Are autonomous vehicles and San Francisco development on a collision course? 

“Oh, give me land, lots of land, under starry skies above
Don’t fence me in.” — Cole Porter and Robert Fletcher

Listen: Fifth and Brannan has come unstuck in time. 

The first thing you notice in the gap between the opaque black fencing and the barbed wire are the oscillating lights atop scores of idle autonomous vehicles.

As if by an occult hand, one ghost car after another is remotely summoned and glides to the exit. It is quickly replaced by a driverless vehicle returning to the Fifth and Brannan streets staging yard after a foray into the outside world.

Surely this is a vision of San Francisco’s future: Boxy Ojais and squat Jaguars — the X-Wings and Y-Wings of the Waymo fleet — silently coasting hither and yon to squire San Franciscans about at the touch of a button. On this day, there may have been 50 of them in the Fifth and Brannan staging area, but there was room for many more. 

And yet, this was also a vision of San Francisco’s past and, specifically, SoMa’s past: A vast sea of open lots and, specifically, parking lots. 

Viewers watching the stunningly low-quality 1970s Blaxploitation film “Dolemite” laugh and point the first time they see a visible boom microphone in the shot. It becomes less novel when you realize there’ll be a boom microphone in most shots.

Autonomous vehicles are in many or most shots here in San Francisco. That’s why visitors to San Francisco stop and gawk at them, but locals don’t. 

For the most part, San Franciscans think critically about autonomous vehicles in terms of how they interact on the road with other vehicles, first responders or, in one unfortunate case, a kitty cat. Less consideration, naturally, is given to what the ghost cars are doing when they’re out of sight, out of mind and not moving around. 

But the people in charge of deploying these technological marvels thought a lot about this, and long ago. And it wasn’t high-level technological marvel thinking: It was old-school business and real estate. 

Almost a decade ago, Molly Turner, a lecturer at the University of California, Berkeley’s Haas School of Business specializing in urban technology, had her first conversation with a higher-up in the world of autonomous vehicles.

“He told me then that his No. 1 priority was finding real estate to park his vehicles on for charging,” she recalls. “And this was before Waymo was really deployed anywhere. Autonomous vehicle fleets need places to park and charge them.” 

But not just to charge them: They also need to be maintained. And cleaned: People do strange and terrible things in autonomous vehicles when there’s no human being to tell you not to — though, rest assured, you are on camera, and your licentiousness is being viewed by someone, somewhere.

Or your slovenliness: Riders may be less than considerate about where they put their trash. Or they may puke. 

Regardless, autonomous vehicles need to be serviced and cleaned by autonomous humans before heading back out to service, and that requires the car to not be moving. So: garages or parking lots. 

But where autonomous vehicle companies would want to put those garages and parking lots and where the city would ostensibly want them to go aren’t necessarily compatible in the long term.

Mission Local has obtained a list of several dozen sites where AV companies put their cars, have attempted to put their cars, or formerly put their cars. These are, by and large, sited near the people the companies would want to serve — which is, business-wise, a great idea. 

But these also tend to be sites where the city may have greater land-use aspirations than garages or parking lots.

“It’s the core,” answers a veteran San Francisco real-estate broker when asked where autonomous vehicle companies want to site their garages and parking lots. “Everyone wants to be in the core. That’s where all the rides are. And, as adoption moves out to the neighborhoods, they’ll go there.”

Nobody ever accused the trillion-dollar companies backing robotaxis of being bad at business — and keep the autonomous vehicles near the people who use autonomous vehicles would seem to be good business theory.

And it’s not just theory, it’s practice: Terrawatt, a company that has made electric-vehicle-fleet charging its raison d’être, has adopted this as a core component of its business model.

“We survey each market to find the strongest overlap of power availability, proper zoning and customer operating needs,” it states on a portion of its site titled “Where rideshare customer demand is highest.” 

If you take a glance at the map of where San Francisco’s autonomous vehicle fleet parking sites are, were or aspire to be, patterns emerge. Only one of the 30 sites is west of Gough. And while a handful, including some of the big ones, are in the city’s less populated and less affluent industrial southeast, the lion’s share are in the crowded northeast, where rideshare customer demand is highest. 

In the abstract, locating a service near the people who desire that service is not a revelatory concept. But, in the concrete — literally and figuratively — locating parking garages and surface lots in populated areas where housing could be developed did surprise land-use professionals and urban studies experts. And not just in San Francisco.  

