Greenlight Re: the discount to book is the return on equity

Greenlight Re lost $29.6 million in the second quarter, and underwriting had almost nothing to do with it. The combined ratio was 100.1%, which on $161.8 million of earned premium costs about $0.2 million. The rest came from the company's own investment account: Solasglas, managed by David Einhorn's DME Advisors, a related party, gave back $27.9 million in three months. That account is 70.7% of shareholders' equity. The underwriting book decides whether this reinsurer is any good. The fund decides what it reports.

At the October 8, 2026 close of $14.97 the shares trade at 0.73 times fully diluted book value of $20.61 and 0.8% below my triangulated Base value of $15.09. The Bear value is $6.78, a 54.7% drop. The required margin of safety (MOS) is 45%, so the maximum buy price is $8.30, which is 44.5% below today's price. The discount to book is genuine. It is also close to what a company earning 9% on equity should trade at when equity costs 12%.

Route: SF-03 Insurance. Subtype: property and casualty reinsurance. Economic classification: Standard (3/12). Primary method: justified price to book value, derived from sustainable return on equity (ROE), cost of equity and long-run growth. Secondary method: a bottom-up bridge to normalized annual earnings, capitalized at 9 times. Values are present fair value estimates, not 12 month price targets.

What Greenlight Re is

Greenlight Re is a Cayman-based specialty reinsurer that writes property and casualty business through three routes: an Open Market book of treaty and facultative reinsurance, an Innovations unit that takes stakes in insurance technology businesses and writes the risk they produce, and Lloyd's Syndicate 3456. It earned $661.1 million of net premiums in 2025, so this is a small reinsurer by sector standards, with a market capitalization of $489 million.

The second engine is the one that makes it unusual. Instead of the conventional bond portfolio, most investable assets sit in Solasglas, a fund managed by DME Advisors, the firm of chairman David Einhorn. Premiums collected today pay claims years from now, and in the meantime the money runs a long and short equity strategy. On June 30 that was $493.4 million against $697.7 million of equity. The result is a company where good underwriting can be buried by a bad quarter in the fund, and the other way round.

The near-term change is Lloyd's. On July 30 the company announced in-principle approval for Syndicate 3456 to convert from a special purpose arrangement into a full syndicate from January 1, 2027. The syndicate made a calendar-year profit of $3.656 million in 2025 at a combined ratio of 89.6%, against 101.1% in 2024. A full syndicate means a wider licence and more business, and also more capital tied up at Lloyd's.

The latest quarter

The $31.5 million of first half catastrophe losses came from the Middle East conflict and an explosion at a QatarEnergy gas facility, with no recoveries booked. On $316.0 million of earned premium those events are worth 9.97 points of combined ratio. Strip them out and the first half looks like an 88% underwriter, which is the number a bullish reader will reach for. It is also the wrong number: a reinsurer that never has a catastrophe year is a reinsurer that is not being paid for one.

Why book value is the metric, and why free cash flow is not

For most companies I value the cash left over for owners. For a reinsurer that measure is misleading, because premiums arrive years before claims are paid. Operating cash flow can look generous while every dollar of it is already committed to reserves and collateral. On June 30 net loss reserves were $889.0 million, or 1.27 times shareholders' equity, and restricted cash was $526.8 million against $76.3 million of unrestricted cash. The restricted cash earns income. It does not belong to shareholders until the claims behind it are settled.

So the metric is fully diluted book value per share and the return the company earns on it. Note which book value: reported basic book is $21.22 and fully diluted book is $20.61, the difference being 972,651 in-the-money options and restricted stock units. Using the basic figure to value a share you can actually buy would add 3% to every number in this article for free. The company reports both and ties its long-term incentive plan to the diluted one, which is the right choice.

Total assets are $2.24 billion against equity of $697.7 million. Financial debt is almost nothing, $9.0 million drawn on a revolving facility. The leverage that matters here is the reserve and investment leverage, not the balance sheet debt line.

Capital, ratings and the share count

Statutory capital is held at the subsidiaries, not at the holding company. At December 31, 2025, the Cayman carrier held $609.4 million of statutory capital against $354.6 million required, 171.9% coverage, and the Irish carrier reported own funds of $66.8 million against a solvency capital requirement of $41.0 million, 163% coverage. Those are the latest published figures and they are nine months old. They also do not add up to one group ratio: capital cannot move freely between regulated entities, and dividends upstreamed to the parent need approval. In November 2025 the operating companies' financial strength rating was upgraded to A with a stable outlook, while the parent's issuer rating sits at bbb.

