San Francisco Housing Market: The AI Rebound Is Real, and Selective

Data as of July 2026

For three years the story about San Francisco was written in advance: emptied downtown, departed residents, a doom loop. The 2026 data does not support that story, and it does not support the opposite one either. Office vacancy is falling faster here than anywhere else in the country. The population is growing again. And a house in the Sunset still draws bids 28% above its asking price.

What the recovery has not done is make San Francisco affordable. This is the city where the standard affordability rule breaks from the income side: even at roughly $140,970, among the highest median household incomes in the country, the arithmetic on a median house does not close. That is the exact mirror image of Miami, where the rule breaks because a third of buyers pay cash and ignore the mortgage rate entirely. Here the buyers are financed, the incomes are extraordinary, and it still does not work.

Three scope notes before the numbers. First, “San Francisco” here means the City and County of San Francisco, which is the cleanest anchor in this series because the city and the county are the same jurisdiction. This follows the Chicago city-first and Miami county-first precedents. SF MLS districts 1 through 10 appear as sub-lenses; Oakland, the East Bay, and San Jose are their own markets and appear only as contrast. Second, the Case-Shiller “San Francisco” indexes cover a five-county metro area on a repeat-sales methodology, an entirely different scope from city medians, and are labeled that way everywhere they appear. Third, because houses and condos behave like different markets, this page runs the full affordability engine twice and publishes two Market Balance Score bands rather than one blended figure.

The market scorecard

San Francisco market scorecard, SF MLS trailing-twelve-month closings through July 2026, with Q2 2026 office and June 2026 rent data; mortgage rate as of the week of July 23, 2026

Median sale price, single-family

▲ up over the stated period, favors sellers: $1,850,000

City & County of San Francisco, trailing 12 months, July 2025 to July 2026

The trailing-twelve-month median is $1,850,000 and rising, with the mix-sensitive Q2 figure up 22.2% year over year. At about 13.1 times the city median household income, this is where the 28% rule breaks from the income side.

Months of supply, single-family

▼ down over the stated period, favors sellers: about 1.0 to 1.1 months

City & County of San Francisco, June 2026

About 1.0 to 1.1 months for houses, the tightest reading in this eleven-city series, with active listings down roughly 37% to 45% year over year. Condos run about 2.3 months citywide and 4 to 5 downtown.

Share of houses sold over list

● unchanged over the stated period: 83%

City & County of San Francisco, trailing 12 months, July 2025 to July 2026

83% of houses sold over list, at a median premium of $360,500 and a 117.9% list-to-sale ratio. The list price is a marketing number here, so a buyer must underwrite to comparable sales and budget the overbid.

New-buyer secured property tax rate

▲ up over the stated period, favors sellers: 1.18268325% of purchase price

City & County of San Francisco, fiscal year 2025-26

Proposition 13 resets your assessed value to the full price you pay, so a new buyer owes 1.18268325% on the purchase price while a long-tenured neighbor pays on a decades-old basis. The 2% annual cap only helps after you own.

Price-to-rent ratio, single-family

● unchanged over the stated period: about 27

City & County of San Francisco, 2026

About 27 on the single-family median against the $5,700 two-bedroom rent, the highest in this series and well past the 25 line that conventionally favors renting. Rents and prices both rose about 22%, so the ratio itself is roughly unchanged.

Case-Shiller condo price index

▼ down over the stated period, favors buyers: 273.34

San Francisco 5-county MSA (Case-Shiller scope), January 2026 , down from 275.08 in December 2025

The condo repeat-sales index eased to 273.34 in January 2026 (a 5-county MSA scope), and downtown units trade near 2015 prices with only 31% to 43% selling over list. The one place a San Francisco buyer has real leverage.

A colored triangle shows whether the change favors buyers: green favors buyers, red favors sellers. A gray dot marks a metric that is neutral for buyers (its direction is in the subtitle). Direction and color are descriptive of each metric's own stated period, not a forecast. Sources are listed in the source registry at the end of the page.

The scorecard is red-heavy, and honestly so. Prices are rising into falling supply, and the Proposition 13 reset means your tax bill is set by what you pay rather than by what your neighbor pays. One card is green: the condo index eased, and downtown units trade near decade-ago prices, which is the one place a San Francisco buyer has genuine leverage. Two cards carry a gray dot because they are structural levels rather than changes: the 83% over-list share and the price-to-rent ratio near 27. Their readings carry the message in words instead of color.

Is San Francisco a buyer’s or seller’s market right now?

Sellers, in both segments, and the second half of that sentence is the surprise this page has to explain.

