Arch Capital (ACGL), a $34B specialty insurer I’ve been researching. Here’s my analysis.

been working through Arch Capital for the past few weeks and wanted to get my notes down somewhere. posting here because I'd genuinely like to hear what I'm getting wrong.

short version of why I bothered. it's a specialty property and casualty insurer, plus reinsurance, plus mortgage insurance. roughly $34B market cap. trades around $97 today. tangible book is $61.71 a share, so 1.56x book. they've grown book value per share at over 15% a year for 23 years. that's the kind of compounding that usually doesn't sit at 1.5x book for very long.

the way insurers actually make money is they collect premiums today, pay claims later, and get to invest the gap in between. that gap is called float. Arch sits on $24B of it. their five-year average combined ratio is 88%, which means after they pay all the claims and expenses, they keep 12 cents on every premium dollar. so the float isn't just free, they're being paid to hold it. that float backs a $47B investment portfolio that threw off $1.62B of investment income last year on top of the underwriting profit.

here is what the picture looks like, just the lines that matter to me:

what made me actually pay attention. property catastrophe pricing softened 10-20% in Q4 2025. most insurers respond to soft pricing by writing more business to keep the top line up. Arch shrank net premiums written by 4% instead. management has been clear they would rather write less at the right price than more at the wrong one. you only see that behavior at insurers that compound long-term, and it's also the behavior the market punishes in the short run.

at the same time the board authorized another $3B in buybacks on top of what they were already doing. so a 19.5% ROE business is buying its own stock at roughly 8.4x earnings while shrinking premium volume on purpose. those are two pretty bullish signals doing the opposite of what a deteriorating insurer would do.

what I'm not as sure about. regulatory capital requirements are the real risk in my view. if global regulators decide insurers need to hold more capital per dollar of premium, ROE comes down by math, not by anything management can offset. I think this is the actual reason the multiple is suppressed, and it's not nothing.

catastrophes are the other one. management's own model says a 1-in-250-year peak zone event is about $1.9B, or 8.2% of tangible equity. survivable, but those events have been showing up more often than the modeling implied for a while now.

I'm less worried about insurtech. specialty casualty reinsurance treaties don't get bound through a phone app, and that's where Arch plays.

where I land. you're paying 1.5x book for a 20% ROE compounder with disciplined management actively buying back stock at single-digit earnings multiples. the math suggests fair value somewhere around $120-125, current price is $97, and book should keep compounding at mid-teens while you wait.

I could be wrong about how regulatory capital plays out, and one genuinely bad cat year could eat 12-18 months of earnings. but I don't see anything that says the underlying business is structurally broken.

happy to be argued with on any of this. holding a position.