What Happens to Marin When San Francisco Gets Liquid

The conversation happens at the edge of the room, usually late, usually after someone's folded.

We play poker in San Francisco most months — a rotating table of founders, early employees, a couple of investors, people who have known each other since a previous company. And for about the last year, a particular exchange keeps recurring. Someone mentions a tender offer, or a filing, obliquely, the way people talk about money they haven't received yet. And then someone else, usually the one with a kid under two, asks the question that has nothing to do with poker:

Where are you guys going to go?

Not whether. Where.

We hear it on Saturday runs too — the groups that leave from the Embarcadero and the Presidio, where the conversation stretches out over six or seven miles and gets more honest around mile four. Same question, different setting. A senior engineer, four to eight years in, a one-bedroom that has started to feel like a decision deferred, and equity that is about to stop being theoretical.

More of them than we expected say north.

What follows is our read, built from conversations we've actually had and public data anyone can check. It's a call, not a forecast — and we'd rather state it plainly than hedge it into mush.

Why we think this cycle points north

The Google and Facebook booms landed on the Peninsula for an unremarkable reason: those companies were headquartered in the South Bay, and people live near where they work.

OpenAI and Anthropic are in San Francisco proper. That's the whole structural argument. Their workforces are city-based, and skew hard toward the years when people start having children. For a household already living in San Francisco, the natural move when the second bedroom becomes non-negotiable runs north across the bridge, not forty miles south. Thirty minutes on a ferry, a childhood spent on Mount Tam, and schools that don't require a private-school budget to be good.

The numbers underneath it are real. In October 2025, more than 600 current and former OpenAI employees sold roughly $6.6 billion in a tender offer — about 75 hit the $30 million per-person cap, and the remaining ~525 averaged around $8.3 million each. In June 2026, OpenAI filed a confidential draft S-1. Anthropic raised at a $350 billion valuation in February 2026 and is reportedly targeting a Q4 listing.

That's several hundred San Francisco households, in the family-forming window, with down-payment money that is no longer hypothetical. They do not need to be a majority of anything to matter here. Southern Marin turns over a few hundred single-family homes a year across all eight towns. A hundred additional serious buyers is not a rounding error in a market that size.

Why the supply side is the stronger half of this argument

Here's where the case stops depending on our read of a poker table and starts resting on public record.

Marin cannot build. Not "chooses not to" — structurally cannot, on a timeline that matters to anyone reading this.

The state assigned Marin's twelve jurisdictions 14,405 housing units for the 2023–2031 cycle, up from 2,298 in the previous one. That's 6.3 times the prior target. Through the most recent reporting, they have permitted 1,040 units — 7.2% of the goal, with roughly 42–44% of the cycle elapsed. Corte Madera is at 3.9%. Sausalito 4.4%. Mill Valley 8.0%. (HCD's RHNA progress data is public and worth clicking.)

The instructive part: in the previous cycle, Marin hit 116% of its target building at roughly the same rate. Corte Madera hit 416%. The capacity didn't change. The target did.

Meanwhile SB 79, the largest transit-oriented upzoning California has passed this cycle, effective July 1, 2026, does not apply in Marin. It covers counties with 15 or more passenger rail stations. Marin's SMART and ferry stops were classified Tier 3, and Tier 3 was removed from the final bill.

Layer on a land base that's been locked since 1972, when A-60 agricultural zoning — one house per sixty acres — took more than 136,000 acres off the table, and the 1973 Countywide Plan funneled all growth into the corridor along Highway 101 where these towns already sit. Fifty-four years later that's still the governing document.

Then Proposition 13, which in high-appreciation coastal markets adds two to three years to average owner tenure. Marin is the maximum-subsidy case. And roughly half of American mortgages are still below 4% against a market rate near 6.6%, which freezes turnover further.

Fixed supply, frozen turnover, and a demand cohort that got liquid. That's the argument.

The generational handoff nobody is pricing

Here's the part of this that we think is genuinely mispriced, and it comes out of a number that looks bad until you turn it over.

Marin is old. The median age is 47.9 against 37.9 statewide. Roughly 27% of the county is 60 or older, and by 2030 it's projected to be about one in three. State Department of Finance data shows the county's population has drifted down since 2020, and deaths now slightly exceed births.

Read that as decline if you want. We read it as the largest transfer of desirable housing stock in the county's modern history, arriving over the next ten to fifteen years.

These are houses bought in the 1970s and 80s, held through Proposition 13 for decades, on streets that will never be replicated because the zoning that protects them also prevents anyone building anything like them again. They are coming to market. Not all at once, and not on a schedule anyone can publish — but they are coming, and the generation selling them is not being replaced from within Marin.

So who buys them? Someone has to, and the county isn't producing those buyers internally. The households with both the means and the motive are in San Francisco: thirty minutes away, in the family-forming years, and about to be liquid.

That's the thesis in one sentence. An aging county with fixed supply, handing off exceptional houses to a cohort in San Francisco that is coming into money at exactly the moment it starts thinking about school districts. The population number that looks like weakness is the mechanism.

Where we could be wrong

Two things, and you should know them.

The best study on this finds the effect arrives at filing, not at cash-out. Hartman-Glaser, Thibodeau and Yoshida looked at 725 California IPOs from 1993 to 2017 and found home prices near headquarters rose about 1.0% after filing and 0.8% after issuing, but approximately zero after lock-up expiration — people buy on anticipated wealth, not realized wealth. If you're waiting for December lock-ups to fire a starting gun, the gun may have gone off at the S-1.

That cuts toward acting sooner rather than later, which is the practical upshot either way.

Timing is not a science. After the 2020–21 IPO wave, the Case-Shiller San Francisco index peaked in May 2022 and fell 12.74% by February 2023. Liquidity is real and lumpy, and the exact month is unknowable. Buy a house you want to live in for a decade and the entry point stops being the deciding variable.

One correction to our earlier version: we included SpaceX, and it mostly doesn't belong here. It priced at $135 in June 2026 and raised $75 billion, the largest IPO ever — but it's headquartered in Starbase, Texas, with its main California operation in Hawthorne, and the coverage of SpaceX wealth reaching real estate is about the South Bay of Los Angeles. OpenAI and Anthropic are the Marin story. SpaceX was us reaching.

What we'd actually do

The useful version of this thesis isn't a price prediction. It's about what to watch, and the signal is public.

Liquidity shows up in terms before it shows up in prices. Contingency waivers first, then appraisal-gap coverage, then flexible close dates, then the number. By the time a median moves, terms moved two quarters earlier. If you want to know whether anything is arriving, don't watch the median — watch how many recent sales in a given town closed without a loan contingency.

For buyers: the asymmetry favors getting positioned before rather than during. Financing pre-underwritten, an agent relationship that's real, off-market network active. And if you find yourself in an eight-offer situation the month after a splashy listing, that is the moment to be most disciplined about your top number, not least. We'd rather lose a deal than hand you a bad one.

For sellers: if you're flexible on timing, the vesting and lock-up calendar of a major employer is public information and worth understanding. We'll walk through it with you rather than treating it as proprietary — it isn't, and any agent implying otherwise is selling access to a Google search.

The size and the timing are genuinely uncertain, and anyone who tells you otherwise is selling something. What isn't uncertain is the structure: supply here is fixed in a way almost nowhere else in the Bay Area is, an unusually large share of the county's best houses will change hands over the next fifteen years, and the people best positioned to buy them are thirty minutes south and about to have the money.

We think that's the most interesting setup in the Bay Area right now. We're telling you it's our read rather than a forecast — but it's a read we're making with our own money and our own families, which is the only version of this worth publishing.