FinExplained Market Balance Score (beta): Single-family

Seller's market an intense seller's market

Direction basis: The measurable single-family inputs (about 1.05 months of supply, 12 to 13 days on market, a 117.9% list-to-sale ratio, and active listings down roughly 41% year over year) land at the top of the seller band, with two of the four clamping at their normalizer ceilings. The honest read is the tightest single-family market in this eleven-city series: houses sell in under two weeks, 83% of them above list, at a median premium of $360,500. The price-cut input is missing and its weight is renormalized, and the supply and days-on-market figures are broker-sourced.

Beta: this score has not yet been backtested against historical market data, and the bands may be recalibrated. Read it alongside the metrics below, not instead of them.

Input	Reading	Normalized (0-100)	Weight applied
Months of supply	1.05 months	88	35%
Days on market	12.5 days	90	24%
Sale-to-list ratio	117.9%	100	24%
Year-over-year inventory change	-41%	100	18%

2 inputs exceed the scale's calibrated range (Sale-to-list ratio; Year-over-year inventory change), so the score is clamped at the top of what it can express. Read this band as a floor, not a point.

SINGLE-FAMILY band. Months of supply: about 1.0 to 1.1 months (June 2026, broker-sourced); the input uses 1.05, the midpoint of the sourced range.

Days on market: about 12 to 13 days; the input uses 12.5, the midpoint.

Sale-to-list ratio: 117.9%, the trailing-twelve-month list-to-sale ratio. This runs so far past the normalizer's 105% ceiling that the input CLAMPS at 100, so the score cannot fully express how extreme the over-bidding is.

Price-cut share: NOT SOURCED per property type. San Francisco publishes no per-type price-cut share, so this input is dropped and its 15% weight renormalizes across the other four (the Austin precedent). Sourcing a per-type figure from Redfin Data Center is a BACKLOG item.

Year-over-year inventory change: down about 37% to 45%; the input uses -41, the midpoint. FALLING inventory is seller-favorable and also clamps at 100 on the normalizer.

THE CAVEAT THAT MATTERS: two of the four available inputs CLAMP at the top of their normalizers, so this band is a floor on how seller-favorable the single-family market is, not a ceiling. Read it beside the metrics, and note that the broker-sourced supply and days-on-market figures carry Medium confidence because San Francisco has no single authoritative publisher for them.

How this score works

Each input is normalized onto a 0-100 scale where higher means more seller-favorable: months of supply maps 0 months to 100 and 9 or more to 0; days on market maps 0 days to 100 and 120 or more to 0; sale-to-list maps 90 percent to 0 and 105 percent to 100; price-cut share maps 0 percent to 100 and 60 percent or more to 0; year-over-year inventory change maps a 40 percent rise to 0 and a 40 percent fall to 100. The design weights are months of supply 30 percent, days on market 20 percent, sale-to-list 20 percent, price-cut share 15 percent, and inventory change 15 percent; when an input is unavailable its weight is redistributed proportionally. Bands: below 40 reads as a buyer's market, 40 to 60 balanced, above 60 a seller's market. We show the band rather than a decimal because the inputs do not support decimal precision. The score is in beta and has not yet been backtested against historical market conditions; read it alongside the underlying metrics, never instead of them.

FinExplained Market Balance Score (beta): Condo (citywide)

Seller's market seller-favorable citywide, soft downtown

Direction basis: The measurable citywide condo inputs (about 2.3 months of supply, 16 days on market, a 100.0% list-to-sale ratio, and listings down about 38%) land in the seller band, below the single-family reading but above balance. The honest read is that citywide condo statistics are tighter than the downtown narrative implies, because they blend strong neighborhood districts with the soft core. The buyer leverage that genuinely exists is concentrated downtown, where only 31% to 43% of units sell over list and days on market run 77 to 106. The price-cut input is missing and its weight is renormalized.

Beta: this score has not yet been backtested against historical market data, and the bands may be recalibrated. Read it alongside the metrics below, not instead of them.

Input	Reading	Normalized (0-100)	Weight applied
Months of supply	2.3 months	74	35%
Days on market	16 days	87	24%
Sale-to-list ratio	100%	67	24%
Year-over-year inventory change	-38%	98	18%

Scope: most of this band's inputs (Months of supply; Days on market; Sale-to-list ratio; Year-over-year inventory change) are not measured on this segment alone, so the band summarises a wider blend than its title. Where a blend hides a split, the per-segment figures on this page are the sharper evidence.

CONDO band, on CITYWIDE inputs. Months of supply: about 2.3 months citywide (June 2026, broker-sourced). Many downtown neighborhoods run 4 to 5 months.

Days on market: about 16 days citywide. Downtown districts 8, 9, and 10 run 77 to 106 days, five to six times longer.

Sale-to-list ratio: 100.0%, meaning the typical condo sold AT list, against 117.9% for houses.

Price-cut share: NOT SOURCED per property type, identical treatment to the single-family band. The input is dropped and its 15% weight renormalizes across the other four.

Year-over-year inventory change: down about 38%. Falling inventory is seller-favorable on the normalizer and is a large part of why this band reads the way it does.