The share count moves the right way. Fully diluted book value shares were 35.60 million at the end of 2023, 34.65 million at the end of 2025 and 33.85 million at June 30, 2026, a 2.31% reduction in six months. Buybacks ran at $9.8 million in 2025 at an average of $13.76 and $19.2 million in the first half of 2026 at an average of $17.42, with a further $3.9 million at $16.42 between July 1 and August 3. Buying at $17.42 against a $20.61 book adds to book value per share. Whether it adds to economic value depends on whether that book earns its cost of capital, which is the whole question in this article.

Equity actually fell over the half, from $708.0 million at the end of 2025 to $697.7 million, because $19.2 million of buybacks was larger than $6.2 million of earnings. Book value per share still rose. That is arithmetic working as intended, not value creation.

Primary valuation: justified price to book

For an insurer the cleanest anchor is the steady-state relationship between the return on equity, the required return and growth:

Justified price to book = (ROE - g) / (k - g), and value per share = book value per share times that ratio.

The input that decides everything is the sustainable ROE. The record: 2025 delivered about 11% on average equity, 2024 about 6%, and the five-year path of fully diluted book value per share from $13.42 at the end of 2020 to $20.43 at the end of 2025 is a compound annual rate of 8.77%, with no dividends paid and some of that growth coming from buybacks below book rather than from earnings. The first half of 2026 annualizes to under 2%. I use 9.0% as the through-cycle sustainable ROE: better than the trailing book compounding, because underwriting has genuinely improved since 2022, and below the 2025 result, because one good year in five is not a run rate.

Cost of equity is 12%. That is my required return for a small reinsurer carrying catastrophe risk and a concentrated, externally managed equity portfolio, not a measured figure from a beta calculation. Long-run growth is 2%, which implies retaining 22% of earnings. The sensitivity grid below is where the judgment is tested.

Sensitivity: value per share at 2% growth

Book value is worth exactly book value when the return equals the required return, which is the 10% and 10% cell. Everything to the right of the diagonal is a company that should trade below book, and that is where Greenlight Re sits on any honest reading of its record. No liquidation floor at book is assumed, no cash is added on top, and the investment fund is not valued a second time: it is already inside the $20.61.

Secondary valuation: what a normal year earns

The second method ignores the book formula and builds annual earnings from the ground up: premiums times underwriting margin, plus the return on each pot of invested assets, minus the costs that sit outside the combined ratio. The asset base is the June balance sheet: $493.4 million in Solasglas and a separate $775.98 million of cash and fixed maturity investments ($172.9 million of fixed maturity, $76.3 million of unrestricted cash and $526.8 million of restricted cash). Solasglas returns are quoted net of the manager's fees, so those are not deducted twice.

The premium, combined ratio, investment returns, expenses and multiples are my estimates. Two of them carry most of the weight. A 97% combined ratio is better than the five-year average of 98.7% and worse than 2025, and it assumes no help from reserve releases. An 8% net return on Solasglas is close to what the fund needs to earn to justify the structure, and it is an assumption rather than a forecast: the same account lost $27.9 million in one quarter this year.

The 9 times multiple is a judgment about a volatile small reinsurer, not a peer median, and it is worth noting that the bridge arrives at a 9.0% return on June equity entirely independently of the primary method, which uses 9.0% as its ROE input. Two different routes, same answer on the return. That is corroboration of the earning power, not of the value: the two methods still differ by 14.3% per share.

History and growth diagnostics

A proper five-year distribution of price to book is not available to me, so I use dated observations against the book value that was actually published at the time, which avoids crediting the past with a book value nobody knew yet.

A persistent discount to book has precedent here, and the Base multiple of 0.70x sits inside the observed range. Trailing twelve month net income through June is $51.0 million, which against the current diluted share count is about $1.51 a share and a trailing multiple near 9.9x. Screens show 10.1x on weighted average shares. Against my normalized earnings the market is paying 8.1x.

