Brian Potter's Blog, page 2

Reading List 08/02/2025

Goddess of Kannon statue, Japan, via @sci_fi_infra.

Welcome to the reading list, a weekly roundup of news and links related to buildings, infrastructure, and industrial technology. This week we look at European multifamily construction, a robot lamp, China’s self-driving car test, a “mini Moravec’s paradox”, and more. Roughly 2/3rds of the reading list is paywalled, so for full access become a paid subscriber.

No essay this week, but I’m working on a more involved piece about the nature of technological progress that should be out next week.

Housing monopolies

A line of argument I haven’t been paying much attention to but has (apparently) been gaining traction is the idea that recent increases in housing costs are driven by monopolies in homebuilding. This theory has been advanced by folks like Matt Stoller, but evidently it's gaining currency with others as well.

This theory didn’t seem completely crazy to me, as over the past several years homebuilding has steadily gotten more concentrated. It used to be that the 10 largest homebuilders were responsible for a very small fraction of total homes built in the US, but today they’re responsible for an increasingly large share. But I had never really looked into the idea.

However, a good piece by Derek Thompson on his substack lets some air out of this theory:

It is a fact that home prices across the country have increased in the last decade. It is also a fact that homebuilders have gotten bigger in the last few years. The question at stake here is: Is the second fact causing the first fact?

I believe there is only one economic paper that unambiguously says yes. It is a 2023 working paper entitled “Fewer players, fewer homes: concentration and the new dynamics of housing supply” by the economist Luis Quintero. This research is load-bearing in the antitrust world. It’s quoted in articles and social media posts by the antitrust advocate Matt Stoller. You’ll find it in research from other antitrust scholars. And it seems to be the only empirical research in Musharbash’s essay.

I called Luis Quintero to ask what level of market concentration in homebuilding he considered to be dangerous. In the most concentrated markets, Quintero said, one or two firms account for 90 percent of new housing. But problems begin to accelerate, he said, if five or six firms account for 90 percent of new housing.

…I tracked down a complete listing of the country’s 50 largest homebuilding markets, from #1 Dallas to #50 Cincinnati. How many meet Quintero’s first oligopoly threshold (two companies = 90 percent of the market)? Zero out of 50. And how many meet his second threshold (six companies = 90 percent of the market)? One: Cincinnati. It turns out that the largest homebuilding markets just aren’t that concentrated, even when you accept the yardstick of the economist who’s offered the only empirical definition of market concentration in homebuilding that I can find.

European multifamily construction

It’s become clear that for many types of construction, European countries have much lower construction costs than the US does. Transit and subway construction is the most famous example of this, but it's my sense that in multifamily apartment construction, Europe is also often substantially cheaper.

With transit construction, some of this comes down to improved efficiency and lower labor costs, but some of it comes down to differences in what gets built: US transit apparently often has much larger, more expensive stations than European transit. It seems like something similar might be true for multifamily construction as well. On Twitter, developer Brendan Whitsitt looks at a recent multifamily project in Paris, and lays out the many differences which allow it to be built for less than a similar project in Canada (which I believe will be fairly similar to US requirements.)

Robot lamp

I’ve spent a bit of time discussing humanoid robots, and considering whether a humanoid form factor will win out, or if other robot form factors might make more sense. In that vein, here’s a company developing a robotic floor lamp, Lume, that will (supposedly) autonomously fold clothes for you when it's released in 2026:

It’s not clear how close they are to an actual working robot (the video is entirely simulated, and I’d be surprised if they end up using a form factor that has so many joints/degrees of freedom), but if this seems compelling, you can place a reservation for the lamp here.

Elon Musk’s self driving car predictions

Tesla has been working on self-driving cars since at least 2013, and has been shipping Autopilot hardware since at least 2014. Over the years, Elon Musk has frequently made predictions about when various self-driving capabilities will be available, which have reliably proved to be overly optimistic. Wikipedia has a helpful list of every prediction Elon Musk has made around self-driving cars, and the status of that prediction.

The list includes 29 predictions in total. Of these, 1 has been met, 3 have been partially met, 7 might still be met but haven’t yet, and 18 haven’t been met.

