Four companies within twenty miles of each other are worth $16.6 trillion, and AI spending drove three quarters of America's growth last quarter. The concentration, measured honestly. Page 5.
7.32 -- the first weekly drop since the climb began in August, four days before the Fed decides whether to take it back. Page 6.
Zillow finally posted September: early winter in forty-one states, spring in nine Bay Area counties -- and Atherton is America's priciest ZIP code for the ninth time. Page 7.
The facts as of press time: Anthropic, the San Francisco artificial-intelligence company, filed confidentially for an IPO on June 1 and is expected to make its S-1 public within days -- possibly the day this letter arrives.† The roadshow is slated for the week of November 9, with a Nasdaq listing targeted for mid-November.† The valuation conversation runs from $1.5 trillion to the $2 trillion the company is reaching for -- either number makes it the largest public offering ever attempted, past June's SpaceX debut.† The business underneath: $16.8 billion of revenue in the third quarter alone, up nearly half from the second -- a company compounding at a pace the public markets have simply never been offered at this size.†
This letter does not pick stocks, and nothing on these pages is investment advice. The reason the listing leads the issue is geography. Anthropic is not a Delaware abstraction; it is an employer whose offices sit in San Francisco and whose engineers, researchers and executives already rent and own across the city and down this Peninsula. When a private company lists, its employees' paper becomes money. At these valuations, the letter's back-of-envelope (e): even a few thousand employees holding ordinary startup equity implies tens of billions of dollars of newly liquid wealth, concentrated among people who already live within commuting distance of the neighborhoods on page 7's Board.
And it lands on a market this letter has measured weekly: eleven-year-low sales volume, inventory welded shut by three-percent mortgages and a frozen tax exemption, and prices that rose through a 7.4 percent autumn anyway. The standoff has been a war of patience between sellers who will not sell and buyers who cannot stretch. A liquidity event of this size does not negotiate with that equilibrium; it simply adds buyers who do not need mortgages. The Honest Ruler's neighborhoods -- Crescent Park at $2,471 a foot, Lindenwood up 25 percent -- were repriced by the last wave of SI-economy money while rates were at two-decade highs. This would be the first wave that arrives with its own bell.
Confidential S-1: June 1. Public S-1: expected any day (not yet on EDGAR as this issue went to press -- this desk checked). Roadshow: week of Nov 9. Listing: Nasdaq, mid-November. Valuation talk: $1.5--2 trillion. Q3 revenue: $16.8B, up ~50% from Q2. Scale: largest IPO in history, ahead of SpaceX's June offering.
What history says such a listing does to the streets around it is a measurable question -- and it has been measured. The precedents, including the one finding almost nobody quotes correctly, are on the facing page.
The cleanest study examined 725 California IPOs from 1993 to 2017 and tracked home prices around each company's headquarters. The finding: prices within ten miles rose about 1 percent more than the surrounding county after the filing date, and a further 0.8 percent after the listing itself -- but showed no additional jump when the lockup expired (e).† The market, in other words, does not wait for the employees' shares to unlock; it reprices the moment the wealth becomes visible. Sellers planning to "wait for the lockup" are planning to arrive after the move.
The single-company evidence agrees. After Facebook's 2012 offering, Zillow found home values in census tracts where Facebook employees lived climbed 21 percent in the following year, against 17 percent for the rest of the Bay Area -- and every ten Facebook employees in a tract added another 1.6 points of appreciation.† Google's 2004 listing, whose lockup lapsed in spring 2005, visibly stepped up the high end of Mountain View, Sunnyvale and Los Altos and set a floor under the south Peninsula for years (e).
Applying it here (e): the filing effect may already be underway -- the offering has been Silicon Valley's open secret since June, and page 6's nine-county September rise happened with this listing in plain view. The listing effect arrives in mid-November, in what is normally the market's quietest month. And the volume effect -- actual transactions by newly liquid employees -- builds into spring 2027, when a standard six-month lockup would release shares just as the market season opens. Every quiet-winter-versus-spring conversation in the quiet book now has a new entry on the ledger: a winter sale meets the anticipation; a spring listing meets the cash.