THE CAVEAT THAT MATTERS, AND IT IS THE BIGGEST ON THIS PAGE: this band reads SELLER-FAVORABLE even though the downtown condo market is genuinely soft. That is not a contradiction, it is a scope problem. All five inputs are CITYWIDE, and citywide condos blend strong neighborhood districts (1, 5, and 7, where condos behave like small houses and district 5 condos sell 72% over list) with the soft downtown core. The index cannot see that split. The figures that DO show it are the district-level ones this page renders instead: only 31% to 43% of downtown condos sell over list, they sit 77 to 106 days, and downtown units trade near 2015 prices. Read this band as a citywide summary and the district figures as the real condo story.

How this score works

Each input is normalized onto a 0-100 scale where higher means more seller-favorable: months of supply maps 0 months to 100 and 9 or more to 0; days on market maps 0 days to 100 and 120 or more to 0; sale-to-list maps 90 percent to 0 and 105 percent to 100; price-cut share maps 0 percent to 100 and 60 percent or more to 0; year-over-year inventory change maps a 40 percent rise to 0 and a 40 percent fall to 100. The design weights are months of supply 30 percent, days on market 20 percent, sale-to-list 20 percent, price-cut share 15 percent, and inventory change 15 percent; when an input is unavailable its weight is redistributed proportionally. Bands: below 40 reads as a buyer's market, 40 to 60 balanced, above 60 a seller's market. We show the band rather than a decimal because the inputs do not support decimal precision. The score is in beta and has not yet been backtested against historical market conditions; read it alongside the underlying metrics, never instead of them.

The single-family band is the most extreme reading in the series, and two of its four available inputs clamp at the top of their normalizers, so it is a floor on how seller-favorable that market is rather than a ceiling. The condo band is the one that needs explaining, because it reads seller-favorable too while everything else on this page says downtown condos are soft.

That is a scope problem, not a contradiction. All five Balance Score inputs are citywide, and citywide condo statistics blend the strong neighborhood districts, where condos behave like small houses and district 5 condos sell 72% over list, with the soft downtown core. The index cannot see that split. The figures that do show it are the district-level ones: only 31% to 43% of downtown condos sell over list, they sit 77 to 106 days rather than 16, and they trade near 2015 prices. Read the band as a citywide summary and the district figures as the real condo story. Neither band uses a price-cut share, because San Francisco publishes none by property type, so that input is dropped and its weight renormalizes across the other four, per the Austin precedent.

What changed

This is the first edition of this dashboard, so the baseline is the story. The moves already visible in the sourced data:

• Office vacancy fell to 29.7% in Q2 2026 from 34.7% a year earlier, a five-point drop and the largest year-over-year improvement in the nation, from a 2023 peak near 35%.
• The population returned to growth, reaching 826,079 as of July 2025 after the 2020 to 2022 decline.
• Single-family active listings fell roughly 37% to 45% year over year, and condo listings about 38%.
• Rents set records: the median one-bedroom crossed $4,000 for the first time to $4,060, and the median two-bedroom reached $5,700, the highest in the nation, on about 22% annual growth.
• The 30-year fixed averaged 6.58% for the week of July 23, 2026, up from 6.43% in early July.
• The city’s projected two-year budget deficit narrowed from $936 million to about $643 million on stronger hotel, sales, transfer, and business taxes.

Why do houses and condos behave like different markets?

Because they are different markets, and the gap is the widest expression of it in this series.

The driver here is demand-side, and it is worth being precise about that because two other studies in this series cover condo markets in trouble for an entirely different reason. In Tampa and Miami, the condo weakness is a regulatory and insurance shock: post-Surfside reserve laws, six-figure special assessments, and mortgage blacklisting. San Francisco has none of that. What it has is remote work, uncertainty about downtown’s recovery, HOA dues that typically run $400 to $1,200 a month, and a wave of high-rise supply that delivered right before the pandemic. The softness is real, and the right word for it is soft or discounted rather than distressed, because no forced-sale data supports the harsher word.

New condo construction has also nearly stopped. Only about 3,300 units were under construction citywide in the first half of 2026, down from 4,545 five years earlier, and just 405 units reached occupancy. Against a state housing obligation of 82,069 units for 2023 to 2031, that is a supply air-pocket forming underneath a recovering demand picture.

Is the AI rebound real, or a story?

Real, measurable, and selective. The clearest number is office vacancy: 29.7% in the second quarter of 2026, down from 34.7% a year earlier and a peak near 35% in 2023. That five-point drop was the largest year-over-year improvement in the country. Kidder Mathews put the first quarter at 28.0%, and other trackers cite 22% to 25% using availability rather than vacancy methodology, so the honest read is a 28% to 30% range with the sources named.