The growth cross-check (SM-19) does not produce a usable PEG ratio: credible synchronized forward earnings estimates with a visible sample size were not available to me, and the cyclical distortion flag is on. The 75% jump in 2025 earnings came off a weak 2024, and the first half of 2026 went the other way. A low trailing multiple next to a big trailing growth rate describes a rebound, not a growth company. The free cash flow quality check (SM-20) does not apply to an insurer: operating cash flow is premiums received before claims are paid, so the book value and earnings equivalents above take its place. Book value per share grew 8.77% a year over five years and the current normalized earnings yield is 12.4%. Reliability of forward growth: low. Neither diagnostic sets the Base, the margin of safety or the verdict.

Valuation triangulation

The primary method gets the larger weight because the equity capital it starts from is reported and audited, while the secondary depends on an assumed investment return on a fund that can move 6% in a quarter. The divergence between them is 14.3%, under the 20% level that would cost a confidence point and add to the margin of safety. The two are not independent: both rest on the same balance sheet and the same view of what this company can earn. Agreement between them is not proof.

What is the market already pricing in?

At $14.97 and a book of $20.61 the price to book is 0.726x. Running the primary formula backwards gives the return on equity the market is paying for:

At my 12% hurdle the market is paying for a 9.3% return, against my 9.0% estimate and the 9.0% the bottom-up bridge produces. Read the other way, if you believe this company sustainably earns 9%, the price implies a cost of equity of 11.6%, which is a sane required return for the risk. The price is not assuming a reserve blow-up or a book value haircut. It is assuming a decent but below-hurdle return, which is what the last five years show. That is the uncomfortable part of the bull case: the discount to book is not an error waiting to be corrected, it is the valuation of the return.

Bear, base and bull scenarios

Each scenario uses the same book anchor of $20.61. The Bear lowers the return and raises the required return, which is how a deteriorating reinsurer actually gets repriced, and it stops there: writing down the book, deducting catastrophe losses from equity and then applying a valuation haircut would count the same bad news three times. In the Bull the two methods are 20.4% apart, so treat that end of the range as the loosest number in the article. I assign no probabilities to the three cases.

One stress sits outside the scenarios. The company's own modelled single event net probable maximum loss for a 1 in 250 year North Atlantic hurricane was $143.1 million at July 1, which is $4.23 a share, or 20.5% of equity before tax effects and before any earnings that offset it. A disclosed +100 basis point parallel move in rates was worth $33 million to the fund, about $0.97 a share. These are modelled sensitivities at a past date, not forecasts, and they do not add together.

How the buy price is set

A Confidence C score carries a base MOS of 35% at the lower end of its range, which is where the verified book value and audited accounts put it. I add 5 points for cyclicality and event risk, because catastrophe losses and investment losses tend to arrive in the same year, and 5 points for the concentration of 70.7% of equity in a single externally managed strategy. Nothing is added for financial leverage, which is negligible, and nothing for method divergence, which is below the threshold. The required MOS is 45%, so the maximum buy price is $15.09 times 0.55, or $8.30. Below the Bear value of $6.78 the thesis would need rebuilding before buying, so the practical accumulation band is $6.78 to $8.30.

That ceiling is 44.5% below today's price, and it will look severe for a stock already at 0.73 times book. It is the price at which this particular combination of below-hurdle returns, event risk and concentration pays for the uncertainty. A cheap multiple is not the same thing as a margin of safety.

Dividend and capital return

No common dividend is paid and none was announced in the filings I reviewed. The modelled cash yield is 0%, and dividend coverage tests do not apply. Capital comes back through buybacks, which is not income: a shareholder who holds receives nothing and simply owns a larger share of the company. Subsidiary dividends upstreamed to the holding company are not a distribution to shareholders either, and they need regulatory approval.

Reasons to own GLRE

• Book value has compounded. Fully diluted book value per share went from $13.42 at the end of 2020 to $20.61 at June 30, 2026, through a hard market, a soft one and two loss years, with no dividend leakage.
• Underwriting has improved. The combined ratio was 94.6% in 2025 against 101.4% in 2024, and the Lloyd's syndicate ran at 89.6% on a calendar-year basis in 2025.
• Buybacks below book. $29.0 million of repurchases in 2025 and the first half of 2026, plus $3.9 million in July, at average prices between $13.76 and $17.42, all below the book value at the time.
• The Lloyd's route. Full syndicate status from January 2027 widens the licence and the distribution, if the incremental business is written at a profit.
• Investment optionality. A differentiated long and short strategy can earn more than a bond portfolio over a cycle. The owner has to accept that it works in both directions.