University of Oregon professors Nico Larco and Robert Binder, two of the authors of a recent study analyzing the potential impact autonomous vehicle expansion may have on parking and development, both said they were initially surprised to see large robotaxi depots in populated, upscale and central Los Angeles neighborhoods. 

“What were these sites zoned for? My guess is they weren’t zoned for light industry,” Binder says. “These are not industrial areas. They are locating themselves in an area where there are customers.” 

Putting a product within easy reach of customers — and minimizing the time vehicles are driving in from industrial zones in the sticks, empty, and not earning revenue — is pretty intuitive business practice.

But putting parking lots in areas zoned for housing and customer-serving businesses is not intuitive city planning practice. Not anymore at least: Downtown Houston in the 1970s is nobody’s role model for ideal urban land-use and transportation.

But here’s where the long-term wishes of autonomous vehicle companies and the city they’ve made their testing ground ostensibly diverge. San Francisco can’t grow and evolve without building more housing. And autonomous vehicle companies can’t grow and evolve without more parking. 

Are autonomous vehicle companies impeding San Francisco housing and development? It’s a question that’s been asked before. The fairest answer is: No, not yet. 

There are many reasons that development in this city is slow and difficult. There are many reasons — global, national and local — that development is presently not booming. It would be a stretch to say that, at the moment, autonomous vehicle parking, charging and staging sites are what’s keeping buildings from being built. 

But prosperity, they say, is around the corner. And, sooner or later, the numbers will make financial sense for housing and development on sites presently occupied by herds of ghost cars. Many of these sites are already entitled for significant housing and development. 

On sites rented on a month-to-month basis in which little infrastructure has been installed, there are fewer impediments to breaking ground for housing when the time is right. But on charging sites — even if the climate becomes ideal to build, build, build — development is less likely to occur.

This is because the upfront investment to install chargers and other infrastructure to service a fleet of electric vehicles is significant.

The Chronicle in 2025 reported that a single supercharging station can cost $400,000. An electrical contractor this month told us that it could run $250,000. That’s before you factor in the cost of additional electrical infrastructure and the electric itself — and fleet-charging sites can have scads of these chargers.  

As such, the industry standard for fleet-charging stations is to lock in a long-term, turnkey lease.

“It’s a pain in the ass to get power from PG&E, and it’s a pain in the ass to build this stuff,” says the real-estate broker. “You’re not spending millions of dollars without a long-term lease. You have to amortize that cost.”

And even on robotaxi sites without significant infrastructure, there may be another impediment to future development — guaranteed rent money to the site owner, reliably paid by the richest companies on earth.

“If you could rent out your land to an AV company and you’re getting a six percent return on land value, why would you sell your land?” asks the broker. “You’d have to get a lot more.” 

If nothing else, getting into a bidding war with the world’s wealthiest companies would drive up the cost of land even more. 

You may ask yourself: But wait! Weren’t autonomous vehicles supposed to do away with the need for parking? 

In theory, yes. In practice: Yes and no. Larco, the Oregon professor who studied the interaction between robotaxis and parking, says that one autonomous vehicle cruising around all day could reduce the need for parking by 15 to 18 spots. But here’s the rub: Those are scattered spots, often in outer neighborhoods or towns.

Meanwhile, autonomous vehicles require middling to large concentrated parking and staging sites, which the companies would prefer to see in central, high-traffic areas. Parking spaces, a city planner tells us, “are not going anywhere. They’re just being repurposed.” 

And if autonomous vehicles mushroom in San Francisco and a far, far, far higher percentage of trips in this city are undertaken by driverless cars, would scattered parking sites then be converted into housing and development?

Larco, Binder and graduate student Matthew Channing studied exactly that question. And their answer is: There would be less development than you’d expect or hope for. 

In Los Angeles, which looks a lot more like Houston in the ’70s than San Francisco, they found that even if robotaxi use exploded, only 3.8 percent of the parcels in L.A. would be more suitable for development.

In San Francisco, they found, the percentage would be even smaller. You need large parking lots to build significant developments and, here, that’s pretty much Stonestown. But, they found, it would still only pencil out to develop a portion of Stonestown’s vast parking lot — say, one-fifth. And that’s if robotaxi use takes off to infinity and beyond. 

This is something to think about while watching the futuristic driverless cars arrive and depart from the retrograde full-city-block surface parking lot on Fifth and Brannan. 

Perhaps someday cars will fold up into a briefcase, like George Jetson’s. In the meantime, San Francisco’s future and past will continue to duke it out, one parking lot at a time.