The honest caveats, printed at full size (e): valuations under discussion have already moved a half-trillion dollars in a month; an IPO priced into a falling tape repriced every precedent above; and employee equity at late-stage valuations is not 2004-style lottery paper. The letter will grade this page the way it grades every forecast -- in print, after the fact.
| Event | When | Historical effect |
|---|---|---|
| Public filing | Any day | +1% near HQ vs county |
| Listing day | Mid-November | +0.8% further |
| Lockup expiry | ~Spring 2027 | No price jump -- but the transaction wave |
What Anthropic does. Founded in 2021 by Dario and Daniela Amodei and a group of researchers who left OpenAI, Anthropic builds Claude, a family of large language models, and sells access to it -- to companies that wire it into their products, to professionals through its apps, and above all to software teams, where its coding tools became the industry standard almost overnight.† What sets the company apart commercially is the pairing (e): frontier capability, especially at writing working code and running multi-step tasks, with a reputation for safety research that made it the vendor cautious enterprises could approve. That pairing is what compounds $16.8 billion of quarterly revenue -- and it is why the firm's research arm matters to the next paragraph.
What a large language model is. Strip the mystique and an LLM is a prediction machine: a program trained on trillions of words to guess, over and over, what word plausibly comes next. The architecture -- the transformer -- is published and well understood, the way a floor plan is understood. Training tunes billions of internal dials until the guessing becomes uncannily good, and somewhere along the way the machine starts doing things nobody explicitly taught it: translating, reasoning through problems, writing code. Even the name tips its hat to the idea's inventor: Claude is widely understood as an homage to Claude Shannon, the Bell Labs mathematician who founded information theory in 1948 and ran the first next-letter guessing games on English text -- the experiment every language model descends from.† An LLM is Shannon's parlor game, played with billions of dials (e).
The paradox. Here is the honest, strange part: we know the blueprint completely and the behavior barely. The dials are not set by engineers; they are grown by the training process, and no one can read a trillion numbers and say why the machine answered as it did. These systems are grown more than they are built, and the young science of opening them up -- interpretability, a field Anthropic itself leads -- is still mapping which clusters of dials mean what, the way early anatomists mapped a brain.† The largest IPO in history is, in part, a bet on a machine its own makers are still learning to explain (e).
The machine after this one. Language models learned from text -- the world as described. The next generation, called world models, learns from the world itself: video, motion, cause and effect. Systems like Meta's V-JEPA, NVIDIA's Cosmos and Fei-Fei Li's World Labs aim to give machines an internal physics -- a sense of objects, space and consequence -- so they can predict what happens next in a room, not just in a sentence.† That is the doorway to robotics and the physical economy: machines that can plan a task, not just describe one (e).
LLMs reshaped knowledge work in four years; world models would carry the change into warehouses, job sites and kitchens. Readiness is not abstract -- it is power lines, data-center land, retraining, and housing for the people who build it. This letter's ledgers say the region inventing that future approves four lot splits a city. The machine is ready before we are. That gap, not the technology, is the thing to watch.
Start with growth. In the first quarter of this year, spending on artificial intelligence -- data centers, chips, power -- accounted for roughly three quarters of all U.S. economic growth.† Bridgewater puts AI capital spending at about half of 2026's growth, rising further next year.† Who directs that spending? Companies run from this bay: Nvidia in Santa Clara, Alphabet in Mountain View, Meta in Menlo Park, Apple in Cupertino, OpenAI and Anthropic in San Francisco. Those first four alone are worth about $16.6 trillion -- four campuses within a twenty-mile drive of each other, worth more than the annual output of every country on earth except the United States and China (e).†
The stock market tells the same story. The fifteen largest S&P 500 companies now hold a record 46.7 percent of the index; strip the AI-linked names out of the market's two-year run and a 142 percent gain collapses to roughly 16.† Most of those names bank, build and badge within an hour of this page.