What could go wrong?

• Thesis breaker: reserves and underwriting deteriorate together. Net loss reserves of $889.0 million are 1.27 times equity. A 3% adverse development on that balance is $26.7 million, more than four times the first half's net income.
• Catastrophe accumulation. The first half already carried $31.5 million of event losses with no recoveries. The modelled 1 in 250 single event hurricane loss is 20.5% of equity.
• Investment concentration. 70.7% of equity sits in one related-party fund that lost $27.9 million in the second quarter. Insured losses and investment losses are not independent in a stressed market.
• Capital is not fungible. The 171.9% Cayman and 163% Irish coverage ratios are nine months old and belong to separate regulated entities. Simultaneous claims and investment losses can constrain both upstreaming and growth.
• Growth that consumes capital. A full Lloyd's syndicate needs funds at Lloyd's. Expansion written at a combined ratio above 100% destroys capital faster than it builds the franchise.
• Related-party economics. The investment manager is the chairman's firm. Fees and conflicts need watching regardless of how the fund performs; management fees were $3.57 million in the first half alone.

Management execution

Book value per share rose 13.8% in 2025 and the annual return on equity was about 11%, against about 6% in 2024. Repurchases were executed below book in both years, and long-term incentives are tied to diluted book value per share, which is the correct target for an insurer. Against that, the five-year underwriting record includes two years above a 100% combined ratio, and the fund that holds most of the invested assets is run by a related party. The reported management fee rate was reduced to 1.25% a year from 1.5%, effective May 1, 2026, which helps and does not remove the conflict. Two independent directors were added in July and September 2026. Assessment: 12/20, good with qualifications. The next test is the third quarter report, with September's fund result and an updated share count.

Stock Analyza scorecard

Business quality is 49/100: moat and pricing power 9/20 (specialty relationships and a Lloyd's platform, in a market where reinsurance pricing is set by capacity), return versus cost of capital 5/20, balance sheet 10/15 (almost no debt, but reserves at 1.27 times equity), earnings and reserve quality 8/15, growth and runway 7/15, management and capital allocation 7/10, dilution and alignment 3/5. The valuation score is 33/100: discount to Base 8/40 (a 0.8% discount against a 45% requirement), protection against the Bear case 3/20 (the price is 121% above the Bear value), agreement between methods 11/15, historical valuation 3/10 (dated observations, no distribution), reverse valuation 5/10 (the implied return is almost exactly my estimate, so there is no asymmetry to exploit), data and model quality 3/5.

For context on the underwriting score, two larger reinsurers reported Q2 2026 combined ratios of 93.1% (AXIS) and 72.8% (RenaissanceRe) against Greenlight Re's 100.1%. Different mixes, different catastrophe experience and different reserve releases, so this is not a like-for-like comparison. It is enough to explain why I do not pay a premium underwriter's multiple here.

Confidence score

Data quality is high because the balance sheet, share count and book value reconcile to the filings, and it is capped below full marks because several current inputs are missing: third quarter results, September's fund return, the October share count and any solvency ratio newer than December 2025. The range from $6.78 to $27.52 is what holds scenario dispersion at 2.

What I would watch from here

Today the readings are mixed. The rolling combined ratio is yellow at 98.1% for the half. Book value per share is green, up from $20.43 at the year end. Reserve development is green, favourable by $1.8 million. The fund is yellow: the second quarter lost $27.9 million and I have no verified September figure. Concentration is yellow at 70.7%. Share count is green, down 2.31% in six months. Data freshness is the weakest item, and it is why the third quarter report is the next review point.

The three most important thesis breakers

1. The combined ratio stays above 100% for two successive rolling years while diluted book value per share falls.
2. Reserves develop adversely by more than 3% of the opening net balance, which would reprice both the earnings and the book.
3. A loss year in Solasglas coincides with a catastrophe year, draining capital and the statutory buffers at the same time.

Adversarial review

I recomputed every figure in this article in a separate model and checked the inputs against the second quarter release, the Form 10-Q for the quarter ended June 30, 2026, the 2025 annual release and the closing price. The first draft carried three issues that changed the output.

