Destructive investing and the siren song of software
Auto manufacturing is a strikingly tough business. Your upfront costs are incredibly high: you need huge factories, massive amounts of raw materials. Design and manufacture is incredibly complex, with each car having thousands of precision-manufactured parts. Sales cycles are very long: most people consider and research before buying a car, since it's the biggest non-house purchase they make. Support cycles, too, are long, with cars staying on the road for a decade or more. Demand is highly elastic and subject to the vicissitudes of broader economic cycles. Storing unsold inventory is incredibly costly. There are very good reasons that new car companies tend to fail, and why existing companies, even though they are massive economic entities with billions or even hundreds of billions of dollars of revenue, and even though they offload every iota of risk they can to supplier and dealer networks, are constantly on the verge of bankruptcy*.
Given all that, why is it that investors—particularly tech investors, self-consciously prone to evaluating companies with the same sort of eyes as a venture capitalist might—are so high on Elon Musk? Why does Tesla, a startup auto manufacturer trying to do precisely the same thing that so many other companies (Tucker, Delorean, Fisker, the list goes on) have failed to do in the past, have stock that trades far, far above the stock of industry leaders like Toyota?
I think the answer lies in a particular habit of mind—delusion is a strong, but maybe accurate, word for it—that has taken hold among investors after the wild successes of venture capital in the era from approximately the early 1990s to the late 2010s. It's a delusion that explains a lot of other seemingly irrational and counterproductive behaviors on the part of investors today, from private equity's strategy of turning profitable, successful businesses into risky, failure-prone ones to "AI" boosters' seemingly iron-clad determination to apply large language models to tasks like producing interesting journalism or movie scripts for which they are manifestly unsuited.
In the mid 1970s, the industry that now gets called "tech" looked very different. The computer industry, as it was then known, was dominated by a small handful of enormous players. IBM, Digital Equipment Corporation (DEC), Prime, Apollo, Sperry and Burroughs were in the business of selling large, expensive computers in small quantities—often just one—to Fortune 1000 businesses. Their business had high capital costs. The amount of steel alone in an old full-size mainframe rack was substantial. It had long support cycles. Customers demanded (and paid for) ongoing white glove support for years after a purchase. It had low margins. Hardware manufacture is fantastically expensive, requiring extensive development cycles, large amounts of raw materials, and complex iterative testing on physical prototypes (there’s a remarkably book about this process by Tracy Kidder, Soul of a New Machine. From a revenue perspective, the business was lucrative, but margins were tight, and so only a small handful of companies (many collected in the nation's premiere tech hub, suburban Boston, in close driving range to MIT) had the necessary scale and resources to even compete. They were funded by public stock sales, ongoing long-term contracts, and large infusions of cold war-era government funding. The idea of a "startup" didn't exist in its modern form.
By the late 1980s the industry had completely transformed. The rise of the personal computer was the most noted bellwether of the transformation, but that was just a leading indicator of a more profound transformation in the business models of computing. In 1975, Bill Gates had released a BASIC interpreter for the Altair 8800, more or less the first microcomputer, albeit one with no display or keyboard sold as a kit to hobbyists. This was a strange and deeply antisocial thing to do in the small community of personal computer hobbyists; software, as a set of instructions to make the computer do something, was generally freely exchanged by people who were lucky or committed enough to get access to a computer. The physical system itself was far more expensive to build and acquire and evidently valuable. The software was simply an enabler. It wasn't even particularly clear at the time why you would bother to pay for software, given that it could be reproduced for no cost whatsoever. It took Bill Gates's combination of entrepreneurial arrogance and affronted woundedness to insist that his interpreter was in fact a creative work on par with novels or symphonies, assert that it was covered by copyright, and create the software industry.
At the same time as Gates was clawing software as such back from the commons, modern venture capital was being created. Firms like Kleiner Perkins and Charles River Ventures were created to funnel funds from the (also pretty new) private equity world into technological developments coming out of tech-oriented universities like Stanford and MIT. Their initial investments were in companies with strong, well-attested technical pedigrees hoping to grow into the sort of "big iron" computer companies that were already existing and successful. DEC had originally been funded this way, as had companies from the first AI boom like Symbolics.