The cause is concentrated. Per Colin Yasukochi of CBRE’s Tech Insights Center, AI companies have accounted for 30% of all leasing activity since 2023 and more than three quarters of net absorption, making the sector the primary reason vacancy is falling. Total active AI leases exceed 7.5 million square feet. OpenAI anchors Mission Bay with roughly a million square feet and somewhere between 4,500 and 7,850 employees depending on the counting method, and Anthropic holds more than a million square feet on Howard Street. Reported IPO speculation around both is exactly that, reported, and this page does not estimate a wealth effect from it.

Two things keep this honest. The recovery is geographically selective: trophy space and the AI corridors are nearly full while Class B and C buildings and the deepest downtown pockets still lag. And the office-value collapse that came before it has left a hole in the city budget, which the Proposition 13 section below explains is structurally durable rather than cyclical. Population, though, is unambiguous: 826,079 as of July 2025, growing again after the decline, with county unemployment at 3.5% in April 2026, among the lowest in California.

Does buying reset my property tax in San Francisco?

Yes, and California is the canonical case of it. Proposition 13 sets your assessed value at your purchase price and then caps growth at 2% a year, but the assessment resets to full market value when the property sells. So a new buyer pays the full fiscal 2025-26 secured rate of 1.18268325% on what they just paid, while the neighbor who bought in 1998 pays on a basis from 1998.

That is worth placing precisely against the rest of this series, because every state does it differently. Phoenix’s limited property value does not reset on sale at all. Texas and Florida reset to market annually for everyone, with caps that protect homesteaded owners and that themselves reset when a home changes hands. California resets once, on sale, and then locks growth at 2% for as long as you hold.

The consequence is golden handcuffs in both directions. Holding is rewarded, and moving is penalized: a seller gives up a low basis, pays the transfer tax, and may face capital gains above the $250,000 or $500,000 primary-residence exclusion on a long-held San Francisco home. That structurally suppresses turnover, which is part of why the city is chronically among the lowest-inventory big markets in the country. Some of San Francisco’s scarcity is tax policy rather than geography.

Proposition 19, effective April 2021, lets homeowners who are 55 or older, severely disabled, or disaster victims transfer their low base to a replacement primary residence anywhere in California, up to three times. It also repealed the old parent-child and grandparent-grandchild exclusions, so inherited property is now reassessed to market unless the heir makes it a primary residence, and even then only the first roughly $1 million of assessed-value increase is excluded. No rigorous published study yet measures whether portability has moved turnover, so this page claims nothing about that.

There is a fiscal twist worth naming. Office properties are about 17% of the city’s property-tax base, and because assessment resets on sale, every distressed office building that trades far below its assessed value locks in a permanently lower basis. The office-value collapse therefore reprices the city’s revenue for a generation rather than for a cycle, which is the mechanism behind a projected two-year deficit that ran $936 million in December 2025 before narrowing to about $643 million.

Are San Francisco prices actually rising?

Yes for houses, and the honest answer needs two measures rather than one. The trailing-twelve-month single-family median is $1,850,000 across 2,156 closings. Monthly medians ran from about $1.61 million in January to a $2.16 million peak in May, and the Q2 figure was $2.15 million, up 22.2% year over year.

That 22.2% is mix-sensitive and this page labels it that way every time it appears. Part of the spring climb is a heavier share of high-end sales as the season warms, not uniform appreciation, and it is not a claim that a given house rose 22%. Meanwhile the Case-Shiller seasonally adjusted index for the five-county metro edged down about 0.6% from January to April 2026, in the same window city medians were climbing.

Both readings are correct. Case-Shiller is a repeat-sales index that tracks the same homes and controls for what sold; a median tracks the middle of whatever transacted that month and moves with the mix. This is the same lesson the Dallas-Fort Worth study draws from its own conflicting measures, and the discipline is identical: do not pick the most dramatic figure and call it the trend. This page uses the trailing-twelve-month median as the engine basis precisely because it is the least mix-sensitive of the available city figures.

The practical consequence is the single most useful thing a San Francisco buyer can internalize: the list price is a marketing number, not an ask. Underwrite to comparable sales, budget the overbid before you write an offer, and read the disclosure package early, because in a market that moves in twelve days there is no time to start that work after you fall in love with a house.

What does it actually cost to own here?

Two answers, because there are two markets. We run the full engine twice: a single-family example at the $1,850,000 trailing median and a condo example at the $1,185,000 trailing median, both at 20% down and the 6.58% rate for the week of July 23, 2026.

The house carries about $11,573 a month in principal, interest, taxes, and insurance, rising to about $13,114 with 1% annual upkeep and about $13,739 once estimated earthquake coverage is added. At 10% down it is about $12,752 before upkeep. The condo carries about $7,335 in principal, interest, taxes, and insurance, about $8,322 with upkeep, and about $9,122 once the $800 monthly HOA dues are counted.

