When a new dollar goes into the capital markets, after being bundled and securitized and lent several times over, where does it end up? When society’s total savings increase, what capital assets do those savings end up invested in?

When economists talk about “capital assets”, they mean things like roads, buildings and machines. When I read through a company’s annual reports, lots of their assets are instead things like stocks and bonds, short-term debt, and other “financial” assets - i.e. claims on other people’s stuff. In theory, for every financial asset, there’s a financial liability somewhere. For every bond asset, there’s some payer for whom that bond is a liability. Across the economy, they all add up to zero. What’s left is the economists’ notion of capital, the nonfinancial assets: the roads, buildings, machines and so forth.

Very roughly speaking, when there’s a net increase in savings, that’s where it has to end up - in the nonfinancial assets.

I wanted to get a more tangible sense of what nonfinancial assets look like, of where my savings are going in the physical world. So, back in 2017 I pulled fundamentals data on ~2100 publicly-held US companies. I looked at all the categories of balance-sheet assets, and picked out the nonfinancial types - things like inventory, patents, plant, property and equipment - then sorted, and found that 100 companies accounted for about half of the total nonfinancial assets. I then downloaded annual reports from all 100 of those companies (actually 102, for reasons I frankly don’t remember), and read through their balance sheets and notes to get a fine-grained idea of exactly what those capital assets look like.

This post is a compressed summary of findings from that project.

Disclaimers

Before we dive in, a few disclaimers. First disclaimer: this methodology is obviously quite biased. It completely ignores capital typically owned by the government (e.g. roads, water infrastructure) or individuals (e.g. houses, cars), and it’s weighted toward types of capital which are concentrated in fewer companies rather than dispersed over many. That said, the upside is that we know what the biases are; the methodology is systematic enough that there probably aren’t many unknown unknowns still hiding.

Second disclaimer: quantifying nonfinancial capital assets is tricky, because they depreciate. A 10 year old machine is worth less than a brand new machine. Different kinds of capital assets also depreciate at different rates. That means all the numbers in this post are sensitive to the details of accounting conventions - and unfortunately they don’t always use the same conventions (though at least different companies in the same industry usually use similar conventions for similar assets). The numbers in this post are good rough numbers on which to hang intuition, but don’t put a lot of confidence in their precision.

Third disclaimer: I have more background knowledge about some of these industries than others. I studied all of them enough to (I think) at least get a basic understanding of the capital assets involved, but there still could be unknown unknowns leaving me totally confused and interpreting numbers completely wrong. Please leave a comment if you see a mistake or if you want to add detail about a particular industry you know well.

Fourth disclaimer: I pulled these numbers in 2017. That was a while ago. Things change. Some things change more than others. Oil is a very cyclical industry; the power grid is pretty stable. AI datacenters weren’t a thing at all in 2017.

Fifth disclaimer: I pulled US companies, so naturally this is all fairly US-focused.

With that out of the way, we’re on to the fun part. Where is the capital?

Overview (With Numbers!)

The nominal grand total of capital assets (which is what I will somewhat-confusingly call the “real stuff” I was looking for) across all 102 companies was around $6.3 trillion. You should not take this number very seriously! It is largely an artifact of accounting conventions, and adding things up somewhat differently could change it by a factor of 2. But it does give a sense of the scale here, suitable for Fermi estimates.

The breakdown across my own off-the-cuff ontology of sectors looks like this:

Next I’ll do a walkthrough of those sectors and what their capital assets look(ed) like.

Oil - 25%

A bit more than half of the capital assets in the oil industry are in the “upstream” subsector, which basically means oil wells. Every well requires drilling a very deep hole, lining it with pipe and concrete, and (for the very large majority of the wells included here) fracking so that oil can flow more easily through the surrounding rock. That’s expensive, and there’s a lot of these things.

Each dot is a well in the Eagle Ford basin, one of several major US oil basins.

A lot of different companies own tens of billions of dollars each of those wells, with names like Anadarko Petroleum, Chesapeake Energy, Encana Corp, or Murphy Oil Corp.

Then there’s the rest of the oil industry:

Transport Infrastructure: pipelines, ports, tankers, storage, etc. Example companies: Cheniere, Kinder-Morgan.
Downstream and Retail: refineries, chemical plants, gas stations. Example companies: Phillips 66, Valero

There are also a few relatively large companies which vertically integrate wells, infrastructure and downstream processing; Exxon and Chevron are the biggest in terms of capital assets (at least according to these numbers).

Power Grid - 16%

Power plants, power lines, substations, etc. In the US, many different companies handle the power grid in different regions, and they all have both fairly predictable revenue streams and fairly steady capital expenditures on maintenance. The industry might as well be straight out of a textbook.

15 power grid companies were big enough to make it into my sample, and they mostly look pretty similar, just operating in different regions. The one exception is Berkshire Hathaway, which has almost half its capital assets in an otherwise-normal-looking power company subsidiary, but does a whole bunch of other stuff too.

Consumer - 13%

This category bundles together companies like:

Retailers: Walmart, Lowes, Home Depot, Costco, Target
Pharmacy: Proctor & Gamble, Johnson & Johnson, CVS, Walgreens
Food: Kraft Heinz, Pepsi, Kroger, Mondelez

The capital assets are what you’d expect: lots of buildings, equipment, and inventory. For instance, here’s Lowes’ breakdown:

Inventory: 9.5B
Land: 7.1B
Buildings & improvements: 17.5B
Equipment: 10.9B

(Note: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the company’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)

Telecoms - 8%

Companies: AT&T, Verizon, CenturyLink, Comcast, Charter, Sprint, T-Mobile, and Dish.

