Life After Google

by George Gilder

  Whereas Google’s “free world” tries to escape the laws of scarcity and the webs of price, you will live in a world brimming with information on the real costs and most efficient availabilities of what you want and need. The proof of your work will trump the claims of top-down speed and hierarchical power. The crude imperatives of “free” will give way to the calibrated voluntary exchanges of free markets and micropayments.

  Whereas the Google world strains you through sieves of diversity and runs you through blenders of conformity, the new world will subsist on the foundational realities of individual uniqueness and choice. Whereas the Google world is stifling entrepreneurs’ access to the public markets through initial public offerings, which are down 90 percent in two decades, the new world will offer an array of new paths to enterprise. From initial coin offerings and token issues to crowd-funded projects, new financial devices are already empowering a new generation of entrepreneurs. The queues of abject “unicorns”—privately held start-ups worth a billion dollars or more—outside the merger and acquisition offices of Google and its rivals will be dispersed, replaced by herds of “gazelles” headed for public markets at last.2

  Whereas Google attempts to capture your eyeballs with ubiquitous advertisements, you will see advertisements at your own volition, when you want them, and you will be paid for your time and attention. Again, Brave is the leader of this movement.

  Money is not a magic wand but a measuring stick, not wealth but a gauge of it. Whereas money in the Google era is fodder for a five-trillion-dollar-a-day currency exchange—that’s seventy-five times the amount of the world’s trade in goods and services—you will command unmediated money that measures value rather than manipulates it. Whereas the Google world is layered with middlemen and trusted third parties, you will deal directly with others around the globe with scant fees or delays.

  Emerging is a peer-to-peer swarm of new forms of direct transactions beyond national borders and new forms of Uber and Airbnb beyond corporate gouges. Whereas the Google world confines you to one place and time and life, the new world will open up new dimensions and options of new life and experience where the only judge is the sovereign you.

  Does the promise that human dignity will once again take its place on the Internet and that human beings will be masters of the cryptocosm sound too good to be true?

  If these principles are enigmatic today, to explain their sources and ultimate success, we must, as Caltech’s Carver Mead tells us, “listen to the technology and find out what it is telling us.”

  CHAPTER 6

  Google’s Datacenter Coup

  The drive up Interstate 84, through the verdant amphitheatrical sweep of the Columbia River Gorge to the quaint Oregon town of The Dalles (rhymes with “pals”), seems like a journey into an alluring American past. Through a filigree of Douglas firs you glimpse ancient basalt bluffs riven by glittering waterfalls. Signs direct you to museums of native Americana, full of feathery and leathery tribal relics. There are farms and fisheries, hillside vineyards, eagles and ospreys riding the winds.

  On the horizon, just a half-hour’s drive away, stands the radiant, snowcapped peak of Mount Hood, site of eleven glaciers, source of half a dozen rivers, and home of four-season skiing. “I could live here,” I say to myself with a backward glance down the highway toward urban Portland. Compared with the billboarded corridor between Silicon Valley and San Francisco, the Columbia valley shimmers as a sylvan dream.

  Then, as the highway comes to an end, the gray ruin of an abandoned aluminum plant rises from a barren hillside. Its gothic gantries and cavernous smelters stand empty and forlorn, a poignant testimony to the evanescence of industrial power.1

  The name The Dalles derives from eighteenth-century voyageur slang for the dangerous nearby rapids on the Columbia River—back when the local industry involved transporting beaver pelts by canoe. Now the beavers are left alone, and the aluminum plants are mostly abandoned, but The Dalles is booming. Here beside the river, six miles west of the dam, Google bought thirty acres in 2005 for the company’s first owned-and-operated data center. The Dalles was to be the spearhead of its new system of the world.

  In nine years, this campus more than tripled in size, when Google (under the name “Moraine Industries”) bought seventy-four more acres in 2014 from the struggling Northwest Aluminum Company. Its total investment in this tiny town mounted close to two billion (out of some $29 billion invested in its global plant). With its cloak-and-lawyer disguise during the negotiations and its lavish charity payments afterwards, Google even managed to make its data centers generally exempt from property taxes.

  The data center itself is wrapped in secrecy, with gates to keep out employees who do not have the correct clearance and airport-style millimeter-wave whole-body scanners for everyone entering the heart of the warehouse. To handle the floods of bits and bytes, each of the three ten-million-cubic-foot glass-walled warehouses of Google’s Dalles fortress now holds 75,000 computer servers, interlinked with fiber-optic lines, arrayed in towering racks.2 These servers, jammed as close together as possible to minimize speed-of-light delays, look like glowing horizontal books shelved in the stacks of a huge futuristic library.