A jumbo caveat that matters here more than anywhere else in this series. The 2026 high-cost conforming loan limit for San Francisco County is $1,249,125. The $1,850,000 house example sits well above it, so a house purchase here normally requires jumbo financing: stricter credit and reserve requirements, 10% to 20% down as the practical norm, and qualification overlays a conforming borrower never encounters. The $1,185,000 condo example sits below the limit and can often use high-balance conforming. Our engine computes standard amortization for both, and this disclosure is how we handle the difference rather than modeling jumbo overlays we cannot source.

The buyer closing-cost band is a labeled estimate. Our sources do not publish a San Francisco buyer closing-cost survey, and because the transfer tax falls to the seller by custom, a buyer’s costs here are escrow, title, recording, and lender fees rather than the transfer tax that dominates buyer closing costs in some metros. Sourcing a proper figure is on our backlog. Itemize your own in the buyer closing cost calculator.

What income do you need to buy in San Francisco?

More than almost anyone earns, and that is remarkable in a city with these incomes. Our engine says the $1,850,000 house example requires about $495,968 a year at 20% down under a 28% front-end ratio, and the $1,185,000 condo about $314,352 on a principal, interest, taxes, and insurance basis.

Set that against what a household actually earns: about $140,970, among the highest medians in the country. The gap on the house example is roughly $354,998 a year. Third-party estimates land in similar territory from different angles: Redfin put the requirement at $443,979 in March 2026, up 7% year over year, and the California Association of REALTORS at just over $400,000. Comfort-not-affordability framings from other publishers run $321,000 to $408,000 depending on method. Our figure sits above most of them because we carry the full tax and insurance load explicitly and hold strictly to the 28% rule.

The honest reading is that the median household is not the marginal buyer here. The effective buyer pool skews far richer than the median: equity-compensated tech workers, large down payments, and buyers selling a previous appreciated home. That is why the market can clear at 83% over list while the median household cannot qualify. Test your own numbers in the home affordability calculator, the how much can I borrow calculator, and the DTI calculator.

Is it cheaper to rent or buy in San Francisco?

Renting, by the widest margin in this series, and it is not close.

Against the sourced $5,700 two-bedroom rent, owning the house example costs about $7,414 more per month before earthquake coverage and about $8,039 with it. The condo example costs about $2,622 more, or about $3,422 once HOA dues are counted. The price-to-rent ratio on the house is about 27, well past the 25 line conventionally read as favoring renting and the highest in this series.

One imperfection stated plainly: a median single-family house is larger than a two-bedroom apartment, so pairing them understates what an equivalent house would rent for. That makes the gap conservative rather than exaggerated. The $5,700 two-bedroom is simply the closest sourced comparable available, and we use one rent anchor for both examples so they stay comparable to each other.

The counterweight is Proposition 13. Renting wins the monthly comparison decisively, but an owner’s tax basis is frozen at 2% growth while rents just rose 22% in a year. That is precisely why San Francisco is simultaneously the strongest buy-and-hold market in this series and the weakest buy-short-term one. Run your own comparison in the rent vs buy calculator, which carries a San Francisco condo preset, or read the ten-city rent vs buy comparison.

What about earthquake risk?

It is the largest unpriced liability of any city in this series, and it deserves plain statement rather than drama. Standard homeowners policies exclude earthquake damage, and earthquake coverage is not lender-required. Only about 12% to 15% of California homeowners carry it, so roughly six in seven are uninsured against the peril the region is best known for.

That is a deliberate contrast with the Tampa study’s flood variant. Flood coverage there is frequently required by a lender, so it is a cost of borrowing. Earthquake coverage here is a choice, which is exactly why so few people make it. Our house example models $7,500 a year, the estimated midpoint of a sourced $6,000 to $9,000 band at the 15% minimum deductible, which is the lowest deductible available on homes valued above $1 million. That figure carries low confidence and firming it with the California Earthquake Authority’s own calculator is on our backlog. We show it as an optional layer rather than folding it into the base payment, because that is what it is.

Two submarket-specific factors matter more than the citywide average: the soft-story retrofit requirements on certain multi-unit buildings, and the liquefaction zones in the Marina, the SoMa fill, and the Mission valley floor, where 1989 Loma Prieta damage concentrated. Ask about both before you buy in those areas.

Regular homeowners insurance is a separate question, and San Francisco is better placed than most of California: the city proper sits largely outside the wildland-urban interface, so it has repriced less than wildfire areas. It is not insulated, though. After the January 2025 Los Angeles wildfires, State Farm received a 17% interim statewide homeowners increase and the FAIR Plan levied a $1 billion assessment.

Which San Francisco submarkets fit your budget?

What should buyers do with this market?

A framework, not marching orders: this is an educational read of the data, not personalized advice.

