Two hundred twenty nine. That was the number of private companies valued at a billion dollars or more—so-called “unicorns”—worldwide in January 2016, according to a report released by VentureBeat. The Wall Street Journal offered a more conservative estimate, but even there the overall trend was clear. In just two years, the number of unicorns worldwide had more than tripled, from 45 to 149.
To prophets of doom and gloom, there was only one explanation for the unicorn explosion: Silicon Valley was in a bubble, and it was about to burst. In April, Benchmark’s Bill Gurley published a blog post claiming the unicorn financing market had become “dangerous … for all involved.” In May, the Wall Street Journal’s Christopher Mims suggested the end might have already begun. In June, San Francisco announced it was working on an “economic resiliency plan” to help the city survive the inevitable downturn. The whole tech world held its breath.
And nothing happened. Sure, there were down rounds. Some old unicorns, like content company Mode Media, shut down entirely. The unicorn explosion slowed its pace, with only a few new companies earning that status. But it’s now 2017, and we’ve seen nothing like the collapse in startup valuations and consequent economic catastrophe that everyone’s been predicting.
Here’s the thing: The collapse isn’t going to happen this year, either. Or next year. Or the next, or the next, or the next. That’s because, despite sky-high startup valuations, we actually aren’t in a bubble. Most of those valuations reflect actual value creation — a level of which we haven’t seen since the second Industrial Revolution. And as innovation accelerates, they’re only going to get bigger.
To see why, we’ll need to take a trip into the economic past.
Information technology as a “general purpose technology”: the third industrial revolution
At the tail end of the 20th century, there was much puzzlement in economics circles over why the information technology (IT) revolution wasn’t doing much to accelerate productivity growth. In 1987, labor economist Robert Solow quipped that “you can see the computer age everywhere except in the productivity statistics.” The so-called “Solow paradox” was not what it seemed, however. Looking back to the first and second industrial revolutions, economic historians like Stanford’s Paul David were able to show that a slowdown in productivity growth after a major technological innovation was not a paradox but a sure sign that a revolution was occurring.
Technological innovations that trigger changes on the scale of the first or second industrial revolutions are known in economic parlance as general purpose technologies (GPTs). General purpose technologies (GPTs) are special because via a single breakthrough they open the door to potential innovations across many, if not all, sectors of the economy. Steam was a GPT in the first industrial revolution and electricity was one in the second. And, as historians like David have argued, IT is the GPT in a third industrial revolution that is still occurring.
The U.S. saw a slowdown in productivity growth occurred between 1890 and 1913, when electricity was just being introduced. This may have been due to the slow pace of factory electrification, as David has argued, or some other cause. But the end result is clear: After taking more than 20 years to reach full penetration, electrification suddenly set off an enormous acceleration of productivity growth for the next 15. If that acceleration hadn’t been cut short by the Great Depression, who knows how long it could have gone on.
Evidence shows a similar pattern with the adoption of IT. Writing in 2005, Jovanovic and Rousseau dated the period of IT adoption as beginning in 1971, when Intel invented the microprocessor that would become the core of the first PCs. Almost immediately, productivity growth slowed as offices began taking baby steps into the PC revolution, only picking up again in the mid-1990s. Jovanovic and Rousseau predicted that if IT adoption was anything like electrification, the economy would continue to experience accelerating productivity growth through the first half of the 20th century.
Here’s the thing, though: That massive productivity growth upswing hasn’t happened yet. After 2004, productivity growth slowed again. Which means that in 2016, we aren’t yet reaping the full benefits of the IT revolution.
In fact, a completely different GPT revolution may be at hand.
Will smartphones drive the fourth industrial revolution?
According to a 2015 McKinsey Global Report, the U.S. economy’s use of technology is still only about 18 percent of its potential. Beyond Silicon Valley, many industries’ tech is stuck decades in the past. Air traffic control uses computer systems from the 1970s; the U.S. nuclear program is still run on floppy disks. We are nowhere close to reaping the benefits of the third industrial revolution yet.