The ROE correction is the one that matters. The Bear and Bull cases both used a sustainable ROE that matched their own earnings bridge, 4.5% against 4.3% and 13.0% against 13.9%. Only the Base was set two points above its bridge, which manufactured most of the apparent upside and most of the apparent agreement between the two methods. Fixed, the methods are 14.3% apart and the upside is 0.8%. The verdict was watchlist before the correction and it stays a hold now, but for a different reason: not that the discount is too small, but that there is essentially no discount.

The strongest counter-thesis, for the bulls: book value has compounded at 8.77% a year with no dividend, buybacks below book keep adding to it, underwriting has improved for two years running, and the Lloyd's syndicate opens a wider market in 2027. If this company sustainably earns 12% rather than 9%, the primary value is $20.61 and the full Bull case triangulates to $27.52. The response is that a 12% sustainable return has to be demonstrated across a full cycle, not asserted from one strong year, and even the Bull value at a 45% margin of safety gives a buy price of $15.14, barely above today's quote. That is an optimistic case clearing the bar by 1%, not a reason to buy.

The strongest counter-thesis, for the bears: a below-book price can still be too expensive if the capital earns less than it costs. At an 8% sustainable return the primary value is $12.37, and at a 14% required return it is $10.31. The response is that the last verified capital position and two years of underwriting improvement do not support treating this as a value trap, which is why the outcome is a hold with a conservative buy ceiling rather than a sell.

Only a 10% cost of equity or a sustainable return above 10% produces meaningful upside, and none of the nine checks brings the price inside the buy zone. Audit flags: the 2021 to 2023 combined ratio and book value series comes from the company's investor presentation and I verified only the 2024 and 2025 figures against the annual release. The modelled probable maximum loss, the interest rate sensitivity and the management fee rate change are company disclosures in the quarterly filing that I did not re-derive. Monthly Solasglas returns after June could not be retrieved from the investor relations page, so no post-June book value update is implied anywhere in this article. Arithmetic: $697.682 million divided by 33.854189 million shares is $20.6084; $14.97 divided by $20.61 is 0.7263; 0.7 times $14.427 plus 0.3 times $16.6505 is $15.0940; $15.0940 times 0.55 is $8.30. The publication gate status is PASS WITH WARNING: the balance sheet and share count reconcile, and the value rests on an estimated sustainable return, an estimated fund return and an estimated required return.

Final verdict: hold at $14.97

Greenlight Re is a better underwriter than it was in 2022 and a cheaper stock than its book value suggests at a glance. It is also a $489 million company holding $889 million of net claims reserves and $493 million in a fund run by its chairman's firm, earning a return that has not covered its cost of capital over a full cycle. The market has worked that out. At 0.73 times book it is paying for a 9.3% return on equity, and 9.3% is roughly what this business earns.

Verdict: HOLD at $14.97, watchlist for new money. Triangulated Base value $15.09, upside 0.8%, Confidence C (15/25). Required margin of safety 45%, maximum buy price $8.30.

A discount to book is only an opportunity when the book earns more than the market thinks. Here the market and I agree on what it earns, which leaves the price doing exactly the job it is supposed to do.

Sources

Primary sources include Greenlight Re's second quarter 2026 earnings release of August 4, 2026, the Form 10-Q for the quarter ended June 30, 2026 (balance sheet, loss reserves, borrowings, catastrophe losses and reserve development), the full year 2025 earnings release of March 9, 2026, the second quarter 2026 investor presentation for the multi-year combined ratio and book value series, the audited 2025 consolidated financial statements for statutory capital, the Irish subsidiary solvency report for 2025, the November 2025 rating upgrade announcement and the July 30, 2026 announcement on Syndicate 3456. The closing price and trailing multiples come from StockAnalysis. Peer combined ratios come from RenaissanceRe's second quarter 2026 release and from reporting on AXIS Capital's second quarter 2026 result.

The reference market price is the October 8, 2026 regular session close of $14.97. Balance sheet and book value figures are as of June 30, 2026, and the latest separately reported basic share count of 32,641,344 is as of August 3, 2026. Third quarter results, September investment returns and the October share count were not available. The sustainable return on equity, the cost of equity, the growth rate, the normalized premium and combined ratio, the investment return assumptions and the capitalization multiple are analytical estimates, not company guidance.