• Underwrite to comparable sales, not to the list price. With 83% of houses selling over list at a median $360,500 premium, the list is a marketing number. Decide your true ceiling from comps before you see the property, and treat the list as an opening position.
• Do the disclosure work early. In a twelve-day market, reviewing the disclosure package, inspection reports, and any permit history has to happen before you are emotionally committed.
• Prepare for jumbo underwriting if you are buying a house. At the median you are above the $1,249,125 conforming ceiling, so expect reserve requirements and documentation that a conforming borrower never sees. Get fully underwritten, not merely pre-qualified.
• Make the earthquake decision deliberately. It is not lender-required and most owners skip it. Price a California Earthquake Authority policy and decide on purpose rather than by default.
• If you are buying a condo, the building is the investment. Review the reserves, the HOA budget and dues trajectory, any pending assessments, and the owner-occupancy ratio, which affects financeability.

What should sellers do?

• On a house, the overbid machine works for you. Price to attract competition rather than to capture value on the list, because the market’s convention is to bid the list up.
• On a condo, price realism is the whole game, particularly downtown, where only 31% to 43% of units sell over list and the typical listing sits 77 to 106 days.
• Understand what you give up by selling. Your Proposition 13 basis does not travel unless you qualify under Proposition 19, and you will pay the transfer tax and possibly capital gains above the exclusion. Run your net in the seller net proceeds calculator.

What should renters do?

• The monthly math favors renting by the widest margin in this series, roughly $7,414 a month against the house example and about $2,622 against the condo before HOA dues.
• But you are renting into record pricing, at $4,060 for a one-bedroom and $5,700 for a two-bedroom on 22% annual growth, so the comparison is not static.
• Know your protections. The Rent Ordinance covers most multi-unit buildings occupied before June 13, 1979, and capped increases at 1.4% for the March 2025 to February 2026 year. Single-family homes and condos are exempt from the caps under Costa-Hawkins but keep just-cause eviction protections.

What should current homeowners do?

• Your Proposition 13 basis is an asset. Protect it. The gap between your basis and a new buyer’s is often worth more than any renovation you could make.
• At 55 or older, understand portability. Proposition 19 lets you carry your base to a replacement primary residence anywhere in California, up to three times, which is the one clean exit from the golden handcuffs.
• Check whether your building falls under the soft-story retrofit requirements, and reprice your earthquake decision periodically. Refinancing near 6.58% is marginal unless you bought at the rate peak; find your break-even in the refinance calculator.

What should investors consider?

• Rent control changes the underwriting entirely. In a pre-1979 multi-unit building you are modeling 1.4% annual increases, not market growth. Costa-Hawkins exempts single-family homes and condos from the caps, which is why those trade differently.
• The condo-discount thesis is real but not free. Buying downtown near 2015 pricing is the clearest value setup in this series, and HOA dues of $400 to $1,200 a month are the drag that decides whether it works. Stress it in the rental property ROI calculator.
• The two-way lock-in cuts against short holds. Between the transfer tax, the reset you trigger for your buyer, and capital-gains exposure, San Francisco rewards long holds and punishes quick ones.

Three scenarios for the next 12 months

No single price forecast here. Instead, three scenarios with the signals that would confirm or break each.

The selective recovery continues (the base case). AI leasing holds, office vacancy grinds toward the mid-20s, houses stay tight, and downtown condos slowly absorb. Confirmation: quarterly leasing stays above 2 million square feet and condo days on market fall from the 77-to-106 range. Invalidation: leasing drops below 2 million square feet for two consecutive quarters.

The condo discount closes. Return-to-office and AI hiring lift downtown demand enough that the near-2015 pricing gets bid up. Confirmation: downtown over-list share rises out of the 31%-to-43% band and the Case-Shiller condo index turns up. Invalidation: HOA dues keep rising and downtown days on market lengthen further.

Rates or the AI cycle break the setup. A sustained move in the 30-year fixed, or an AI funding pullback that slows leasing and hiring, would hit the top of the house market and the rent records together. Confirmation: PMMS moves materially off 6.58% or announced AI expansions pause. Invalidation: leasing pace and rent growth both hold.

What to watch next month

• The AI leasing pace and the office-vacancy trajectory, the clearest read on whether the rebound holds.
• Condo absorption downtown, especially the over-list share and days on market in districts 8, 9, and 10.
• The Freddie Mac PMMS 30-year fixed against the current 6.58%.
• The BUILD Act transfer-tax proposal, which is a proposal and not law.
• Housing production against the 82,069-unit obligation, now under annual state review through SB 423.

Run your own numbers

Every dollar figure on this page came from our tested calculator engine at stated assumptions, and each of these tools lets you swap in your own: rent vs buy, mortgage, home affordability, how much can I borrow, buyer closing costs, seller net proceeds, refinance, DTI, property tax, and rental property ROI. San Francisco is the mirror image of our Miami housing market study, where the same affordability rule breaks from the cash side rather than the income side, and the inverse of our Chicago housing market study, where a high tax rate rather than scarcity is what makes the total cost bite. Compare the rest of the series: Austin, Tampa, Phoenix, Denver, Nashville, Atlanta, Charlotte, and Dallas-Fort Worth, and for how metro costs shape long-term plans, the FIRE number by metro study.

