How Bubbles and Megaprojects Parallelize Innovation

Welcome back to The Diff. Here are the subscribers-only posts you missed this week:

Coming up in the next few weeks, I’ll be talking about: why US and European payment regulations promote different kinds of financial innovation, the theory that “a mine is a hole in the ground with a liar on top” as a long-term driver in natural resource equity returns, a profile of an interesting conglomerate, a writeup of a boring company that can earn far higher profits in a post-Covid world, gold, a general theory of why US equities have outperformed so much of the rest of the world for so long, and the economics of state-backed airlines.

Subscribe today to receive those posts, view the archive, and join today’s subscriber call with polling wizard David Shor.

Subscribe now

This is the once-a-week free edition of The Diff, the newsletter about inflections in finance and technology. The free edition goes out to 16,574 subscribers, up 266 week-over-week.

In this issue:

How Bubbles and Megaprojects Parallelize Innovation
Vaccine Economics and Emerging Markets' Tough Choices
Big Tech and the Missing Middle in Emerging Asia
The Personal, The Political, and Ant
Form, Function, and Ads
Understanding Value’s Underperformance

How Bubbles and Megaprojects Parallelize Innovation

There are two equivalent ways to describe a financial bubble:

It’s when investors pay up for assets without regard to their real valuation, whether that means high six-figure no-doc loans to people making minimum wage in 2006, massive checks to the fifth or tenth online pet food company in 1999, or a meme-themed crypto project in late 2017.
It’s when the flow of money into a market directly or indirectly validates the thesis of investors in that market.

Explanation #1 is simplistic, because it doesn’t posit a mechanism. It makes calling a bubble easy in retrospect. Just look at what assets went up a lot and then went down. Explanation #2 is more powerful, because it offers an explanation and some guidance on when the bubble can break down. Many famously irrational bubbles turn out to have internal logic like this. For example, here’s how Paul Graham explains Yahoo’s valuation in 1999:

By 1998, Yahoo was the beneficiary of a de facto Ponzi scheme. Investors were excited about the Internet. One reason they were excited was Yahoo’s revenue growth. So they invested in new Internet startups. The startups then used the money to buy ads on Yahoo to get traffic. Which caused yet more revenue growth for Yahoo, and further convinced investors the Internet was worth investing in. When I realized this one day, sitting in my cubicle, I jumped up like Archimedes in his bathtub, except instead of “Eureka!” I was shouting “Sell!”

And here’s how Greg Lippmann’s subprime short thesis is described in The Greatest Trade Ever:

[Deutsche Bank quant Eugene] Xu split the country into quartiles. He discovered that states with the lowest rates of default, like California, Arizona, and Nevada, also claimed the highest growth in home prices. The quartile with the highest rates of default, on the other hand, had the slimmest growth in home prices. Florida and Georgia, for example, seemed similar in many ways, but Xu’s numbers showed Florida had a much lower rate of default than its northern neighbor, which seemed to be due solely to its soaring home prices.

…

“Holy shit,” Lippmann exclaimed to Xu on Deutsche Bank’s trading floor while reading over his work, “if home prices stop going up, these guys are done.”[1]

Let’s talk about another bubble. This one also involves rocketing market values, scams and frauds, volatile markets, and crazy self-fulfilling extrapolations. It’s the story of the auto industry. In the early 1900s, a handful of entrepreneurs in Detroit had a zany idea: some day, cars would not be driven by eccentric hobbyists; they’d be a common mode of transportation! Meanwhile, at Standard Oil’s headquarters in New York, another nutty thesis was brewing: maybe the rise of the electric light bulb didn’t mean that the oil industry, which up to that point depended on kerosene lamps for its revenue, was doomed.

For either of these to be right, they both had to coincide: for cars to be a ubiquitous means of commuting, gas had to be available everywhere. Otherwise, a car was just a fancy toy.[2] It would not do to drive a long distance, run low on gas, and then learn that the automobile revolution hadn’t reached your destination, and you didn’t have enough fuel to get back. Gas stations solved this problem, but they came fairly late. The Prize estimates that in 1920, there were fewer than 100,000 places to buy gas, and that most of them were “grocery stores, general stores, and hardware stores” that sold it by the can. The first gas station was founded in 1907, and by 1929 there were 300,000 stores that sold gas, almost all of which were gas stations. The same book notes that oil production rose from 1.0m barrels per day in 1919 to 2.58m in 1929, and that 85% of 1929’s consumption was gas and fuel oil.

