%, 107.7%, 106.8%, 103.2%, 110.3%); AM Best via Bermuda:Re, 17 May 2019 ("has not reported an underwriting profit since inception"). https://www.bermudareinsurancemagazine.com/news/third-point-re-ratings-outlook-revised-to-negative-4752 · https://s27.q4cdn.com/660241321/files/doc_financials/2018/Exhibit-99.1-Earnings-Press-Release-(12_31_18).pdf · https://s27.q4cdn.com/660241321/files/doc_financials/2020/q4/Exhibit-99.1-Earnings-Press-Release-(12_31_20)-FINAL.pdf
¹⁰ Watford Holdings results 2016-2020 (combined ratios 101.7%, 112.6%, 104.5%, 109.7%, 106.1%; book value per diluted share $47.08 at 31 December 2020) and the $35.00 per share take-private by Arch, Warburg Pincus and Kelso, 2 November 2020. https://www.sec.gov/Archives/edgar/data/1601669/000160166921000005/q42020earningsrelease.htm · https://www.businesswire.com/news/home/20201102005577/en/
¹¹ Reinsurance News, "Third Point Re, Sirius Group to combine in $788mn transaction", 7 August 2020; SiriusPoint Form 10-K for 2023 (Third Point Enhanced Fund exposure $878m at end-2021, $100m at end-2022) and full-year results for 2023-2025 (combined ratios 84.5%, 88.3%, 88.3%). https://www.reinsurancene.ws/third-point-re-sirius-group-to-combine-in-788mn-transaction/ · https://www.sec.gov/Archives/edgar/data/1576018/000157601824000018/spnt-20231231.htm · https://www.sec.gov/Archives/edgar/data/1576018/000157601826000019/exhibit991-earningspressre.htm
¹² Lloyd's, "More interesting claims" - "1980 - Computer leasing insurance ... The market lost $400m."; Itel Corporation company history - Lloyd's insured Itel from 1975 to 1978 against lessees returning computers early; IBM's 4300 series. https://www.lloyds.com/about-lloyds/history/catastrophes-and-claims/more-interesting-claims · https://company-histories.com/Itel-Corporation-Company-History.html
¹³ Business Insurance, "Burned at the box office", 4 June 2000 - industry losses under gap financing policies estimated at $1.4-1.5bn; Business Insurance, "HIH to appeal ruling in film finance case", 2 April 2006 - "at least 50 pieces of litigation in five jurisdictions" (Eversheds estimate). https://www.businessinsurance.com/burned-at-the-box-office/ · https://www.businessinsurance.com/hih-to-appeal-ruling-in-film-finance-case/
¹⁴ Business Insurance, "Bank, insurers settle Enron surety dispute", 5 January 2003 - $1.09bn claimed by JPMorgan Chase, settled for $654m by eleven insurers. https://www.businessinsurance.com/bank-insurers-settle-enron-surety-dispute/