Frequently asked questions

What is the median home price in San Francisco in 2026?

About $1,850,000 for single-family homes and $1,185,000 for condos on trailing-twelve-month SF MLS medians through July 2026. A blended all-types figure runs about $1,489,000, but the split is the real story and this page does not lead with a blend.

Is San Francisco real estate recovering?

Yes, selectively. Houses are in a strong seller’s market at about 1.05 months of supply, office vacancy fell to 29.7% from 34.7% on AI-driven leasing, and the population returned to growth at 826,079. Downtown condos remain soft, trading near 2015 prices.

Why are San Francisco condos relatively cheap?

Remote work, uncertainty about downtown’s recovery, HOA dues of $400 to $1,200 a month, and a wave of high-rise supply delivered just before the pandemic. Unlike Florida, this is demand-side softness, not a reserve-law or insurance shock.

How much income do I need to buy a house in San Francisco?

Our engine computes about $495,968 a year for the $1,850,000 median house at 20% down under a 28% front-end ratio. Third-party estimates run $443,979 (Redfin, March 2026) and just over $400,000 (California Association of REALTORS), against a median household income near $140,970.

What is the property tax rate for a new buyer in San Francisco?

1.18268325% for fiscal year 2025-26, applied to your full purchase price, because Proposition 13 resets assessed value when a property sells. That is the 1% Proposition 13 base plus about 0.18% of voter-approved bond debt.

Does Proposition 13 help a new buyer?

Not at purchase. You pay the full rate on the price you paid, while a long-tenured neighbor pays on a decades-old basis. It helps only after you own, by capping your assessed-value growth at 2% a year for as long as you hold.

Is it cheaper to rent or buy in San Francisco?

Renting, by the widest margin in this series. Against the $5,700 two-bedroom rent, owning the median house costs about $7,414 more per month before earthquake coverage. The price-to-rent ratio near 27 is well past the 25 line that conventionally favors renting.

Do I need a jumbo loan to buy in San Francisco?

Usually, for a house. The 2026 high-cost conforming limit for San Francisco County is $1,249,125, well below the $1,850,000 single-family median, so most house purchases are jumbo. Many condos fit within high-balance conforming.

Is earthquake insurance required in San Francisco?

No. It is excluded from standard homeowners policies and is not lender-required, which is why only about 12% to 15% of California homeowners carry it. A policy for a home in the $1.5 to $2 million range runs roughly $6,000 to $9,000 a year at the 15% minimum deductible.

What are San Francisco rents in 2026?

Records. The median one-bedroom crossed $4,000 for the first time to reach $4,060 and the median two-bedroom reached $5,700, the highest in the nation, both in June 2026, on about 22% annual growth, the fastest of any US city.

What is the San Francisco transfer tax?

A six-tier schedule from 0.50% to 6.00%, customarily paid by the seller on residential sales. At both example prices on this page the rate is about 0.75%. Critically, a tier’s rate applies to the entire price rather than marginally, so crossing a threshold is a cliff.

Will San Francisco home prices rise in 2026?

House prices are rising on tight supply and AI-driven demand, and downtown condos are stabilizing. That is a description of current direction rather than a forecast, and the Case-Shiller repeat-sales index for the metro actually edged down slightly in early 2026 while city medians rose on mix.

Methodology

Where the numbers come from. Every market figure on this page is transcribed from a named source with its geography, period, and confidence level in the source registry below, and is never presented without its period. Where sources differ, a city median and a five-county repeat-sales index, a broker supply figure and another broker’s, we show them with their scopes rather than averaging them.

Why the City and County of San Francisco. It is the cleanest anchor in this series because the city and county are the same jurisdiction, and search intent for this market is city-scoped. This follows the Chicago city-first and Miami county-first precedents. SF MLS districts appear as sub-lenses; the five-county Case-Shiller MSA appears only where those indexes do, labeled as a different scope and methodology; Oakland, the East Bay, and San Jose are their own markets and are referenced only as contrast.

Why two worked examples and two Balance Score bands. Houses and condos diverge far enough here that one blended example or one blended band would describe neither. We run the engine twice and publish two bands, following the pattern the Miami study established. Both bands drop the price-cut input, because San Francisco publishes none by property type, and renormalize the remaining four weights per the Austin precedent. The condo band reads seller-favorable on citywide inputs even though downtown is soft, and the card says so directly: that is a scope limitation of a citywide index, and the district-level figures are how this page tells the real condo story.