This means an oil wildcatter in the 20s was making a bet on the widespread adoption of a cutting-edge technology, the car. Meanwhile, the car manufacturers were making a bet on the continued productivity of the oil industry. If either industry had failed, both would have: a shortfall in car production would have flooded the market with oil, wiping out the (almost always financially overextended) oil entrepreneurs. A shortage of oil would have made the total cost of ownership for cars prohibitive, and that would have slowed down the shift towards car-friendly cities.

The internal combustion engine dates back to 1876. Oil is older (it’s mentioned in Herodotus), but oil as an actively exploited energy source dates back to the Titusville well in 1859. But the most influential boom both went through happened decades later, and exactly in sync.

Semiconductors and software had a similar tandem bubble cycle, with each generation of software justifying the next generation of chips. And still later, the glory days of ISPs as growth stocks lined up with the rise of publicly-traded dot-coms: VCs who invested in e-commerce were indirectly subsidizing AOL and Compuserve, and those companies were indirectly subsidizing e-commerce.

Sometimes, the bubbles fall slightly out of sync. Fiber optic infrastructure received far too much investment in the late 90s and very early 2000s, well before there was a ready supply of streamable content. By the time streaming started to turn viable, this infrastructure was still in the ground, and now very cheap. (Not as cheap as it could have been, since Google was buying it up.)

But in general, the paired-bubble concept is a powerful one. As long as both sides of the bubble have a lag between when the decision to spend is made and when the results are realized, they can leapfrog each other: in period 1, company A invests; in period 2, company B invests in response; in period 3, company A’s investment creates a broader market for company B’s product, which comes along in period 4; this success encourages A to launch another round of spending, repeating the process.

Bubbles aren’t the only mechanism for coordinating parallel innovation. Scientific megaprojects can, too. One of the impressive things about the Manhattan Project was how much of it got started on the assumption that other parts of the project would finish successfully. The uranium enrichment plant at Oak Ridge, for example, was designed to use an uncertain enrichment technique at scale. It was possible that the plant would not be able produce uranium in sufficient quantity and purity to be useful in the atomic bomb (only a few years before, estimates for how much uranium would be required varied by an order of magnitude in either direction). Once completed, the plant would require electricity, and would, in fact, require the single largest power plant ever constructed up to that time. So a power plant was built for a facility that might not function at all, and, if it functioned, might not do the job it was intended to do, and, if it did that job, might end up producing raw materials for a project that wasn’t viable for some unrelated reason.

Atomic theory and the reality of nuclear weapons were linked by a long chain of technical uncertainties, and resolving them serially would take too long. So everything got built at once, based on rough estimates that were rapidly, continuously refined.

Any researcher interested in nuclear weapons in, say, 1935 could have looked at the information published up to that point and concluded that such weapons were possible. But penciling out all the technical problems that would have to be solved to be sure they were viable was daunting, and building only part of the project was worthless; a theoretical design for a bomb was pointless when the largest available samples of pure U-235 were barely visible to the human eye. Refining larger amounts would have been a ludicrously expensive idea without a ready blueprint for how to use them. (The entire Manhattan Project ended up requiring an investment equivalent to the value of the US auto industry at the time.) A megaproject parallelized a set of tasks that would never get done serially.

Today, it’s possible to look at trends that have a similar parallelization trend going on.

