The Wealth-Creation Speed Limit
Is there a speed limit for wealth creation? Two recent Twitter threads asked this question in two different ways, which is the sign of an interesting problem.
In both cases, the question is: subject to an extreme constraint — time or headcount— how rich can you get? This is a salient question today, because many people in the technology sector have gotten very, very rich. It’s not a coincidence that the people asking this question either got rich from investing in startups, or are running a startup in the process of making their investors rich.
We don’t have a lot of data on individual programmers’ earnings, but we do have data on how quickly companies scale, and it looks like the pace at which companies hit $50 billion in inflation-adjusted revenue is accelerating:
I ran a regression to fit years-since-start against years-to-$50 billion. The start date is arbitrary. You could pick the first scalable tech product with network effects (Boulton and Watt’s steam engine, 1765), the rise of financial markets that didn’t exist to fund wars (Lloyds of London, 1686), or 0 AD (a nice round number).
In the time-honored tradition of finance, I considered several viable stories and went with the one that had the highest r-squared, which happened to be 1686.
I propose a simple rule that defines the speed limit of wealth creation: if you’re good at solving social problems, your success will be defined by engineering constraints; if you’re good at solving engineering problems, your success will be defined by social constraints. Any big tech company is an example of the latter. For the former, you can look at US energy policy: nobody with the political savvy to champion a modern energy policy can do enough arithmetic to determine that that means lots of nuclear power.
The Single-Person Speed Limit
What would it take to build a huge one-person company? Let’s work backwards:
- You can’t do retail, or anything that requires customer support.
- You can’t manufacture, unless you manufacture something defined by the creator, i.e. art.
- You can’t do anything that requires complex infrastructure you own. So natural resource extraction is out, although I suppose you could make a billion dollars as a solo entrepreneur if you owned a lot of land with a lot of oil under it.
- You can’t do anything that requires more than one specialist, which rules out lots of hardware businesses, and — unless there’s some low-hanging fruit we’ve all missed for a generation — all of biotech.
- You can’t do anything regulated; a billion-dollar business creates enough compliance work for at least one full-time lawyer even if it’s in a totally deregulated business. That rules out the rest of biotech, and several other fields besides.
So that leaves two possibilities: running a hedge fund that outsources the back-office, or being an artist. Picasso might have been the first person to make a billion dollars doing something that he theoretically couldn’t scale. An estate that’s complicated by “authentications, rights, and licensing deals,” sounds like the result of a business with more than a few lawyers on staff, but presumably he could have ditched all that, cranked out a couple more Femmes d’Alger, and hit the billion-dollar number with less headache. J.K. Rowling is another example: the boring stuff, like licensing rights, all gets outsourced to publishers and agents. She can just focus on Harry.
A hedge fund seems like a better bet. Investing is basically a business that allows you to outsource every single thing a company does, and charges you for the privilege. The difference between a company’s market capitalization and the book value of its assets is the fee that you pay to have someone else take care of the dirty work of turning those assets into as much free cash flow as possible. Warren Buffett ran his hedge fund with a tiny staff, and Berkshire’s corporate HQ is still pretty minuscule, with 25 staffers running a business with 377,000 employees.
There are some investment groups that make do with a small staff; there are many more large funds that don’t. Past a certain point, it takes some personal perversity to insist on doing all the work yourself, or outsourcing literally every boring part of the business besides security selection. That said, some people have the attitude to make it work. There’s one anecdote that makes me suspect that if Thomas Peterffy had been born just a few decades later, he would be the first person to run a one-person company that earned a billion dollars a year:
Three decades ago, to bypass a Nasdaq rule requiring all orders to be entered on a keyboard, his team built a robot to do it. “On active trading days, the robot typed so fast it sounded like a machine gun,” National Public Radio reported in 2012.
Satoshi Nakamoto, assuming he’s one person and not a team, might be a special case of a hedge fund. Even though the work was a lot of C++ and message board flamewars, the money all came from sitting on a big long position in a single asset.
The Company Speed Limit
The Bloomberg data on how quickly companies reached a $50bn run-rate show a pattern: over time, it’s happened faster. We can think broadly about three cases:
- The degenerate case of private equity companies and acquisitive consolidators. Most of the big PE complexes didn’t make the list, but if you add up all their portfolio companies, a few PE complexes do over $50 billion in annual revenue. Berkshire Hathaway probably counts, too; they didn’t make their money in any one industry, but by sourcing cheap capital from insurance and deploying it opportunistically. We can discount these companies, because there are multiple moving pieces: the way to get big in private equity is some combination of high returns and cheap capital, and since cheap capital is most abundant at a time when checks are largest, it swamps the returns effect. So, let’s ignore these. We can also ignore some of the companies that won through consolidation (so, let’s throw out banking and health insurance).
- Better operators in legacy businesses. Skimming the list, I generally see companies that are in retail or manufacturing, and had a cost advantage. These are both hard businesses. (Unless you have a cost advantage.)
- Companies that created or redefined an industry. These are the interesting ones.