  Although evergreen mazes, mountain majesties, and always-on skiing played a role, two amenities in particular made this an auspicious site for a dominant data center. The first is a fiber-optic hub linked to Harbour Pointe, Washington, two hundred miles northwest of The Dalles, on the far side of Mount Rainier. This is the coastal landing base of the massive cables of PC-1. Named Pacific Crossing 1 by its builder, Gary Winnick’s ill-fated Global Crossing, this network ganglion is a fiber-optic artery built in 2001 to handle 640 gigabits per second (billions of bits per second). Upgraded twelve-fold to 8.4 terabits (trillions of bits per second) a decade later, it connects Asia to the United States across six thousand miles of the Pacific.

  A glassy extension cord snakes through the town’s major buildings, tapping into the greater Internet though NoaNet, a node of the once leading-edge set of standards called Internet2. Under the indomitable Urs Hölzle, Google’s “cloud” careened forward with ten new data centers in 2017 under a still more advanced regime called “Internet3.”

  The other amenity is the Dalles Dam and its 1.8-gigawatt power station. Built in 1957 between Klickitat, Washington, and Wasco, Oregon, by the Army Corps of Engineers for Bonneville Power, the half-mile-long dam channels The Dalles rapids into cheap subsidized electrical power. Once essential to aluminum smelting, it is now a strategic spearhead for computing. Indeed, Google is not the only Silicon Valley titan to depend on the Columbia River, which provides electricity at about a fifth of the cost of power in the San Francisco Bay area.

  This condensation of big data and vast computing power in the “cloud” is unprecedented in the history of computing. These machines gain control of their environment by excelling all others in the speed and density of their computations and transactions and in the size of their data stores.3 Such servers are behind new centers of dominance in industries as diverse as retailing and finance, insurance and real estate. But Google’s are the most dominant of all (save, perhaps, by the measure of profitability, a financial rival in the East).

  Moore’s Law, which describes the growth in capacity of integrated circuits, has a corollary named after Gordon Bell, the legendary engineer behind Digital Equipment Corporation’s breakthrough VAX line of minicomputers of the 1980s and now a principal researcher at Microsoft.4 According to Bell’s Law, every decade a hundredfold drop in the price of processing power engenders a new computer architecture.

  Just last century—you remember it well, across the chasm of two economic crashes—the PC was king. Deposed and deceased was the lordly computer mainframe, which had sustained the dominance of IBM in information technology in the 1970s and the Digital Equipment and Data General minicomputers and their client-server systems of the 1980s.5

  Google’s cloud defines the current Bell’s Law regime. B
ut as recently as the late 1990s, Larry Page and Sergey Brin were nonprofit googoos working in the Gates Center at Stanford seeking to search their 150-gigabyte index of the Internet. At the time, when I wanted to electrify crowds with my uncanny sense of the future, I would talk terascale (10 to the twelfth power), describing a Web with an unimaginably enormous total of 15 trillion bytes of content.

  The Google worldwide warehouse arises from this once futuristic terabyte paradigm, but its operating environment is now the peta-scale—petabytes, petaops, petaflops. “Peta” means a quadrillion (that is, 10 to the fifteenth power, a million billion) but also, by felicitous coincidence, evokes petere, the Latin verb “to search.” Today Google reigns over a database of thousands of petabytes, called exa-bytes, swelled every twenty-four hours by scores of terabytes of Gmails, Facebook pages, presidential twitter feeds, and videos—a relentless march of daily deltas, each larger than the whole Web of a decade ago. Google handles a billion YouTube videos and 3.5 billion searches every day and 1.5 trillion searches per year. Doubling annually, its internal bandwidth was up fifty times in six years through 2014 and is expected to expand another tenfold through 2018. According to Google’s operations chief, Hölzle, that number will surge another tenfold within another two years.6

  Replicated around the globe, this Bell’s Law machine in The Dalles is at the heart of Google’s hegemony. It is the coup on the Columbia that underlies Google’s supremacy.

  Back in 1993, in a midnight email to me from his office, Eric Schmidt, then the CTO of Sun Microsystems, described the future: “When the network becomes as fast as the processor, the computer hollows out and spreads across the network.” Sun publicized this notion in a compact phrase: “The network is the computer.” But Sun’s hardware honchos failed to absorb Schmidt’s CEO-in-the-making punch line. In which direction would the profits from that transformation flow? “Not to the companies making the fastest processors or best operating systems.” At the time that would have been Sun with its SPARCstations, its RISC (reduced instruction set computer), its Java virtual machines, its Solaris OS—all competing with the rising leviathan Microsoft and the still ascendant IBM. No, Schmidt wrote in his midnight email, the profits would flow to “the companies with the best networks and the best search and sort algorithms.”7

  This insight I dubbed Schmidt’s Law. Schmidt was not just a midnight email doodler. He soon left Sun and, after a stint as CEO of Novell trying to build the best networks and search engines in Utah, he joined Google and soon became CEO. There he found himself engulfed by the future he had predicted. While competitors like Excite, Inktomi, AltaVista (DEC), and Yahoo were building out their networks with SPARCstations and IBM mainframes, Google designed and manufactured its own servers from cheap commodity components made by microprocessor star Intel and hard-drive king Seagate.

