The number that stopped me was a ranking, not a price. Among the 40 largest housing markets in the United States, Silicon Valley real estate came in last for year-over-year price growth — dead last, at a median around $1.6 million, according to a Homes.com analysis of CoStar data reported in the regional business press at the time. Last place. The most expensive metro in the country, and the slowest-moving one.
I have watched people do two very different things with that number in the same week. One group read it as the end of an era: the engine is stalling, the talent is leaving, sell the fourplex. The other read it as noise: prices are high, growth off a high base is naturally slow, nothing to see. Both readings are doing the same thing — treating a median sale price as if it were a diagnosis.
It isn't. It's a symptom, and a laggy one. So before you make a leveraged decision on it, it's worth being precise about what that figure counted, what it left out, and which of the two things you're actually worried about — a downturn or a change in the region's composition — it can speak to.
What the number actually measured
A median closed price, compared to the same month a year earlier, in a market with very thin volume.
Each of those clauses matters. Closed means the transaction cleared escrow, which means it was priced 30 to 60 days before it showed up in the data, on expectations set 30 to 60 days before that. You are reading last quarter's confidence.
Median means the middle sale of the ones that happened. It is not a value estimate. If the sellers who transact this year skew toward smaller homes, older housing stock, or the estates and relocations that have to sell regardless of conditions, the median falls even if no individual house lost a dollar. In a market where a meaningful share of owners are sitting on sub-4% mortgages and have no reason to list, the pool of transactions is self-selected in ways that move the median independently of value.
Thin volume amplifies all of it. Santa Clara County's monthly closed-sale counts are small enough that a few dozen transactions in the wrong price band can swing a ZIP-level median several percent. Rank that against 39 other metros and you get a headline. You do not necessarily get a trend.
None of this means the weakness is fake. Compass and other brokerages tracking the region have been consistent that inventory sat longer and price reductions were more common than the national pattern. The engine did cool. But the number tells you that fewer confident buyers showed up at a specific price point in a specific quarter. That is a real fact with a narrow reach.
What it doesn't measure
Three things, and they're the three you actually care about.
It doesn't measure withdrawal. A price median counts completed sales. It says nothing about the listings that came off the market unsold, the owners who decided not to list at all, or the buyers who stayed in the rental pool for another year. In a stalled market, the most informative population is the one that didn't transact — and by construction, it's invisible in the metric.
It doesn't distinguish departure from non-arrival. These look identical in a price series and mean opposite things for your ten-year outlook. If 3,000 people are laid off — Meta's cuts in that stretch were reported at roughly that scale — most of them do not sell and leave. They take severance, they take a lower offer at Cisco or Adobe or a Series C nobody's heard of, they stay. That's a cyclical dent in demand that repairs itself on the next hiring cycle. If, instead, the twenty-six-year-old who would have arrived on an H-1B and rented in Sunnyvale for four years and bought in Fremont in year six never enters the region at all, nothing shows up in this year's price data. It shows up in 2032's.
It doesn't measure household formation. Joint Venture Silicon Valley has documented for years that roughly 41% of the region's population is foreign-born. That's not a demographic footnote in a housing market; it's the demand pipeline itself. The Bay Area housing market has, for three decades, run on a conveyor: arrive, rent, get promoted, get a green card, buy. Slow the front of that conveyor and the back end doesn't notice for half a decade. Then it notices all at once.
The forty-mile counterpoint
Here's the fact that keeps me from writing the obituary. While the South Bay posted the weakest growth in the country's largest markets, San Francisco — a forty-minute drive up the peninsula — was doing fine. Better than fine, in the price tiers where AI money concentrated.
That divergence is the most useful piece of information in the whole story, and it's the one the headline number can't carry. If this were sector-wide collapse, San Francisco would be falling too. It isn't. What we're watching is a sorting event: demand concentrating around a small number of extremely well-capitalized firms in a dense urban core, while the broad mid-tier of the industry — the enterprise software companies, the hardware groups, the internal tools org at a company you've heard of — absorbs the layoffs and the flat comp. Suburban Santa Clara County housing was priced on the health of that broad middle. That's what softened.
So the question you should be asking isn't "is tech okay." It's whether the middle of the industry recovers its hiring, and whether the arrival pipeline that has always refilled the bottom of the ladder stays open.
How I'd actually read the next twelve months
I'd stop watching medians. They're the noisiest available signal and the slowest. Here's what I'd watch instead, and what each is honestly good for.
New listings versus pendings, monthly, county level. Best at: telling you within weeks whether demand is thinning or supply is thickening, which the median blends into one useless number. Wrong for: anything about long-run value.
Months of supply, split above and below $2 million. Best at: separating the layoff story from the immigration story. First-time and move-up buyers cluster below; the equity-rich cluster above. If the bottom band stalls while the top band clears, you're watching pipeline, not panic. Wrong for: small samples in individual cities — keep it countywide.
Class A rental absorption in Sunnyvale, Santa Clara, and Fremont. Best at: catching arrivals, because newcomers rent first. This is the earliest honest read on whether people are still coming. Wrong for: anyone who needs a clean number — you'll be reading operator reports and asking-rent trends, not an index.
Visa and consular throughput, plus employer petition volume. Best at: the structural question, directly. Wrong for: timing anything, because the effect lands years out.
Headcount announcements by city, not by company. Best at: seeing the sorting event in real time. Wrong for: sentiment — announcements lag decisions.
Who should care, and who is over-reading this
If you're a leveraged small landlord in the South Bay, this number matters and you should already be stress-testing against flat rents and longer vacancies, not against a price crash. Your risk is time-on-market and refinance timing, not valuation.
If you're on city or county finance staff, the price series is close to irrelevant to you in the near term — Proposition 13 means your assessed base doesn't reprice with the market. Transfer tax volume and permit applications are your leading indicators, and both are worth pulling now.
If you're a would-be buyer who's been priced out for six years, this is the softest the market has been in a while, and the honest thing to say is that it's still a $1.6 million median. Weakest-in-40 growth on that base is not an opening. It's a pause.
If you're an economist or an allocator asking whether this is cyclical or structural, the answer available from this data is: the layoffs are cyclical and the price weakness mostly reflects them. The immigration slowdown is the structural variable, and it is not yet visible in any price series, because the people it affects haven't reached the age where they buy. That's not reassurance. That's a warning that the number you're reading is the wrong one.
The region has absorbed worse. It absorbed 2001, when the office vacancy in Santa Clara County went to numbers that looked like typos, and it absorbed 2008. What it has never done is run without new arrivals.
Layoffs move the price. Immigration moves the market. Watch who arrives, not what closes.