Reading List 07/26/25

Central control building for Kalyon Karapinar solar farm, Turkey, via @sci_fi_infra.

Welcome to the reading list, a weekly roundup of news and links related to buildings, infrastructure, and industrial technology. This week we look at an important FAA rule change, construction microfactories, recycling data center waste heat, the rise and fall of MATLAB, and more. Roughly 2/3rds of the reading list is paywalled, so for full access become a paid subscriber.

FAA MOSAIC rule change

The Federal Aviation Administration (FAA) regulates aircraft in the US, and those regulations exist on a continuum of strictness depending on the type of aircraft and the passengers it will carry. On the ultra-strict end, you have regulations for commercial airliners (Part 25 of Title 14 of the Code of Federal Regulations), which are designed to be strict enough to limit catastrophic events to 1 every billion flight hours. On the less-strict end, you have things like experimental aircraft and ultralights, which don’t need to be built to any particular set of standards (and in the case of ultralights don’t even require a pilots license to fly).

In the middle of the regulation continuum is a category called Light Sport Aircraft. Aircraft in this category have more regulation than experimental aircraft (they must be built to consensus standards), but less than commercial aircraft or general aviation aircraft. The FAA recently updated its rules around light sport aircraft (known as Modernization Of Special Airworthiness Certification or MOSAIC) to greatly expand what sorts of aircraft fall under this category. Eli Dourado, in a post on the proposed rule change from 2023, explains why this is a big deal:

After World War II, expectations for general aviation were sky high. The war had driven significant advancements in aeronautics, navigation, and communication technologies. There were a large number of military surplus planes and a lot of trained pilots returning from the war. For someone alive in 1945, who had perhaps witnessed the social transformation wrought by the automobile, it would seem only natural that general aviation would be the next step in personal transportation.

Of course, it didn’t play out that way. The government significantly tightened certification requirements for general aviation in 1945 and 1965. With the market flooded with cheap, surplus military aircraft, it was hard for manufacturers to make money investing in new, more advanced models. And, of course, piloting an airplane using conventional flight controls is not as simple as driving a car, requiring significantly more skill, with lapses in attention more catastrophic.

If general aviation is to make a comeback, something like the MOSAIC rule is a prerequisite. By making a category of aircraft that doesn’t require type certification, is actually useful for transportation (250 knots and 4 seats), and can be flown with less skill via simplified flight controls, FAA is opening the door to a bigger market and vastly more innovation.

Without the need for type certification, manufacturers can iterate on their designs more rapidly without going through the costly supplemental type certification process. They can include cheaper uncertified avionics. They can do over-the-air software updates.

Meanwhile, simplified flight controls and making LSAs actually useful could greatly increase demand for these aircraft as transportation. That increased demand feeds directly into manufacturing investment and the pace of iteration. The more progress there is in making these small planes really great, the more the demand for them will increase, creating a flywheel effect.

If we can get LSAs into mass manufacturing, production costs of the airframe could go down further…

World’s biggest dam

Today the world’s biggest dam by hydroelectric capacity, and the biggest power station in the world is the Three Gorges Dam in China, which can generate 22.5 GW of electric power. (By comparison, a large nuclear reactor can generate around 1 GW of electric power, and the Hoover Dam has around 2 GW of generation capacity.) Now China has started construction on an even bigger hydroelectric dam in Tibet, the Motuo Hydropower Station. When complete, it will have an estimated capacity of 60 GW of electric power. Via the BBC:

Chinese authorities have begun constructing what will be the world's largest hydropower dam in Tibetan territory, in a project that has sparked concerns from India and Bangladesh.

Chinese Premier Li Qiang presided over a ceremony marking the start of construction on the Yarlung Tsangpo river on Saturday, according to local media.

The river flows through the Tibetan plateau. The project has attracted criticism for its potential impact on millions of Indians and Bangladeshis living downriver, as well as the surrounding environment and local Tibetans.

Beijing says the scheme, costing an estimated 1.2tn yuan ($167bn; £125bn), will prioritise ecological protection and boost local prosperity.