The money that funds the next generation is even more concentrated: in the first quarter, more than 80 percent of all U.S. venture capital went to Bay Area companies.† Even Sacramento rides the wave: personal income tax supplies about two thirds of California's general fund, and this year's budget was rescued by a $21 billion upgrade driven by capital-gains realizations -- the state's checkbook is, functionally, a derivative of Bay Area stock gains.†
Now the honest paragraph, because concentration is not a victory lap (e): an economy carried by one region's one industry is an economy exposed to it. If the AI buildout disappoints, the same arithmetic runs in reverse -- the index, the state budget, and yes, the bid under Peninsula housing would all feel it together. The letter's read for property owners is narrower and sturdier: whatever one believes about the boom, the land under its headquarters is the scarcest input in the entire system. Chips can be fabbed elsewhere and models trained anywhere; the thirty-mile shoreline where the people who decide these things insist on living is not expandable. That is what page 7's nine-county September says, and what every ledger this letter has printed says underneath.
AI capex: ~75% of Q1 U.S. growth. Four Bay companies: ~$16.6T. Top 15 stocks: record 46.7% of the S&P. Venture capital to the Bay: 80%+ of the U.S. total last quarter. California's general fund: two-thirds personal income tax, swung $21B by capital gains. One peninsula, carrying all of it.
The tape: Freddie Mac's 30-year fixed averaged 7.32 percent, down from 7.40 -- the first weekly decline in nine weeks of prints, after seven rises and a plateau.† The 15-year eased to 6.68.† The grade on our own page: "a plateau, not a peak that breaks" survived exactly one week before the plateau broke -- downward, which is the direction the page did not predict. Marked as a half-miss, at full size. The cause is mundane and therefore sturdy: Treasury yields drifted off their highs as the war premium in oil faded and last week's hot CPI got digested rather than extended.
What 8 basis points buys: about $110 a month on a $2 million loan (e) -- real, and not yet a regime change. The refinance file stays closed. The underwriting discipline eases from 7.5 to 7.375 for the careful (e).
The asterisk is dated October 27--28, four days after this issue: the Fed meets with markets still pricing a meaningful chance of a hike after September's 3.7 percent CPI -- and December-hike chatter already circulating.† A hawkish surprise takes the week's 8 basis points back by Friday; a plain hold lets the crack widen. For the fall market's last six weekends, the practical read (e): buyers who locked at 7.4 should re-shop the lock; sellers get one more argument for the prepared-winter sale -- the first rate drop in two months is the kind of headline that reactivates paused buyers, and page 2's bell is about to reactivate a different kind entirely.
30-yr: 7.32%, −8 bp -- first decline in 9 weeks.† 15-yr: 6.68%.† A year ago: 6.19%.† Next verdicts: FOMC Oct 27--28, October CPI mid-November, and the election in between.
The national picture: Zillow's September report finds newly pending sales down 8.5 percent, with winter arriving early across most of the country as buyers retreated from seven-percent money.† California closings fell 6.5 percent from a year ago.† Volume, nationally and statewide, is the standoff this letter has measured all autumn.
And then the nine counties around this bay: every one of them rose in September, for the second straight month.† San Francisco climbed 2.0 percent in a single month -- its thirteenth consecutive rise -- to a typical value near $1.4 million. Marin matched the streak at $1.5 million. San Mateo reached $1.6 million. Santa Clara and Sonoma inched 0.4; Alameda, Contra Costa and Napa 0.2; Solano 0.1.† Prices rising through a 7.4 percent October, in every county at once, while sales fall -- the standoff's signature, now in Zillow's own ink.
The line for this letter's readers: Atherton was named the most expensive ZIP code in America for the ninth time -- typical home value $8.6 million, up 16.2 percent in a year.† Set that beside last issue's ledger: 123 new houses applied for since 2022, 61 under construction this minute, a median build of 2.8 years. The country's priciest dirt is also its busiest estate construction site -- the two facts are one fact. And the September rise happened with page 2's listing in plain view, which is exactly what the precedent research on page 3 predicts the filing phase looks like (e).
| County | Typical value | Month |
|---|---|---|
| San Mateo | ~$1.6M | +0.8% |
| Marin | ~$1.5M | +0.7% |
| San Francisco | ~$1.4M | +2.0% · 13th straight |
| Santa Clara | -- | +0.4% |
| Sonoma | -- | +0.4% |
| Alameda · C. Costa · Napa | -- | +0.2% |
| Solano | -- | +0.1% |
Zillow Home Value Index, September 2026. Atherton: $8.6M typical, +16.2% y/y, priciest U.S. ZIP a ninth time.