What we computed ourselves. Monthly payments, PITI, income needed, cash to close, the escrow split, and the rent-versus-own gaps are computed by the FinExplained calculator engine (decimal-precise, tested) from the stated assumptions, for both examples. PMI below 20% down is disclosed and excluded. Golden tests pin every published figure for both examples and both bands.

The jumbo disclosure. The single-family example exceeds the $1,249,125 county conforming ceiling and therefore requires jumbo financing in practice. Our engine computes standard amortization, and we disclose the difference rather than modeling jumbo overlays we cannot source. The condo example sits below the ceiling.

The transfer tax is seller-side. San Francisco’s transfer tax is customarily paid by the seller on residential sales, so it is rendered as seller-side context and is never included in a buyer’s cash to close, the same treatment the Dallas-Fort Worth study gives a seller-paid line. The BUILD Act is a proposal and is never presented as enacted.

Labeled estimates and derived figures. The earthquake premium ($7,500) is the estimated midpoint of a sourced $6,000-to-$9,000 band and carries low confidence. The condo HO-6 insurance figure ($1,500) is a conservative labeled estimate because no exact San Francisco figure is published. The buyer closing-cost band (1% to 2%) is a labeled estimate. The median household income is a secondary ACS-based estimate. The price-to-rent ratio is derived from the trailing median and the two-bedroom rent. All four are on our backlog to firm.

Charts. Every chart states its geography and period and carries a text description. Two chart ideas were deliberately not built for lack of sourced series: a year-by-year condo price series (the registry supplies one monthly index reading, not anchors, so the condo chart is a statement treatment instead) and a long-run city median series. San Francisco stays out of our cross-city tax-compare chart family, as Atlanta, Charlotte, Dallas-Fort Worth, and Miami do, because its tax story is the Proposition 13 reset mechanic rather than the rate level, and it gets bespoke treatment.

Source registry

Every figure used on this page, with value, geography, period, source, and confidence:

Assumptions and limitations

• The single-family example assumes a $1,850,000 home; the condo example $1,185,000. Both assume 20% down unless stated, a 30-year fixed at 6.58%, the 1.18268325% new-buyer secured rate on the full purchase price, 1% annual upkeep, and buyer closing costs of 1% to 2%. Insurance is a labeled $3,800 estimate on the house and $1,500 on the condo; the house adds $7,500 of estimated earthquake coverage and the condo $9,600 of HOA dues.
• The single-family example requires JUMBO financing (above the $1,249,125 county conforming ceiling). The engine computes standard amortization and the difference is disclosed rather than modeled.
• Proposition 13 resets assessed value to the purchase price on sale and then caps growth at 2% a year, so the modeled rate is the new buyer’s actual basis. Proposition 19 portability is stated factually, with no claim about its effect on turnover.
• The transfer tax is customarily SELLER-paid and is excluded from buyer cash to close. The BUILD Act is a proposal, not law.
• The earthquake premium ($7,500) is a Low-confidence estimated midpoint of a sourced band, shown as an optional layer because earthquake coverage is not lender-required. The condo HO-6 figure ($1,500) and the buyer closing-cost band are labeled estimates. Firming all three is on our backlog.
• The median household income ($140,970) is a secondary ACS-based estimate; replacing it with the exact Census ACS 2024 1-year figure is on our backlog.
• Months of supply and days on market are broker-sourced (Ascendre and Legacy) because San Francisco has no single authoritative publisher for them, and the single-family inventory input uses the midpoint of a sourced -37% to -45% range. Their Medium confidence carries through to the registry.
• Neither Balance Score band uses a price-cut share, because none is published by property type; the input is dropped and its weight renormalized, disclosed on both cards. Two of the single-family band’s inputs clamp at their normalizer ceilings, so that band is a floor rather than a ceiling.
• The rent-versus-own comparison uses the $5,700 two-bedroom rent for both examples. A median house is larger than a two-bedroom, so the gap is conservative rather than exaggerated.
• The Q2 +22.2% figure and the spring monthly peaks are MIX-SENSITIVE and labeled as such; the engine uses the trailing-twelve-month median. Case-Shiller is a five-county repeat-sales index on a different scope and moved the other way in early 2026; the two measures are never blended.
• Downtown condo weakness is described as soft or discounted, never distressed, because no forced-sale data supports the harsher word. Submarket figures are directional.
• Everything here is educational analysis of market data, not financial, investment, tax, or legal advice, and not a recommendation to buy, sell, or rent any property.

Data freshness

This edition carries SF MLS trailing-twelve-month closings through July 2026, Q2 2026 office data, June 2026 rents, and a July 23, 2026 mortgage rate, was published July 15, 2026, and refreshes monthly: the next update is planned for late August 2026 once the July SF MLS and Compass reports close. A notable rate move, a change in the AI leasing trajectory, passage of the BUILD Act, or a swing greater than 10% in inventory or price triggers an off-cycle update. Corrections follow our corrections policy and are logged in the changelog.