When completed, the project - also known as the Motuo Hydropower Station - will overtake the Three Gorges dam as the world's largest, and could generate three times more energy.

Construction microfactories

An idea that is becoming more and more popular with construction startups is the use of microfactories. The idea is, rather than using a large, expensive, conventional factory to produce buildings, you can use flexible manufacturing technology (robots, CNC machines, etc.) within a small footprint factory that can be quickly setup and torn down near the jobsite. The hope is that this will let you get many of the benefits of factory production (automation and improved efficiency of previously labor-intensive operations), without many of the drawbacks (transportation costs, expensive factory overheads). On Thesis Driven, Nick Durham gives an overview of the companies that are experimenting with microfactories:

A shippable microfactory is a compact, self-contained production unit designed for deployment on or near the site of building construction. It combines automation, modular equipment, and digital coordination to enable localized, on-demand manufacturing. It can perform tasks like cutting, framing, and assembling parts using software-driven automation.

Each project’s digital design feeds directly into the microfactory’s machines, enabling components to be custom-fabricated on demand. The key distinction from a traditional prefab plant is scale and mobility: a conventional facility is large, stationary, and optimized for high throughput (but comes with high overhead), whereas a microfactory is small, movable, and deployed per project.

By trading size for agility, microfactories deliver many of the same benefits – precision, speed, controlled environment—in a temporary, on-site setting. Once a project is finished, the microfactory can be packed up and shipped out for the next job. Portability is the real breakthrough. In a sense, the factory itself becomes a productized tool that a builder can deploy as needed, rather than a permanent asset that must be kept busy at all times.

Wood Mackenzie on Big Beautiful Bill energy impacts

Energy consulting company Wood Mackenzie has an analysis of how the tax credit changes in the Big Beautiful Bill will affect energy infrastructure construction. Wind and solar buildout is projected to be reduced, as is energy storage due to rules about foreign sourcing. Oil, gas, and coal all benefit:

The balance of incentives in the US tax system was tilted strongly in favour of low-carbon energy by the IRA. It has now been levelled out.

Wood Mackenzie analysts have published a detailed assessment of the impact of the new legislation on every energy sector. The big picture is that we expect investment in wind and solar power to fall well short of what it would have been if the IRA incentives had remained in place.

It will not collapse completely. Investment in renewables will be supported by growing electricity demand, state and corporate policies, challenges to the growth of gas-fired generation, and a levelised cost of electricity (LCOE) that is highly competitive in many places.

But over 2025-2030, wind and solar installations will be about 100 gigawatts lower than if the IRA incentives had remained in place, Wood Mackenzie analysts predict. That would lead to total installed wind and solar capacity growing by about 25% over that period, instead of a projected 55% growth under the IRA framework.

The impact will vary widely across technologies and across sectors. For solar, we expect a surge in installations in 2025-2026, as developers rush to remain eligible for tax credits. But on a 10-year view, we expect installations to be 17% lower than in our previous base case forecast.

Where are Vacation Homes Located in the US?

As of 2023, the US has around 142.3 million housing units: roughly one home for every 2.4 people in the country. The vast majority of these homes – 127.5 million – are occupied. The remaining 14.8 million homes are vacant. Of these, around 4.8 million homes, or around 3.5% of the total, are vacant because they’re seasonal, or vacation, homes.

I’ve spent a lot of time writing about patterns of housing and home construction in the US, but virtually none of it has been looking at vacation homes specifically. I wanted to better understand patterns in the distribution of vacation homes in the US, for a few different reasons:

It tells us about patterns of recreation in the US: where are the popular vacation spots?

It tells us about patterns of elite consumption in the US: where do people who can afford it have second homes?

It points us to places that might have unique housing challenges due to very high demand from vacationers and tourists.

To look at US vacation homes, I used Census data, which tracks (to varying degrees of success, depending on the census year) the number of seasonal homes in the US going back to the 1940s.1

Vacation homes in the US

The map below shows vacation homes in the US by state.

Florida leads the pack with over 800k vacation homes, followed by California at 360k and New York at 309k. Michigan (at 260k) and Texas (at 201k) round out the top five. At the bottom of the list, you have several Great Plains states: Kansas at 12.9k, South Dakota at 13.7k, and North Dakota at 11.6k.