Nationally, the needle has stopped moving: the generic ballot holds in the D+7 to D+8 band it reached two weeks ago, and prediction markets still make the House roughly nine-in-ten Democratic and the Senate about six-in-ten.† The blue-sweep tail this letter flagged in early October has stopped fattening and started waiting. Our printed scenarios -- split decision, blue sweep, red hold -- get their grades next issue, with the returns beside them.
The housing stakes have not changed since page 7 of last issue: a sweep makes federal demand-side housing programs likelier, and the bond market that sets page 6's rate will be watching the margin as closely as the map.
In California, the September PPIC survey put Proposition 37 -- the $25 billion middle-income mortgage bank -- at 61 percent support, consistent with the three-in-four passage odds the markets have priced all month.† Proposition 45's CEQA streamlining still trails badly, and the local locks -- Los Altos Measure D and Menlo Park's Measure P -- head to their own verdicts the same night.† The two-doors read stands: demand passing, supply failing, and now a two-trillion-dollar neighbor walking through the open door (e).
The three scenarios from No. 10 · Props 1, 37, 45 · Los Altos Measure D · Menlo Park Measure P · and this page's "stopped fattening" call. All of it in print, at full size, next issue.
The bell changed the book's arithmetic before it rang. The Los Altos owner with the just-finished new build -- quiet winter versus public spring, three issues running -- now faces the version of that question with page 3's timing table in it: a prepared winter sale meets the anticipation phase, a spring listing meets the lockup cash and every other seller who had the same idea. The lean has shifted to quiet-now-with-a-price- discipline; the family reads this letter, and this page, and knows it. The mid- Peninsula couple running the frozen-exemption arithmetic advanced a square: their CPA has blessed the two-year structure in principle, and the remaining question -- lease to strangers or to the neighbors -- is exactly the one page 3 of last issue flagged for substance. Progress is slow because careful is slow.
The west-Atherton trust convenes eleven days from now, after the returns. The family's newest datum is not political: it is Zillow naming their ZIP code the country's most expensive for a ninth time, up 16.2 percent in a year, which strengthens the sell-as-it-stands argument another notch (e).
Elsewhere: the museum-grade Eichler's preservation buyer produced the lender letter, and this week's 8-basis-point break is precisely the nudge that turns letters into offers -- the weekend ahead will tell. And the cleared-lot thread from last issue's Atherton ledger has begun to answer: conversations are open with owners of two of the twenty-two lots, both of whom cleared land for a project that 7.4 percent money then shelved. A two-trillion-dollar neighbor with a mid-November bell is, among many things, the best argument those lots have heard in two years for becoming houses again -- or for changing hands to someone who will build one (e).
No addresses in print. Subscribers called first, in subscription order. Vetting both ways. Conversations confidential without exception.
Payne Sharpley · 650.384.3546 · [email protected]
Reply to this email with the word CALLS and the kind of property you watch for. Replies reach this desk directly, and the book's first-call order is strictly first-come.
For ten issues this letter has described a market in perfect, frustrating balance: owners who cannot afford to sell, buyers who cannot afford to stretch, prices held aloft by the scarcity both sides created. Every force in that standoff was a subtraction -- a mortgage rate, a tax freeze, a ballot measure bolting a parking lot. This week the first large addition in years stepped onto the field, and it happens to be the largest ever assembled. The letter's job next is discipline: the precedents on page 3 are averages, the valuation is still being argued a half-trillion dollars at a time, and page 4 was honest that the machine under the money is imperfectly understood even by its makers.
But the standoff has always been a question waiting for a force majeure, and it is hard to name a likelier one than tens of billions of dollars of new liquidity, held by local hands, arriving into the thinnest inventory this region has recorded. The ice on this market was never going to melt from patience. It melts from heat. The bell is heat. Eleven days to the election, four to the Fed, and a filing that could land before this letter finishes sending. The letter will be here Friday, measuring all of it.
| Any day · Anthropic's public S-1 -- read here when it lands | p. 2 |
| Oct 27--28 · FOMC -- hold or take the 8 bp back | p. 6 |
| Nov 3 · The midterms + Props 1/37/45 + Measures D & P -- graded in print | p. 8 |
| Week of Nov 9 · The roadshow begins | p. 2 |
| Mid-Nov · The listing · October CPI · Nvidia earnings | Standing |
| Before Thanksgiving · Building Ledger III: Los Altos & Mountain View | Standing |
| Dec 11 · Federal funding expires -- round two | Standing |
| Dec 31 · Atherton RM-10 trigger | Standing |
The Fed's verdict, the election graded at full size -- and the S-1, dissected, if it has landed. Reply with what you want covered. I read and answer every response.