Historically, the connection between inventing new things and getting rich from them has been somewhat tenuous. GE was a lucrative business that made J.P. Morgan a lot of money, Thomas Edison a decent sum, and Tesla basically nothing. What they contributed to the electric business generally had an inverse correlation with what General Electric did for them personally. Ford tinkered a lot with… Fords… but he was somewhat exceptional. The senior executives at tech hardware companies in the 60s, 70s, and 80s often retired with two or three commas to their names, but they were diluted pretty heavily in the process — if you’re selling something to the masses, you need to mass-produce it, and that means giving up equity.
Software is the first business in a long time that’s sufficiently capital-efficient that the founders can own a large share of the stock by the time the company is mature enough to IPO, but stable enough that the IPO usually won’t represent the business’s high water mark. One obvious feature of software businesses, which allows them to grow quickly over time, is the low marginal cost of the product. Spend a few billion dollars developing a new operating system; spend a few cents on bandwidth to deploy it to one incremental machine.
A non-obvious consequence of the low marginal cost of software businesses is that, if they have any network effect whatsoever, the optimal choice is to go for higher user count now and figure out monetization later. You see this pattern again and again with consumer-facing software companies: at first, ad loads are minimal and prices are cheap; over time, as user growth slows, more of the revenue growth is driven by monetization. But there are few high-value software companies with a declining userbase; the math seems to work best if there’s continuous growth in the number of people on a platform, higher growth in value per user, and higher-still growth in value captured by the company.
Which means that, when a software company hits a revenue milestone, the revenue number dramatically understates the company’s social impact. Imagine! Governments have been toppled by oil companies, even banana companies — and a consumer-facing software company of equal size has a cultural footprint many times larger.
This is why solving the engineering challenge (“How do I keep uptime high with a billion concurrent users?”) becomes less important than the social challenge (“How do I keep heads of state happy with me? How do I prevent politicians from making me a metonym for the unchecked power of elites-who-aren’t-them?”). It’s hard to plan ahead for this: even very smart people underestimate the power of compound interest, and if the number that’s compounding is “share of human attention I indirectly control,” the social consequences are dire.
Managing this problem is hard to parallelize, and it’s hard to spend money on. If you just buy politicians, you have three problems: you’re an obvious faker, so nobody trusts you; you’re an easy mark, so you’re the high bidder for people who are only influential on paper; and you’re choosing a side, which often means accidentally choosing the losing side. It’s not a coincidence that tech companies seem blindsided by politics; they succeed in part because they’ve optimized themselves to grow in metrics-driven, politics-free fields.
The Steady State
My trichotomy of enormous companies — the financial engineers, the low-cost legacies, and the innovators — is fuzzy. Eventually, every company ends up being a bit of all three. A good business throws off more free cash flow than it can effectively reinvest, so it ends up acquiring adjacent businesses. Every old and boring business was once fresh and interesting, so all startups eventually become part of the establishment. At the same time, every industry tends towards irrelevance, so any company that survives for a long time needs to have some sort of startupy nature to it. (The big three video game hardware companies were founded as a playing card company, an electronics shop, and a programming language company. Survivors, all.)
Politics seems to be the fiercest of the limiting factors. It’s probably cyclical. Governments are getting more intrusive — in many countries, both the left and the right have shifted towards more interventionist policies, reversing a decades-long trend. While it might be a coincidence that the rise of software businesses happened at the same time as the rise of populism, I doubt it: a global communications system provides the infrastructure for spreading viewpoints that are less filtered through the traditional media, which allows wilder swings in public opinion. While those swings could go in any direction, software and globalization have both increased income inequality within countries, which tends to lead to more populism.
So while technological trends are pushing on the accelerator, the accompanying social trends pump the breaks. Recently, technology has been winning: companies hit the $50 billion mark faster than ever. But we could easily be approaching an inflection point, where the abundance of information and the short attention span of the average online media consumer means there’s an endless hunt for the next big target.
This explains the phenomenon that a mature tech company’s manager ends up either a) spending so much time doing PR that people assume he’s running for President, or b) spends so much money on lobbying that if he calls Mitch McConnell or Nancy Pelosi, they put the President on hold. Get big enough to worry the government, and suddenly you’re moving at government speed. The mystery is: why does it happen so late?
This might be evidence, albeit in a vague and low sample-size way, that tech companies are excessively focused on their business and insufficiently focused on the transformative power of technology itself. A truly ambitious company monetizes just enough to keep the lights on, and has a lower discount rate for power than for dollars. (Since power eventually translates into wealth, this gives them a cost of capital of roughly zero.) You’ll know technology companies are as ambitious as they ought to be when the founders are testifying before Congress before they’ve made the cover of Forbes.
 One of the hard constraints an artist faces is that the demands of fans scale roughly linearly with the number of fans, so if you write back to the people who write you fan mail, eventually that becomes most of your output. I’m glad Neal Stephenson precommmitted to being a bad correspondent before any of his books got made into movies. It would be terrible if the world were deprived of more Stephenson novels, or more specifically first-two-thirds-of-Stephenson-novels, because he was litigating endless flamewars about the minutia of his books.
 Misha Gurevich points out that Notch, the creator of Minecraft, may also count as a single coder who made over a billion dollars. Notch shows how close you can get, but even in 2012, when Minecraft had hit $100m in revenues, his company had 25 employees. Minecraft shows that an introvert who isn’t especially motivated by wealth can run a small operation, but that it’s hard for such a person to be solo if they sell a product for $20/copy.