There’s a few interesting patterns observable on this map — the dominance of Florida, the relative popularity of Midwest states like Michican and Wisconsin as a place for vacation homes, the fact that Alaska has nearly as many vacation homes as Hawaii — but to a large extent this map is a population map. 4 of the top 5 states for vacation homes are also in the top 5 by overall population (only Michigan breaks this trend, coming in at number 10 in overall population). To account for this, we can look at the fraction of homes in each state that are vacation homes.

By percentages, the pattern changes somewhat. Florida still ranks highly, with over 8% vacation homes, but it's soundly beaten by the New England states of Maine, New Hampshire, and Vermont, which have 15.6%, 10.2%, and 15.1% vacation homes respectively. Even Delaware (8.6%) and Alaska (8.9%) have a greater fraction of vacation homes than Florida does. Michigan (5.7%) is now closer to the middle of the pack, along with places like Idaho (6.2%) and Arizona (5.5%). Great Plains states are still near the bottom (Kansas at 1%, Nebraska at 1.8%), but the state with the lowest fraction of vacation homes is actually Illinois at 0.8%.

At a high level, vacation homes are common in a few geographic regions: New England (Maine, New Hampshire, Vermont), the upper Midwest (Minnesota, Wisconsin, Michigan), the Southeast (Florida, the Carolinas), and the Western states between the Great Plains and the West Coast (Wyoming, Idaho, Utah, Arizona).

How have these patterns changed over time? The chart below shows the total number of vacation homes by state and year.

And this chart shows the fraction of homes that are seasonal by state over time:

Some states have always been popular places to have vacation homes: Vermont, New Hampshire, and Maine have led the pack continuously since the 1940s. Some states have gotten more popular over time. Florida and Arizona increased in popularity substantially between 1960 and 1990, presumably due to the spread of air conditioning (and the construction of places like Disney World). Western states like Montana, Idaho, and Wyoming also get more popular over time.

And a few states have gone down in popularity. New Jersey, Rhode Island, and Connecticut have all fallen in the fraction of homes that are vacation homes. But most states have stayed at relatively similar fractions of vacation homes over time, outside of a broad trend of vacation homes getting more popular in the US as a whole and percentages gradually ticking up everywhere.

Vacation homes by census tract

State-level data can reveal broad patterns in seasonal homes in the US, but we can get a better sense of where specifically the popular vacation spots are by looking at more granular data. The map below shows every census tract in the US where vacation homes are 20% or more of the total number of housing units. This is a small fraction of overall census tracts — 3,372 out of over 85,000 census tracts in the US, or around 4% of the total. But 53% of all vacation homes in the US are in such tracts.2 (Almost 20% of vacation homes are in census tracts where more than 50% of the homes are vacation homes.)

(click to enlarge)

!function(){"use strict";window.addEventListener("message",(function(e){if(void 0!==e.data["datawrapper-height"]){var t=document.querySelectorAll("iframe");for(var a in e.data["datawrapper-height"])for(var r=0;r<t.length;r ){if(t[r].contentWindow===e.source)t[r].style.height=e.data["datawrapper-height"][a] "px"}}}))}();

Looking at our census tract map, we see the same patterns we saw in our state-level map — high concentrations of seasonal homes in Florida, New England, the upper Midwest, and certain Western states — but we now see much more detail. Vacation homes in Florida are concentrated on the coasts, and to a lesser extent in central Florida, near Disney and other theme parks like Universal Studios. Vacation homes in the upper Midwest states of Michigan, Wisconsin and Minnesota are concentrated in the top half of those states.

Broadly, looking at this census tract map, we see vacation homes concentrated in a few types of geographies.

On the beach

Census tracts with high fractions of vacation homes in Florida. Yellow is 20-30% vacation homes, orange is 30-40%, red is 40-50%, and dark red is > 50%. Opacity indicates population density, with higher population densities having higher opacity.