And yet, we may already be in the midst of a fourth — this one driven by smartphones and AI. An Economist special report notes that the post-2004 dip in productivity growth “seems to have coincided with an apparent acceleration in technological advances as the web and smartphones spread everywhere and machine intelligence and robotics made rapid progress.” In other words, we’re in for yet another wild ride.
The Economist report goes on to blame the productivity-growth dip on the fact that globalization has opened up access to armies of cheap labor abroad and thus removed the incentive to invest in labor-saving devices at home, “trapp[ing] rich economies in a cycle of self-limiting productivity growth.” But as labor costs rise in China and other developing economies, the cycle is bound to break. When it does, we’ll see productivity growth accelerate again, perhaps extremely rapidly, as the gains of the fourth revolution will be piggybacking on the as-yet-unrealized gains of the third. It’ll be like nothing ever before seen.
We don’t have to wait until then, however, to get a preview of how impactful the revolution will be. There are firms today that use existing technology to their full potential. They benefit from two main mechanisms that drive increasing returns: network effects and data effects. A network effect is where the value of your product increases as your consumer or user base increases (e.g. Uber, Facebook). A data effect is where the productivity of your R&D is an increasing function of your stock of consumer and/or user data (e.g. Google search).
In both cases the leading firms are able to generate the fastest growth in productivity and market share. So the winners keep winning and pull far out ahead of the others. This what we’re seeing now. A recent Brookings Institute report found that these winning “frontier” firms saw productivity shoot up in 2001-2013, while the losing “non-frontier” firms’ productivity was nearly stagnant. In other words, the lagging productivity statistics in that period are concealing vast gains for companies that have fully implemented available technology.
This is why, in 2016, we’re seeing such an enormous differential between the haves and the have-nots, the unicorns and the rest. Companies that embrace today’s tech fully are already seeing an unimaginable acceleration of productivity growth that has not reached the rest of the economy yet. When it does, those sky-high unicorn valuations will no longer seem noteworthy; instead, they’ll be par for the course.
Whether companies IPO or stay private, GPTs drive an influx of capital
There weren’t any unicorns when the second industrial revolution was just getting underway, but there was a stock market. Jovanovic and Rousseau note that the number of IPOs spiked in 1895, just five years after the introduction of electricity, and that IPOs continued to take up larger-than-usual market share for the duration of the GPT adoption period. In other words, investors were infusing new businesses based around the GPT with enormous amounts of cash, while incumbents faced with high adoption costs faltered. Though Jovanovic and Rousseau don’t point this out specifically, their numbers show that IPOs spiked again in 1920, a few years after productivity growth began accelerating again.
In the case of IT adoption, Jovanovic and Rousseau note, the reaction was more delayed. True, IPOs steadily increased their percent share of stock market value beginning in about 1977, but there wasn’t a true spike in IPOs until the late 1990s, perhaps because IT adoption initially was even more expensive than electrification. Jovanovic and Rousseau argue that this delay means some of the impacts of IT adoption might be delayed compared to electrification — in other words, there is more yet to come.
It’s possible that our glut of unicorns in 2014-2016 is the privately funded equivalent of the spike in IPOs in 1920. It’s a second wave of wealth creation happening now that the benefits of the IT revolution have largely taken root. Or it could be the equivalent of the IPO spike in 1895 — a sign of the beginning of another revolution.
Either way, there are strong reasons to believe tech companies’ high valuations are here to stay. Online distribution tends to present the possibility of “increasing returns to scale.” This means that as technology-based firms grow large, their profit margins can expand rather than shrink — as is the case when diminishing returns to scale prevail. As a result, technology-based firms should trade at much higher valuation multiples than firms subject to diminishing returns. This is not fully appreciated by investors because increasing returns are still represented as an aberration in the textbooks and financial analysts still model earnings growth on the assumption of diminishing returns. But it is yet another reason the unicorn explosion is not, necessarily, the sign of a bubble but of a revolution.
[This story was written with contribution from the Hippo Thinks research network.]