Pull up two San Francisco housing reports published within days of each other this summer and you will find two different cities.
One says the median sale price across the city fell from $1.64 million in July 2025 to $1.33 million in June 2026, a drop of roughly 19 percent. Another, built from the same San Francisco Association of Realtors MLS data that every licensed brokerage in the city can pull, says the median single-family home sold for $2.15 million in the second quarter of 2026, up 22.2 percent over the same quarter last year.
Neither number is wrong. Neither is cherry-picked. They are measuring different things, and the gap between them is the most useful piece of information a buyer or seller can have right now if they are trying to compare one San Francisco neighborhood to another.
The Basket Problem
The blended citywide median includes every closed sale: houses, condos, TICs, co-ops, everything. When condo sales make up a larger share of what closes in a given month, the citywide number drops even if not a single individual property lost value. That is what has been happening. Condo inventory and condo prices have moved at a fraction of the pace of single-family homes, so as condos take up a bigger slice of the transaction mix, the blended median gets pulled down while houses keep climbing underneath it.
Vanguard Properties' own Q2 2026 market update, drawn from SFAR MLS and BrokerMetrics data, makes the split explicit: single-family homes at $2.15 million, up 22.2 percent, against condos, TICs and co-ops combined at $1.3 million, up a much more modest 6.9 percent. Inventory tells the same story from the supply side. As of July 2026, single-family listings were down 42.11 percent year over year to 154 homes citywide, and condo months of supply had fallen from 3.6 a year earlier to just 1.7, tight but nowhere near as constrained as the house market.
If someone tells you San Francisco is up or down without saying which of those two markets they mean, they are guessing.
Where the Money Actually Went
The obvious explanation for the single-family surge is the AI sector. It is also the correct one, but the mechanism matters more than the label.
Redfin's chief economist, Daryl Fairweather, put it plainly to CNN this month: AI is different from past San Francisco tech booms because of how narrowly it concentrates wealth, landing on the relatively small number of people who work at or hold equity in companies like OpenAI and Anthropic before either has actually gone public. Roughly one in three Bay Area home sales closed all-cash between April and June 2026, a sign of buyers who are not waiting on financing because they do not need it.
That concentration is the whole story. One recent estimate put the pre-tax AI equity created so far at roughly $198 billion, but noted it belongs to fewer than 10,000 people, most of whom already live in San Francisco. That detail changes the picture entirely. This is not a wave of new arrivals spreading demand evenly across the city's districts. It is a small number of already-local, already-wealthy households trading up, in place, in the neighborhoods they already know.
That is why the effect shows up as a handful of extremely hot single-family submarkets rather than a citywide rising tide.
The Neighborhoods Carrying the Number
Trailing twelve-month MLS closings through mid-July 2026 make the concentration visible at the district level.
| Neighborhood | Trailing-Year Single-Family Median | Sales Volume |
|---|---|---|
| Cow Hollow (District 7D) | $7.89 million | 20 sales |
| Pacific Heights (District 7B) | $6.85 million | Among the city's highest-value closings |
| Eureka Valley / Dolores Heights (District 5K) | $3.48 million | 49 sales |
| Noe Valley (District 5C) | $2.88 million | 110 sales, the deepest luxury pool in the city |
Marina and Cow Hollow, along with Russian Hill, each posted single-family medians above $9 million in Vanguard's Q2 2026 report, the highest tier in the city. Meanwhile Cole Valley, the Haight and the Sunset saw the fiercest bidding relative to list price, with homes selling for an average of more than 136 percent of asking in that same quarter. High price and high overbid percentage are not the same phenomenon, and a neighborhood can lead in one without leading in the other. That distinction matters if you are trying to decide where your offer needs to stretch versus where the sticker price alone already tells the story.
The Duboce Triangle Test Case
If you want to see the mechanism in a single listing, look at 160 Noe St, a turnkey Edwardian in Duboce Triangle that hit the market this spring for just under $3 million. The house sits a few blocks from Hayes Valley, the neighborhood techies have nicknamed Cerebral Valley for its concentration of AI hacker houses and early-stage startups. The listing agent, Kelly Bennett, described a fear-of-missing-out dynamic among buyers wondering aloud what happens to prices once Anthropic and OpenAI actually complete their public offerings.