Nvidia shares are up 124% this year. It’s an expensive stock. But every new generation of GPUs encourages more GPU-intensive applications, and those applications encourage Nvidia to produce its next generation of GPUs. If everyone trying to automate vehicles, build better deepfakes, or produce machine-generated text for business treats Nvidia’s progress as a given, then Nvidia can treat demand as a given, too.
UPS and Fedex have also outperformed this year, as retail shifted towards e-commerce. But even the shift back leaves more room for delivery: the last mile can be fulfilled by in-store shopping, curbside pickup, or delivery, and when every big-box store is optimized to offer all three, it means the biggest complement to the delivery business is suddenly cheaper.
Amazon (Disclosure: I’m long) constantly pushes these lagged self-sustaining bubble effects. The site’s userbase means that Amazon can supply an effectively limitless number of customers for any given product offered by a third-party seller, which means that the expected revenue from new products is higher. Amazon creates the demand for obscure niche products, and then harvests profits from directing that demand to them. AWS pushes a similar shift in economics: when the unit cost of scaling is known, it’s an implicit subsidy for any business that expects to scale.
Gold and Bitcoin (Disclosure: I’m long both) look, on the surface, like cases where this recursive process should not apply. And yet, in a sense it does. Money is the bubble that never pops, and both Bitcoin and gold are sometimes used as substitute currencies, at least for the purpose of maintaining reserves. Goldbugs who argue that gold is superior to any form of fiat currency have a lot of history to argue against, but asset allocators including central bankers continuously run the numbers on the key question they’re trying to answer: if they want to preserve purchasing power for emergencies, what mix of currencies and currency-like assets is the most appropriate? When nominal rates are low and inflation is a concern, the 0% yield and inelastic supply of gold and Bitcoin looks relatively more attractive. Bitcoin has a small market value relative to the currencies central banks choose between, but this means that every Bitcoin price boom makes it a more viable candidate for a substitute store of value.
Solar power and batteries are complements, and seem to be relentlessly driving one other’s unit costs down. Solar power is getting cheap on the margin, but can’t supply continuous power that matches typical use patterns. Energy storage lets power companies think about average consumption over a period of years rather than minutes, so every round of cheaper solar panels raises the available market for batteries, while every new battery manufacturing plant grows the market for solar.

Every financial mania requires suspension of disbelief, but sometimes that’s entirely rational. Early twentieth century progress in cars and late twentieth century progress in computers were both literally unbelievable to anyone who watched them happen at the time. As it turns out, sometimes the intersection of finance and technology implies a double negative: when two industries producing complementary products embrace a shared irrational delusion, the delusion comes true.

[1] Looking back at these anecdotes, one of the striking similarities is that figuring out how a bubble works is an emotionally moving experience (“Eureka!” “Holy shit!”), whether you’re about to cash in your options or buy a bunch of CDS contracts.

[2] In his excellent Science Since Babylon, Derek de Solla Price says “Amongst historians of technology there seems always to have been private, somewhat peevish discontent because the most ingenious mechanical devices of antiquity were not useful machines but trivial toys.” Given how many technologies turn out to have older antecedents that were never put to widespread use, it may be the case that technology requires the intersection of a toy and a speculative mania or megaproject.

This piece is adapted from a forthcoming book Tobias Huber and I are working on. Stay tuned for more.

Discussion question: are there other current bubbles that are parallelizing innovation? I’m opening comments to free as well as paid subs for this one.

Elsewhere

I joined Jordan Schneider on ChinaTalk to discuss vulnerabilities in China’s financial system with Lauren Gloudeman and Logan Wright of the Rhodium Group, who have both done some very interesting work on the topic. It’s a fun dive into the quirks of the system, the incentives and information asymmetries of regulators, and how to track how stressed China’s banks are.

Vaccine Economics and Emerging Markets' Tough Choices

1.35bn doses by year-end 2021, compared to a world population of 7.6bn, forces some difficult choices. One of the hardest will be for the leaders of emerging markets, who need to handle a more drawn-out period between when the pandemic’s effects are felt and when it’s no longer a problem. As The Economist points out ($), this is especially challenging for India. Their economy “does best when the rest of the world does well—but not too well. India’s exports benefit from global growth. But when the world economy gains too much momentum, interest rates and oil prices can rise uncomfortably high, hobbling a country that is a net importer of both capital and crude.” In the rich world, deficit spending is a reasonable way to deal with the problem of a recession whose catalyst has a known end date, but poorer countries are constrained in how much they can borrow.

One compelling possibility: immigration restrictions in the US and travel difficulties everywhere will reduce emigration, meaning that some of the skilled workers who typically leave India to work in richer countries will stay behind. That turns a brain drain into a source of service exports.

Big Tech and the Missing Middle in Emerging Asia

Pondering Durian has an excellent writeup of how big tech companies will reshape economies in Southeast Asia. In the rich world, there’s a relatively smooth distribution from many small companies to a smaller set of mid-sized ones to a handful of huge corporations. In developing markets, the gap is often wider. As the Durian notes:

In India & Southeast Asia, there isn’t much of a middle to hollow out. Given the distribution of employment, the biggest opportunities come not from digitizing corporations, but from digitizing SMEs, solidifying the classic bifurcation of big tech & the micropreneurs they serve. This follows the Chinese examples of Taobao, Meituan, Didi, and even Douyin to bring the sprawling SME / artisan class into the 21st century.