This new funding model interacted explosively, in the Bay Area, with the growth of software as a business. Software companies did not required extensive, proven out academic research. They required very little startup capital. Their marginal cost of producing additional units was zero. Reasonably quickly in the 1980s venture-funded software companies started clearing returns of 30% or more, about an order of magnitude more profit than computer companies had previously been able to achieve. The older big iron companies like DEC found themselves competing as providers of computerized business systems against rivals that had to a first approximation zero capital costs. The east coast computer industry was decimated, caught totally flat-footed by this new business model, and a large number of investors and engineers got startlingly, even appallingly, wealthy.
Then, the same thing happened again, at far greater scale. The business boom caused by the introduction of the graphical web browser burned through the businesses of communications technology like a hydrogen explosion. Orders of magnitude more and larger fortunes were made in Silicon Valley, and whole swaths of high capital cost industries—long distance communications, newspapers, even television—were thrown into disarray or extirpated as they found themselves competing against dynamic new players with essentially zero capital costs who could undercut any prices they could possibly charge.
After the second time, the venture capitalists began to believe that this was something that they had done on purpose, and could rely on. "Software is eating the world," proposed Marc Andreessen, who had parlayed his brief time building the first graphical web browser into a career as a sage among a set of people overwhelmingly convinced of their own importance. Endless treatises were written on how transformative new business strategies had allowed for—would continue to allow for—this kind of transformation. VCs invested in companies based on what they described as the "hockey stick" growth curve. They were looking for businesses where scale and revenue would hit an inflection point where they would start to grow something close to exponentially, vastly outstripping the underlying costs. They wanted this because they had had it before, because they had come to believe it was not merely their due, but something close to the ineluctable purpose.
These two startling and frankly largely coincidental moments in the history of venture capital I think help explain some of the more puzzling recent behavior of the investor class today.
Elon Musk's genius—and I think it is a variety of genius, although that is obviously a word that should be applied to that particular person only with very great care—lies in his ability to convince the people who are most sweatily chasing the dream of massive profit born of zero marginal cost that he can effect the kind of transformation that was seen in computing and communications systems to other capital intensive, highly regulated, difficult-to-enter industries. The reason behind his push for autopilot in Teslas is not that he truly believes that autonomous vehicles are the transformative future of transportation but because a self-driving car is a car where its primary function happens in software. The central differentiating function of a self-driving car is its self-driving system, which can be priced and updated and sold like a cloud software product. The actual auto manufacture is, in some ways, incidental. To investors chasing the dragon of VC’s historic successes this is intensely desirable, enough so that they're willing to disregard questions about its fundamental plausibility. The mere chance that it might happen, that autos might become, in a business sense, like the software business, is enough to convince them to inject millions upon millions of dollars into his business, even as promises go unfulfilled for years or even decades.
The same dynamic is at play when private equity—or large conglomerates behaving like private equity in acquisitions—buy newspapers or websites on the premise that they will achieve vast and cost-less achievments in efficiency, whether by replacing writers with AI or some other more-or-less implausible snake oil. It might not work, indeed it probably won't work, but history tells them that the payoff if it does work is so extravagant that the risk, to them, seems perfectly worth it.
This deluded drive to repeat the early business successes of tech is so compelling, in fact, that it has proven itself able to outlast a vast collection of signals that the transformations that built the reputation of venture capital—and transformed the nature of private equity—were outliers, vanishingly unlikely to be repeated. The financiers chasing that dragon will keep tearing down productive industry after productive industry, throwing money at obvious charlatans, far past the point of no return, because they have convinced themselves like a degenerate gambler that their early wins were the product of some inherent magic in their fingers, and all they can do is try to recapture it.
* You can find a great discussion of this, particularly as it relates to Tesla, in Ed Niedermeyer’s essential book Ludicrous