That anxiety is rational. Both companies have been valued at roughly $1 trillion in their most recent funding rounds, and an IPO would convert a slice of that paper wealth into cash for employees who can then deploy it into exactly this kind of house, in exactly this kind of walkable, transit-close neighborhood adjacent to where they already work.
Pacific Heights offers the other end of the same trade. Real estate advisor Alexander Lurie, who recently represented the seller of a Pacific Heights home listed just under $4 million that closed at $7 million, described the ultra-luxury tier as facing a mansion deficit, with roughly 50 buyers actively shopping for $20 million homes against fewer than 10 such properties listed citywide in an entire year. The city's chief economist, Ted Egan, called the scale of this concentration unlike anything he has seen in two decades, even as San Francisco has lost roughly 40,000 jobs, mostly in technology, over the same three years. Contraction and windfall are happening inside the same industry, at the same time, in the same city.
What Still Has Room
Not every corner of this market is behaving like Cow Hollow or Pacific Heights. Condos, TICs and co-ops citywide rose just 6.9 percent in Q2 2026 against the single-family market's 22.2 percent, and while condo inventory has tightened meaningfully over the past year, it has not evaporated the way single-family inventory has. Even inside the house market, the picture is not uniform. One agent working the Outer Sunset described houses listed at $990,000 selling for $2.5 million, a spread wide enough to suggest the printed asking price there has become close to meaningless as a signal of what a property will actually command.
That spread is the practical lesson. A list price in a neighborhood absorbing AI wealth is often closer to an invitation to bid than a prediction of value, while a list price in a comparatively quieter submarket may still track closer to what a home ultimately sells for.
What This Means If You Are Comparing Neighborhoods Right Now
Before you treat any single median as a neighborhood comparison tool, it is worth asking three questions of whatever number you are looking at.
- Does this figure include condos, or only houses? A blended number will understate what a single-family home actually costs in a hot submarket, and a houses-only number will overstate what the broader market looks like.
- What quarter or month does it cover, and how many sales is it built on? A district posting fewer than 20 closings in a period can swing wildly on the strength of one or two high-end transactions.
- Is the neighborhood in question one where AI-adjacent wealth is actively concentrating, or one further from that specific demand? Proximity to Cerebral Valley, Mission Bay's AI office corridor, or an established luxury pocket like Pacific Heights or Cow Hollow behaves differently than a neighborhood outside that gravity.
None of this makes San Francisco easy to read from a portal search. It does mean the reader who understands why the numbers disagree is working with real information the headline alone will not give them.
A Few Common Questions
Does the falling blended median mean home values are actually dropping across San Francisco? Not on the evidence available. The blended citywide figure is sensitive to which property types happen to close in a given month. A year with a heavier mix of condo sales and fewer expensive houses will pull the blended median down even if individual properties held or gained value.
Will this cool off once OpenAI and Anthropic actually go public? The people closest to the market, including the agent working the Duboce Triangle listing near Cerebral Valley, are asking the same question and do not have a settled answer. What the current data supports is that a meaningful share of today's activity is already happening ahead of any IPO, driven by pre-IPO liquidity events and existing equity.
Is every San Francisco neighborhood seeing overbids like Cole Valley or the Haight? No. The 136 percent-plus average in Cole Valley, the Haight and the Sunset in Q2 2026 sat well above the citywide pattern. Overbid intensity and headline price level are two separate measures, and a neighborhood can be expensive without being especially competitive on a percentage-over-list basis, or vice versa.
If you are weighing a San Francisco neighborhood against another one this year, the number worth asking for is not the citywide median. It is the one specific to the property type, price point and district you are actually considering, and the story behind why it moved. Frank Nolan and The Nolan Group track these district-level shifts across San Francisco's micro-markets every quarter. Start a confidential conversation if you want the read on your specific neighborhood before you make a decision based on a headline that may not apply to it.