Around 1.2 million vacation homes in the US, slightly more than 1/4th of the total, are in census tracts within 5 kilometers of the ocean. Virtually the entire East and Gulf Coasts are lined with vacation homes. Interestingly, this is somewhat less true of the West Coast. There are still many coastal census tracts with large fractions of vacation homes, but there are many fewer than on the East Coast. Florida has 1.5 to 2.5 times California’s coastline (depending on how you calculate it), and around 60% of its population, but around 5.6 times as many ocean-adjacent vacation homes. In fact, both New Jersey and Massachusetts have more coastal vacation homes than California does. In the continental US, there are roughly 5 times as many vacation homes on the Atlantic coast as there are on the Pacific coast.

!function(){"use strict";window.addEventListener("message",(function(e){if(void 0!==e.data["datawrapper-height"]){var t=document.querySelectorAll("iframe");for(var a in e.data["datawrapper-height"])for(var r=0;r<t.length;r ){if(t[r].contentWindow===e.source)t[r].style.height=e.data["datawrapper-height"][a] "px"}}}))}();On the lake

Census tracts with high fractions of vacation homes in the upper midwest.

Census tracts with high fractions of vacation homes around Lake Tahoe.

There are also a large number of vacation homes around lakes both big and small in the US. There are over 160,000 vacation homes near the Great Lakes, around 30,000 near Lake Tahoe in California/Nevada, another 30,000 near Lake of the Ozarks in Missouri, and 12,000 vacation homes near Lake Champlain in New York/Vermont. Altogether there are around 280,000 vacation homes near the 50 largest lakes in the US, and 37 states have a lake, pond, or reservoir with at least 1,000 vacation homes in its vicinity.3

And outside of large lakes, states often have numerous small lakes lined with vacation homes. In large part, this explains why Minnesota, Wisconsin, and Michigan have such large numbers of seasonal homes. Each state has a large number of lakes (2nd, 6th, and 8th-most in the country, respectively), which are frequently lined with summer lakehouses. Around 1.3 million vacation homes in the US are near a lake, pond, or reservoir at least 3 km^2 in size. (A lake this size is big enough to have several hundred vacation homes around it.)

Even relatively small lakes can accommodate a lot of vacation homes. Big Star Lake in Michigan, which is around 3.64 square kilometers in size, has several hundred vacation homes around its shores.Near skiing

Census tracts with high fractions of vacation homes in Colorado. White circles are major ski resorts.

Census tracts with high fractions of vacation homes near Salt Lake City. White circles are major ski resorts.

Looking closely at the census tract map reveals that many clusters of vacation homes are near ski resorts: Aspen and Vail in Colorado, Lake Placid in New York, east of Salt Lake City in Utah. The pattern of high-vacation home census tracts overlaps nicely with the distribution of ski resorts in the US. The map below shows our census tracts with large fractions of vacation homes, along with the 200 biggest ski resorts in the US as listed by skiresort.info.4

(click to enlarge)

You can see that most large ski resorts are surrounded by census tracts with lots of vacation homes, and that many of the large regional vacation home clusters — the northeast, the west, and even the upper midwest to some extent — overlap with the location of ski resorts. Overall there are around 380,000 vacation homes in census tracts that are within 5 miles of a major ski resort.5

Outside these three major categories, you also see large numbers of vacation homes near certain other amenities. Some are located near mountains outside of ski resorts. The southern Appalachians, for instance, have large numbers of vacation homes in places like Blue Ridge, Georgia, Gatlinburg, Tennessee, and Black Mountain, North Carolina.

Census tracts with high fractions of vacation homes northwest of Great Smoky Mountains National Park.

Golf appears to be another major amenity for a lot of seasonal homes. Popular vacation spots like Hilton Head in South Carolina, Long Island in New York, the Highlands in North Carolina, and Palm Springs in California all have very large numbers of golf courses. There are almost 90,000 vacation homes near 20 of the largest golf resorts in the US.

Census tracts with high fractions of vacation homes in Palm Springs, which has 124 golf courses in the area, supposedly the largest concentration of golf courses in the world.

And finally, there are tens of thousands of seasonal homes near Walt Disney World and the central Florida theme parks. (Possibly there are other theme parks that are enough of a draw to generate the construction of vacation homes, though given how dominant Disney and central Florida are in theme park traffic in the US, this would surprise me.)