Looking into these gave me an appreciation for how the technical considerations of different telecom tech influences the capital expenditure and business model. AT&T is the oldest on the list, and their assets include an absolutely massive amount of “central offices”, i.e. the type of buildings where phone operators once sat, now populated by electronics.

A mobile-focused company like Sprint or T-mobile, on the other hand, needs tons of cell towers, but does not have a bunch of central offices. Comcast mostly owns coax cable originally intended for broadcast but now used to communicate in both directions for internet service, which is a whole different situation from AT&T’s legacy copper wire or Dish’s satellites and stations. (And presumably in the years since I looked into it, everyone involved has acquired a lot more fiber optic line.)

These are very different companies, for reasons mostly boiling down to their different legacy technologies.

Railroads - 8%

The “transportation” category in the pie chart above is actually all railroads. There are four companies in my sample: Union Pacific, Norfolk Southern, CSX, and most of the second half of Berkshire Hathaway’s capital assets.

As a sample, here’s how CSX’s assets broke down:

Rail & track: 7.3B
Ties: 5.4B
Grading: 2.6B
Ballast: 2.9B (gravel and crushed stone that holds track in line)
Bridges etc.: 2.3B
Signals & interlockers: 2.5B
Buildings: 1.2B
Locomotives: 6.1B
Freight cars: 3.4B
Work equipment: 2.1B
Other: 4.6B

(Note: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how a railroad’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)

Healthcare - 8%

Think pharma companies like Pfizer, and medtechs like Medtronic. As a sample, here’s Pfizer’s breakdown:

Inventory: 6.8B
Buildings: 9.8B, Machinery & Equipment: 11.2B, other stuff: 7.1B
“Developed technology rights”: 83.4B
Intellectual Property R&D: 10.1B
Brands: 9.0B

Compared to Medtronic’s breakdown:

Inventory: 3.5B
PPE: 4.8B (net)
Customer-related assets: 18.6B (i.e. Medtronic salespeople’s relationships with their clients)
Purchased tech and patents: 11.4B

(Note, as usual: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the companies’ assets are distributed, but don’t compare the absolute total to other numbers in this post.)

Tech - 6%

Now we’re getting to smaller categories, so I’ll rattle through them a little faster. This is e.g. Google, Amazon, Apple, Intel, etc. Sample breakdown for Google:

Land and buildings: 16.5B
IT assets: 13.6B
Construction: 7.3B
Leasehold improvements: 2.6B
Patents: 6.6B
Customer relationships: 1.3B

(Note: not adjusted, relative sense, blah blah hopefully you read one of the earlier warnings.)

Industrial - 5%

Think GE, Boeing, or Caterpillar (the company which makes excavators, backhoes, and other fun construction vehicles). At the time of this sample, GE was winding down their brief foray into owning and renting out aircraft, so their numbers took me by surprise. Anyway, here’s the breakdown for Caterpillar, as a sample:

Inventory: 8.6B
Buildings: 7.4B
Machinery & equipment: 15.4B
Software: 1.6B
Equipment leased to others: 6.1B
Customer relationships: 2.4B
IP: 1.5B

(Note: … ok just go look at the previous notes if you want the warnings for these numbers.)

Mining - 3%

Three companies - two with “gold” right there in the name, one mainly in copper. Here’s the breakdown for Goldcorp, no points for guessing what they specialize in:

Plant and equipment: 6.7B
Exploration potential: 8.0B
Reserves & resources: 16.3B

(Note: y’know, I should probably actually copy the warning occasionally, so: those numbers aren’t adjusted for depreciation in the same way as some other numbers, so use them to get a relative sense of how the company’s assets are distributed, but don’t compare the absolute total to other numbers in this post.)

Real Estate - 3%

Three real estate holding companies. One of them is Simon, which owns a bunch of malls. It’s exactly what you’d expect, at this point.

Automotive - 2%

Ford and General Motors. Here’s Ford:

Inventory: 8.3B. Raw material etc.: 4.0B, Finished: 5.3B
Leased vehicles: 29.7B
Buildings and land improvements: 10.0B (net)
Machinery & Equipment: 33.2B (net)
Software: 2.6B

(Note: vo chugga depreciation, ne degada forsooth, uncompare, caution.)

Logistics - 1%

Fedex and UPS. Here’s UPS’ breakdown:

Vehicles: 8.6B
Aircraft: 15.7B
Land: 1.4B
Buildings & improvements: 7.1B
Plant equipment: 8.4B
Tech equipment: 1.7B

(Note: go check the warnings in above sections. Preferably one which actually explains the warning.)

Miscellaneous

Finally, a fun section. Three companies defied categorization: Carnival, Disney, and Visa.

Carnival has about half of the global cruise market (including Carnival Cruise Line, but also a bunch of other brands). As you’d expect, the overwhelming majority of their capital assets are cruise ships.

Disney, you’ve heard of. Their parks are the big item.

Finally, Visa, the credit card company. I didn’t look into what was going on with their assets enough to know what was going on; the large majority was some kind of accounting artifact having to do with a big acquisition.

Learnings

My biggest single takeaway from this project is infrastructure. If you ask “where does the capital end up?”, the main answer is infrastructure. Power grid, oil wells and pipelines and refineries, telecom lines and equipment, trains and airplanes and trucks. Then buildings and manufacturing equipment.

I also learned a long tail of how lots of different pieces of the economy work and why; every category above could be a whole essay on how its industry works and what it looks like. Indeed, that was originally my intent, but it’s been 8 years and that project will clearly never be done (at least by me… I would love it if somebody else tackled it, I would 100% buy that book).

Lastly: if you want to see my raw notes on the assets of all the companies, they are here.