Anthropic IPO: confidential June 1 filing, public S-1 expected imminently, roadshow week of Nov 9, mid-November Nasdaq target, $1.5--2T valuation range, Q3 revenue $16.8B: Forge Global, PYMNTS, Yahoo Finance and Luminix reporting, Oct 2026. EDGAR checked directly by this desk Oct 23; no public Anthropic S-1 posted at press time. Employee- wealth arithmetic is the editor's (e). IPO housing precedents: 725-IPO working paper via the San Francisco Chronicle (filing +1%, listing +0.8%, no lockup jump, within 10 miles of HQ); Zillow Research on Facebook-employee tracts (+21% vs +17%, 2012--13); Google/2005 lockup step-up via contemporaneous coverage. Company background and interpretability: Anthropic public materials and transformer-circuits.pub. World models: Meta V-JEPA, NVIDIA Cosmos, World Labs public materials. Characterizations of model opacity and readiness are the editor's (e).
AI capex ~75% of Q1 2026 GDP growth: BEA third estimate via TFTC; ~half of 2026 growth: Bridgewater via Edward Conard roundup. Market caps (NVDA $5.5T, AAPL $4.9T, GOOGL $4.3T, META $1.8T): Motley Fool market-cap tables, Oct 2026. Top-15 concentration 46.7%: Crypto Briefing; ex-AI index arithmetic: CNBC, Oct 15. Venture share 80%+: SF Examiner on Q1 data. California general fund and $21B capital-gains swing: Governor's Budget revenue estimates and CalMatters. Country-GDP comparison is the editor's (e).
30-year 7.32% (from 7.40), 15-year 6.68%, year-ago 6.19%: Freddie Mac weekly survey, Oct 22. FOMC dates: Federal Reserve. December-hike discussion: rate commentary, mid-Oct. Payment arithmetic (e) is the editor's.
National early winter, pending sales −8.5%: Zillow September 2026 Market Report. Nine-county September values and streaks, Atherton $8.6M typical / +16.2% y/y / priciest ZIP ninth time: Zillow Home Value Index via Bay Area press, Oct 2026. California September closings −6.5%: WRE News on C.A.R. data. Ledger cross-references: this letter's No. 11.
Generic ballot and chamber odds: polling aggregates and prediction markets, mid-to-late Oct. Prop 37 at 61%: PPIC Statewide Survey, September 2026. Measures D and P: city materials and The Almanac. Scenario readings (e) are the editor's, graded after Nov 3.
Figures marked † are from the named sources above, as published and dated. Figures marked (e) are the editor's estimates, derivations, opinions or speculation. Corrections print at the size of the original claim -- this issue grades No. 11's "plateau" call as a half-miss, at full size, on p. 6.
This issue is not investment advice. Nothing on these pages is a recommendation to buy, sell, or subscribe for any security, including any interest in Anthropic's offering; the editor holds no position in the company and is a real estate licensee, not an investment adviser. IPO outcomes, valuations and lockup terms change; historical housing effects are averages, not promises. Tax strategies referenced from No. 11 continue to require a CPA and tax attorney.
Disclosures. PAYNE is written and published by Payne Sharpley, CalRE #02195155, Intero Real Estate Services, a Berkshire Hathaway affiliate · Payne Sharpley Group · member, Forbes Global Properties. Office: 10275 N. De Anza Blvd, Cupertino, CA 95014 · 650.384.3546 · [email protected]. Nothing herein is a solicitation of property already exclusively listed with another broker. Election and policy commentary is analysis, not advocacy. Equal Housing Opportunity.
© 2026 Payne Sharpley. Subscribe free: paynesoldit.com/payne · One email a week, one-click unsubscribe always. Mobile edition of this issue: paynesoldit.com/payne/no12.