Census tracts with high fractions of vacation homes near Disney World, Florida.Major cities

You might expect major cities to be another popular place to have a vacation home; people certainly like taking vacations to big cities, after all. But while major cities do have substantial numbers of vacation homes, proportionally the numbers are surprisingly small. Only New York City has more than 50,000 vacation homes: in other words, 75% of the vacation homes in New York state are outside of NYC.6 Only three major cities in the US have more than 20,000 vacation homes: even a major tourist destination like Las Vegas only has around 19,000. Altogether, the 50 largest cities in the US only have around 350,000 vacation homes in their vicinity (fewer than the number of vacation homes around ski resorts), and the proportion of vacation homes in said cities is substantially lower than the national average (around 1%, compared to around 3.4% in the US as a whole).7

So what are my overall takeaways from this tour of the US’s vacation homes?

Popular vacation home locations tend to remain popular. Most states that have large proportions of vacation homes have maintained that status over time. The three states with the largest ratios of vacation homes, Maine, Vermont, and New Hampshire, have been the leaders since at least the 1940s.

This shouldn’t be that surprising. Places are popular sites for vacation homes in large part due to geographic amenities — the climate, the beach, the mountains — or the activities that those geographic amenities allow (golf, skiing) and those amenities typically don’t change over time. The big exception to this trend, the rise of Florida as a vacation destination (and the South more generally), can be thought of as resulting from a change that did occur in geographic amenities, in the form of air conditioning making the climate much more tolerable.

Vacation homes are highly concentrated. Most vacation homes (> 50%) are in a relatively small fraction of census tracts, and are concentrated in a few geographic regions of the country (the coasts, the Great Lakes, New England, some western states). Similarly, vacation homes tend to be clustered near similar types of amenities. A large fraction of vacation homes are near a large body of water like an ocean or a lake, or are near mountains/skiing.

This is another consequence of the geographic nature of seasonal home construction: climate/natural amenities vary by location, some areas will have much more attractive amenities than other places, and they will accumulate a disproportionate share of vacationers.

Growth in vacation homes is slower than the trend of economic growth might suggest. Between 1940 and 2010, GDP per capita in the US increased by a factor of nearly five in real terms, and the median household income increased by a factor of around 2.5 (also in real terms) between 1953 and 2020. Naively, you might expect the fraction of vacation homes to have increased at comparable rates. But in fact, the fraction of vacation homes in the US has lagged this significantly.

This is another consequence of the lack of improvement in construction productivity, and the fact that zoning rules and land use restrictions makes it so difficult to build homes. We have fewer seasonal homes than we would if construction was cheap and building homes easy, and anyone who wanted one could afford a second home.

1

The 1970 and 1980 censuses of housing have implausibly small numbers of vacation homes for reasons which aren’t clear to me. According to the 1970 census, the number of seasonal homes declined from 1.7 million in 1960 to 1 million in 1970, which I’m almost certain is an error. 1980 shows similar low numbers. I’ve excluded those years from analysis.

2

This isn’t because the population is concentrated in a small number of census tracts more generally. Census tracts are sized so they (very roughly) have similar amounts of people in each one, and distribution of population in them is not particularly asymmetric. The 8,000 most populous census tracts in the US only have around 18% of the total US population.

3

“Near” meaning “In a census tract that intersects with a body of water at least 3 square kilomters in size”, or within 5 kilometers of a body of water at least 50 square kilometers in size. Because some large census tracts will end up having vacation homes that aren’t anywhere near the body of water in question, this will include some vacation homes that aren’t actually near a lake. But it will also exclude some vacation homes that are near smaller bodies of water (or have been subdivided by the census into smaller bodies), so on balance I don’t expect this to introduce too much error.

4

“Biggest” refers to total length of ski runs, not attendance.

5

Some of these homes are also near major bodies of water, such as the homes around Lake Tahoe, and some of the resorts in Michigan. If you exclude homes we’ve already counted as being near a major body of water, you get around 270,000 additional homes.

6

There are, interestingly, a few census tracts around Central Park with a large percentage of vacation homes.