The Price of a Mile

Kris Van Meter · July 31, 2026 · Real Estate

By Kris Van Meter | The Upland Group July 31, 2026 Real estate · Decentralization

People keep discovering decentralization like it is a new app category.

This week it showed up as autonomous miles at commercial scale. Last month it was open weights versus closed models, and who rents the rack. On the money side it is settlement rails versus correspondent fog. Same pattern in different clothes. Scarcity piles up in the center. The center overfills. Technology and capital open the edge. People and industry move out. Then the center gets interesting again, and the cycle breathes.

Real estate is where that breath becomes a deed.

Gather, crowd, leave, return

Before cities, people followed food. Settlement stuck when agriculture made a place worth defending. Then density paid: markets, walls, specialization, ports. Then density hurt: disease, fire, rent, crime, commute. Then something new (streetcars, cars, highways, broadband) made distance cheaper, and the edge filled. Then the edge got expensive or boring, and a cohort came back for amenities and short walks. Then the core got expensive again, and families looked outward for bedrooms and yards.

That is not a morality play about suburbs versus downtown. It is a cost curve on distance.

Postwar America wrote the loudest chapter with FHA mortgages, interstates, and mass car ownership. Central-city share of metro population fell for decades while suburbs absorbed most growth. Later waves brought “back to the city” among young college-educated adults, then a suburban and exurban reassertion as housing costs rose and millennials hit family formation. COVID did not invent remote work. It removed the last excuse for pretending every knowledge job needed a badge swipe at 8:55 a.m.

Your housing book already lived on that map: Battery Belt plants, ports, corridors, SFR where jobs and land still pencil. Self-driving was always on the list next to Starlink and power. What changed in 2025–2026 is not the thesis. It is the evidence that autonomous miles left the pilot phase.

The commute is a housing input

Census ACS data put mean one-way travel time to work around 27 minutes recently (about 27.2 minutes in 2024 for workers who do not work from home). Round trip, two hundred fifty commuting days, and you are near 225 hours a year. Call it five to six full work weeks staring at brake lights or gripping a wheel.

Those hours are not just suffering. They are a shadow rent on where you can live. A household pays for housing and pays again for the minutes between home and work. Raise the effective cost of those minutes (traffic, gas, parking, wreck risk) and the viable map shrinks toward the job. Lower that cost (train, remote work, cheaper safer rides) and the map expands.

Autonomous driving attacks the commute on two fronts:

  1. Time quality. Even when the body is still in a seat, cognition can return if the vehicle is truly driverless and the passenger trusts it. That is not the same as “free hours,” but it is not full cognitive lock either.
  2. Risk and money. Human driving kills tens of thousands of Americans a year. NHTSA’s early full-year estimate for 2025 is about 36,640 traffic deaths, down from 39,254 final in 2024. Waymo’s peer-reviewed rider-only analyses (e.g. Kusano et al., Traffic Injury Prevention, 2025, on tens of millions of driverless miles) and independent IIHS work (2026) report large reductions in injury and police-reportable crash rates versus human benchmarks in the cities studied. Methods and geo-mix matter. The direction is hard to ignore: software is compressing one of the largest accidental death risks in American life, which is also an insurance and household-budget story.

When the price and risk of a mile fall, housing demand does not vanish. It relocates.

The metric: price of a mile

Fleet operators and research shops will fight forever over exact all-in costs. The useful discipline is simpler.

A personally owned car, fully loaded (depreciation, insurance, fuel or energy, maintenance, parking), still lands near about a dollar a mile in common consumer all-in estimates. Robotaxi and shared autonomous operators are racing that number down as utilization rises and hardware costs fall. Public discourse (including ARK-style scenarios and fleet-owner Twitter) floats long-run targets in the low tens of cents per mile if utilization and regulation cooperate. Treat those as scenarios, not a Propwell underwriting cell.

What you underwrite is the direction and the gate:

Same logic as compute: rent the scarce hour. There it is a token or a GPU-second. Here it is a mile. Utilization is the difference between a science project and a business.

What commercial scale actually looks like (without the PDF)

You do not need a ninety-page deck to see the phase change. Public company and operator disclosures, plus mainstream research coverage, put autonomous ride-hail past pure demo theater in multiple U.S. cities, with Waymo alone reporting cumulative paid driverless rides in the tens of millions and weekly volumes in the hundreds of thousands by late 2025 into 2026. Last-mile robots are already boring on enough campuses that the interesting question is HOA and sidewalk rules, not whether the box can roll.

China’s stack is closing gaps on the same problem with different capital and regulatory physics. That matters for hardware cost curves and for the Battery Belt thesis (vehicles and batteries still get built somewhere). It does not mean every Sun Belt zip code gets robotaxi density next year.

The bottleneck everyone serious names is the same one power markets name: permission. Cities, insurance regimes, liability, and labor politics decide where fleets densify. Arizona-style regulatory clarity became a case study for a reason. Capital follows the permit as much as the algorithm.

Who gets disrupted (and who still needs a house)

When human error stops being the default assumption for every mile:

Your older strategy docs already said the quiet part: purpose-built rentals, flexible plans, energy, less car worship over time. AV is the transport leg of that sentence. Starlink was the work leg. Power is the live leg.

Decentralization is the theme because the cycle is loud

This week’s stack is not coincidence.

AI decentralizes cognition to the edge when open weights and local hosts make sense, then recentralizes training into power-dense halls. Blockchain fights over who settles value without a correspondent middle. Housing fights over who can afford the center and who takes the edge when minutes get cheaper.

Cities overcrowd. People leave. The edge fills. The center reinvents. People return until rent and chaos push them out again. Farmers settled after gatherers. Streetcar suburbs followed tenements. Levittown followed the war. Zoom towns followed the pandemic. Robotaxi density will write another chapter for the metros that allow it.

The mistake is treating any single chapter as the end of history. The center does not die. The edge is not a utopia. Capital wants the deed where the next binding constraint is loosest for the household it serves.

For a Southeast SFR book, that still looks like:

Autonomous miles improve the last line in cities where they are real. They do not invent a job in a county with no employers.

How to underwrite without LARPing the future

  1. Score mobility by MSA, not by Twitter. Owned car default, transit, human ride-hail, robotaxi legal density, truck ADAS/autonomy corridor exposure.
  2. Keep job gravity primary. Battery Belt and port housing are demand from paychecks. AV is how paychecks travel.
  3. Watch $/mile and insurance the way you watch rate and taxes. They move household capacity.
  4. Design optionality. Plans that work with a car and without a second car age better than plans that require three SUVs of garage.
  5. Ignore $11 trillion headlines in the acquisition model. Use them as a reminder that transport is a large industrial market with long duration, not as a cap rate input.
  6. Separate fleet equity fantasies from land. Owning robotaxis is a different business than owning houses near the people who still show up to work.

Close

The gatherer followed the herd. The farmer stayed. The city crowded. The suburb sprawled. The laptop untethered the desk. The autonomous mile untethers the hands on the wheel.

None of that makes real estate abstract. It makes location a function of the price of connection: to work, to goods, to other people. When connection gets safer and cheaper, maps redraw. When the center overcharges for the privilege of density, the edge fills. When the edge gets thin on jobs or culture, the center calls people back.

We are in another redraw. The operators who win will not be the ones who bought a robotaxi narrative. They will be the ones who kept buying bedrooms next to real economic engines, while the price of a mile quietly changed what “next to” means.

Sources

Not investment advice. Scenario framing for operators and allocators.

Operator seat, not a research desk

If you underwrite housing against jobs, logistics, and mobility and want a sharper conversation, start here.

Contact