In the long term, my departure from Silver Lake worked out for everyone. The second Silver Lake fund got off to a rocky start, as my cofounders struggled with stock picking, but they figured it out and built the firm into an institution that has delivered good investment returns to its investors.
Silicon Valley Before Facebook
I think technology really increased human ability.
But technology cannot produce compassion. —DALAI LAMA
The technology industry that gave birth to Facebook in 2004 bore little resemblance to the one that had existed only half a dozen years earlier. Before Facebook, startups populated by people just out of college were uncommon, and few succeeded. For the fifty years before 2000, Silicon Valley operated in a world of tight engineering constraints. Engineers never had enough processing power, memory, storage, or bandwidth to do what customers wanted, so they had to make trade-offs. Engineering and software programming in that era rewarded skill and experience. The best engineers and programmers were artists. Just as Facebook came along, however, processing power, memory, storage, and bandwidth went from being engineering limits to turbochargers of growth. The technology industry changed dramatically in less than a decade, but in ways few people recognized. What happened with Facebook and the other internet platforms could not have happened in prior generations of technology. The path the tech industry took from its founding to that change helps to explain both Facebook’s success and how it could do so much damage before the world woke up.
The history of Silicon Valley can be summed in two “laws.” Moore’s Law, coined by a cofounder of Intel, stated that the number of transistors on an integrated circuit doubles every year. It was later revised to a more useful formulation: the performance of an integrated circuit doubles every eighteen to twenty-four months. Metcalfe’s Law, named for a founder of 3Com, said that the value of any network would increase as the square of the number of nodes. Bigger networks are geometrically more valuable than small ones. Moore’s Law and Metcalfe’s Law reinforced each other. As the price of computers fell, the benefits of connecting them rose. It took fifty years, but we eventually connected every computer. The result was the internet we know today, a global network that connects billions of devices and made Facebook and all other internet platforms possible.
Beginning in the fifties, the technology industry went through several eras. During the Cold War, the most important customer was the government. Mainframe computers, giant machines that were housed in special air-conditioned rooms, supervised by a priesthood of technicians in white lab coats, enabled unprecedented automation of computation. The technicians communicated with mainframes via punch cards connected by the most primitive of networks. In comparison to today’s technology, mainframes could not do much, but they automated large-scale data processing, replacing human calculators and bookkeepers with machines. Any customer who wanted to use a computer in that era had to accept a product designed to meet the needs of government, which invested billions to solve complex problems like moon trajectories for NASA and missile targeting for the Department of Defense. IBM was the dominant player in the mainframe era and made all the components for the machines it sold, as well as most of the software. That business model was called vertical integration. The era of government lasted about thirty years. Data networks as we think of them today did not yet exist. Even so, brilliant people imagined a world where small computers optimized for productivity would be connected on powerful networks. In the sixties, J. C. R. Licklider conceived the network that would become the internet, and he persuaded the government to finance its development. At the same time, Douglas Engelbart invented the field of human-computer interaction, which led to him to create the first computer mouse and to conceive the first graphical interface. It would take nearly two decades before Moore’s Law and Metcalfe’s Law could deliver enough performance to enable their vision of personal computing and an additional decade before the internet took off.
Beginning in the seventies, the focus of the tech industry began to shift toward the needs of business. The era began with a concept called time sharing, which enabled many users to share the use of a single computer, reducing the cost to everyone. Time sharing gave rise to minicomputers, which were smaller than mainframes but still staggeringly expensive by today’s standards. Data networking began but was very slow and generally revolved around a single minicomputer. Punch cards gave way to terminals, keyboards attached to the primitive network, eliminating the need for a priesthood of technicians in white lab coats. Digital Equipment, Data General, Prime, and Wang led in minicomputers, which were useful for accounting and business applications but were far too complicated and costly for personal use. Although they were a big step forward relative to mainframes, even minicomputers barely scratched the surface of customer needs. Like IBM, the minicomputer vendors were vertically integrated, making most of the components for their products. Some minicomputers—Wang word processors, for example—addressed productivity applications that would be replaced by PCs. Other applications survived longer, but in the end, the minicomputer business would be subsumed by personal computer technology, if not by PCs themselves. Main-frames have survived to the present day, thanks in large part to giant, custom applications like accounting systems, which were created for the government and corporations and are cheaper to maintain on old systems than to re-create on new ones. (Massive server farms based on PC technology now attract any new application that needs mainframe-class processing; it is a much cheaper solution because you can use commodity hardware instead of proprietary mainframes.)
ARPANET, the predecessor to today’s internet, began as a Department of Defense research project in 1969 under the leadership of Bob Taylor, a computer scientist who continued to influence the design of systems and networks until the late nineties. Douglas Engelbart’s lab was one of the first nodes on ARPANET. The goal was to create a nationwide network to protect the country’s command and control infrastructure in the event of a nuclear attack.
The first application of computer technology to the consumer market came in 1972, when Al Alcorn created the game Pong as a training exercise for his boss at Atari, Nolan Bushnell. Bushnell’s impact on Silicon Valley went far beyond the games produced by Atari. He introduced the hippie culture to tech. White shirts with pocket protectors gave way to jeans and T-shirts. Nine to five went away in favor of the crazy, but flexible hours that prevail even today.
In the late seventies, microprocessors made by Motorola, Intel, and others were relatively cheap and had enough performance to allow Altair, Apple, and others to make the first personal computers. PCs like the Apple II took advantage of the growing supply of inexpensive components, produced by a wide range of independent vendors, to deliver products that captured the imagination first of hobbyists, then of consumers and some businesses. In 1979, Dan Bricklin and Bob Frankston introduced VisiCalc, the first spreadsheet for personal computers. It is hard to overstate the significance of VisiCalc. It was an engineering marvel. A work of art. Spreadsheets on Apple IIs transformed the productivity of bankers, accountants, and financial analysts.
Unlike the vertical integration of mainframes and minicomputers, which limited product improvement to the rate of change of the slowest evolving part in the system, the horizontal integration of PCs allowed innovation at the pace of the most rapidly improving parts in the system. Because there were multiple, competing vendors for each component, systems could evolve far more rapidly than equivalent products subject to vertical integration. The downside was that PCs assembled this way lacked the tight integration of mainframes and minicomputers. This created a downstream cost in terms of training and maintenance, but that was not reflected in the purchase price and did not trouble customers. Even IBM took notice.
When IBM decided to enter the PC market, it abandoned