The Indian & Southeast Asian ecosystems are following a similar path; the race to be the preferred partner for SMEs is on. In Southeast Asia, Shopee, Lazada, Tokopedia, Bukalapak, Sendo, Tiki and more are jostling to build supplier liquidity on C2C marketplaces. Grab & Gojek are fighting over drivers and food-delivery. Grab Financial, GoPay, Momo, VNLife, Mynt, Paymaya, and Truemoney are atop the scramble to modernize payments, O2O marketing, and digital financial services for SMEs. B2B marketplaces like Ralali, Telio, Bukalapak Mitra, Warung Pintar are clashing (or partnering) with PoS / basic accounting tools like Moka or BukuWarung - striving to own the offline merchant workflows where they are running smack into eWallets. And the enterprise software solutions are duking it out, fighting tooth and lol… gotcha, no enterprise titans :). India is the same story, different names: Flipkart, Swiggy, Zomato, Google, Facebook, Jio, Paytm and a host of others striving to embed themselves with the economy’s largest opportunity - the SME.

Now that we’re all epidemiologists, it’s much easier to throw around analogies from the world of infectious diseases: the US and Western Europe have had a long time to develop a reasonable immune response to big companies, that involves politics (but not too much!), a robust set of small and mid-sized companies with lots of collective economic heft, and financial systems that can fuel the growth of challengers when the dominant firm in an industry gets too lazy. In a more vulnerable population, the large-company model has fewer obstacles to runaway growth.

The Personal, The Political, and Ant

The WSJ has a detailed look at how Ant’s IPO came to be undone ($): Xi Jinping was personally furious at Jack Ma’s speech talking up Ant and talking down legacy banks (translated text here). Xi suggested tighter regulations on fintech, and approved a stricter-than-originally-planned rule that would require Ant to fund 30% of its loans.

In one sense, China’s government is legible, because it’s impossible to coordinate the behavior of almost 1.4bn people in a one-party state without writing down, in detail, what those people are supposed to do. A less centralized government, or one that’s less central to people’s lives, can actually afford to be more vague. But the necessity of many written rules doesn’t imply the nonexistence of unwritten rules, and apparently one of those unwritten rules was to avoid criticizing China’s banks.

Form, Function, and Ads

New companies grow by doing something radically different, but as they mature they end up converging on the rest of their industry. And this process works in the other direction, too: incumbents end up borrowing ideas from challengers, often at scale. Online businesses built on direct-response marketing, like Booking.com and Expedia, have found that the best way to keep growing is to use branded TV advertising, for example. Two more examples:

TikTok enables small, insurgent brands to punch above their weight and market themselves to a huge audience through organic, viral content. Naturally, big retailers are recruiting their employees to make astroturfed TikTok videos, more or less cancelling this out. The more a social network skews young, the more likely it is that a company with an army of minimum wage workers will, through the law of large numbers alone, employ at least a few people who could be TikTok stars. While a single store can’t bet on this, a large chain can.
Amazon has grown its profits by treating its retail arm as a source of first-party purchase data that can be used to target ads. Now Walmart and Kroger are ramping up their own ad efforts ($, WSJ). Holding the quality of the business itself constant, a low-margin retailer has an advantage in advertising relative to a high-margin one: at a given return on equity for their core business, low margins mean access to a larger dollar value of transactions, which means more information for targeting ads. (This advantage is even stronger for companies that take a small amount of revenue from touching a large dollar volume of transactions, like Shopify, BigCommerce, or payments companies.)

Understanding Value’s Underperformance

Lyall Taylor has a long and occasionally quite pugnacious look at why value stocks have underperformed. As he notes:

Examples abound of formerly highly-rated franchises that were part of the high multiple growth and quality parts of the investment universe that suffered massive disruption and falls from grace, and the tendency of a not immaterial minority of highly-rated quality/growth stocks to fail to live up to their expected potential and suffer major de-ratings has been a fundamental driver of the long term underperformance of high multiple stocks as a group.

Normally, value investors are characterized as cynical and pessimistic: they’re not going to pay 80x earnings for Netflix because they’re skeptical that any company could be worth such a high price relative to its current profits. But in a meta sense, value investors are the optimists: they’re making a bet that the future hasn’t been figured out just yet, and that the best business models investors can back today are inferior to the businesses of the future. A growth investor is a local optimist and a global pessimist, who believes that Facebook, Amazon, and Apple will continue to do well and nobody will disrupt them. A value investor is a global optimist, who thinks banks and energy companies aren’t down for the count just yet, and that every big tech company today is an IBM waiting to meet its Microsoft.