Certain intervals of time we accept as givens. The earth rotates on its axis once every 24 hours; seven days make a week; a half-hour sitcom is really 22 minutes plus commercials; Apple does a big-deal iPhone launch every other year. You know: the fundamentals.

In 1965, Intel co-founder Gordon Moore identified one of these intervals in a way we still associate with his name. Thanks to miniaturization, he observed, the number of tran­sistors that could fit onto a single microchip was doubling every year, making computers exponentially more powerful, energy-efficient, and inexpensive. In 1975, he revised Moore's law, as it was by then known, to set the period of doubling at 24 months, where it has remained ever since.

Until recently. In the past few years, one chipmaker after another has acknowledged the undeniable: Moore's law is no more. At 10 nanometers or smaller, transistors threaten to become so small that quantum uncertainty­--a fuzziness that issues from the weird physics that govern at the atomic scale--will make them unreliable. With billions of dollars' worth of R&D at stake, efforts to engineer a way out of the impasse have been quietly dropped in favor of more promising avenues, such as specialty chips for applications like virtual reality or machine learning. It's the end of an epoch for an industry whose nickname derives from the substance those chips are made of. "We've gone from an era of computing abundance to an era of incremental improvements," says Steve Blank, the influential startup guru whose résumé includes stints running two semiconductor firms.

That could be bad news for all of Silicon Valley, considering what all that abundance has built. Until now, Moore's law functioned as a sort of click track, helping the entire electronics and computer industry keep the same beat. Through the coordinating role of a body called the International Technology Roadmap for Semiconductors, device makers knew what to expect from future generations of processors. That foreknowledge, in turn, let software developers plan for future generations of devices. If you wanted to build anything requiring a lot of computing power--say, an augmented-reality video game--it was possible to engineer the most ambitious version knowing that other parts of the ecosystem would eventually serve up the components needed to make your vision pan out. That assumption is no longer operative. You can still build it, but the chips might not come. So are you sure you want to build it?

 inline image

After a few decades of hyperfast innovation, it's easy to believe things will always be that way--until they aren't. Gas-powered car engines have barely evolved since the advent of fuel injection. Not only have humans not landed on Mars, we haven't even been to the moon since 1972. There's no reason to expect computing will be any different. "Planes stopped getting faster before we got eight-minute flights from Tokyo to New York," says Jason Kelly, CEO of Ginkgo Bioworks, which engineers microbes for industry using synthetic DNA. A breakneck pace for innovation "runs out sometime," he adds. "If this slowdown in chips had happened prior to the iPhone, we wouldn't have smartphones."

Right. The iPhone. There may be no better illustration of what's at stake here. Its skip-a-year release cycle wasn't just to satisfy some aesthetic quirk, like Steve Jobs's preference for rounded corners; it was a downstream function of Moore's law. As chip innovation decelerates, the most profitable product in history is slowing down too. In 2016, when it debuted the barely changed iPhone 7, Apple quietly hinted it might not stick to the two-year thing much longer. Investors understood: We'd hit peak iPhone--which Apple, finally, all but acknowledged in November, when it said it would stop reporting sales numbers for the device.

It's not just Apple. Amazon, Alphabet, Facebook, and Microsoft all are outgrowths of the computing boom. In any given month, these might be the five most valuable companies in the world. Wall Street valuations are supposed to be calculations of future earnings, but these companies are grounded in past growth rates, so their ascendancy today says little about who will rule tomorrow.

That doesn't mean tech companies are doomed to stagnate. The biggest have diversified so much, it's hard to find one rubric to hold them all. Alphabet and Amazon have made big bets on health science and robotics; Apple, Facebook, and Microsoft are well positioned if the hype around virtual and augmented reality pans out; all expect A.I. to change pretty much everything. But it does mean the technologies that got them to this point--mobile computing, web search, cloud storage--won't keep improving at anything like the rate of the past few decades.

One leading contender to become the dominant industry of the 21st century is synthetic biology, where the price of reading and printing DNA (as Ginkgo Bioworks does) is falling exponentially in the same way the price of bits was 50 years ago. But maybe it will be robotics or space exploration or carbon capture. "The asteroid has hit, the dinosaurs are dying off, and the little mammals that were hiding underground are sticking up their heads and saying, 'It's our turn now,' " says Blank.

In Silicon Valley, disruption is supposed to be a good thing, and the end of Moore's law doesn't have to be an extinction-level event. But companies that want to survive it will have to do more than just evolve. They'll have to discover a new way to evolve, and perhaps a new way to be profitable--one befitting a universe where Silicon Valley's innovation cycles are no longer based on silicon.