  In a 2005 technical article, Google’s operations chief, Hölzle, explained why. The price of high-end processors “goes up nonlinearly with [their] performance,” he observed. That is, Intel’s high-end microprocessors cost increasingly more than they are worth in incremental output. The chips hit what might be called Mundie’s Wall. When he was Microsoft’s technical chief, Craig Mundie said:

  We have now run into a brick wall. What brought all of us faster computing was raising the CPU’s clock rate [its speed per compute cycle measured in hertz or cycles per second]. A faster clock increased power consumption. We could only increase the clock rate without consuming more power because we could lower the voltage. But we can’t do that anymore because we’re down into electron volts where quantum uncertainty takes over. If you can’t lower the voltage, you can’t raise the clock rate without using a lot more power.

  It is harder to accelerate the clock rate and reduce the heat emissions than it is to multiply transistors storing bits in memory chips. As memory grows more rapidly than microprocessor operations, faster microprocessors tend to bog down in memory accesses. Hölzle’s solution, driven by Larry Page, was promising: jamming together innumerable cheap processors in parallel and linking them with fiber-optic lines at the speed of light. Ingenious new software made them work together. That was at least a theoretical path to a scalable system in which bang for the buck didn’t diminish as the system grew.

  Today, Hölzle’s architecture, embodying Schmidt’s insight, has been vindicated, conferring on Google its global sway. Schmidt is often seen in elite circles from Aspen to Davos to Cannes wearing the goofy grin of a comp-sci geek who has become a master of the universe.

  The essential first step in the coup, the plant in The Dalles, was the product of what Schmidt calls “some of the best computer science ever performed.” By building its own infrastructure rather than relying on commercial data centers, Google got “tremendous competitive advantage,” Schmidt told analysts at the time.

  In every era, the winning companies are those that waste what is abundant—as signaled by precipitously declining prices—to save what is scarce. Google has been profligate with the surfeits of data storage and backbone bandwidth. Conversely, it was parsimonious with that most precious of resources, users’ patience with delay—how long you wait for a webpage or a search result.

  The continuing explosion of hard disk storage capacity makes Moore’s Law look like a cockroach race. In 1981, a gigabyte drive cost $500,000, and an Intel 286 processor ran at six megahertz and cost $360. In 2018, a gigabyte costs less than two cents and a three-gigahertz processor costs roughly three thousand dollars. In constant dollars, the price of processing has dropped some five hundred–fold, while the price of a hard drive has dropped 250,000 times. By this crude metric, the cost-effectiveness of hard drives grew five hundred times faster than that of processors.

  You would think that the cost-conscious folks at Google would have filled their warehouses with hard drives. But the miraculous advance of disk storage concealed a problem: The larger and denser the individual disks, the longer it takes to scan them for information. The little arm reading the disks can’t move fast enough to keep up with the processor.

  Google’s solution was to deploy huge amounts of fast random access memory chips. By the byte, RAM is some one hundred times more costly than disk storage. Engineers normally conserve it obsessively, using all kinds of tricks to fool processors into treating disk drives as though they were RAM. But Google understands that the most precious resource is not money but time. Search-users, it turns out, are sorely impatient. Research shows that they’re satisfied with results delivered within a twentieth of a second. RAM can be accessed some ten thousand times faster than disks. Measured by access time, then, RAM is one hundred times cheaper than disk storage. So Google has long led the world in the use of RAM.

  It’s not enough to reach users quickly. Google needs to reach them wherever they are. This requires access to the Net backbone, the long-haul fiber-optic lines that encircle the globe. Google interconnects its hundreds of thousands of processors with hundred-gigabit-per-second Ethernet lines, now moving up to four hundred gigabits. Placing gigantic data centers near major fiber-optic nodes is well worth the expense.

  Wasting what is abundant to conserve what is scarce, the G-men have become the supreme entrepreneurs of the new millennium. It is the Google era. But hovering over the massively parallel, prodigally productive petascale computer like a midday emanation over Death Valley is a shimmering haze of heat.

  Air conditioning will be the prime cost and conundrum of the petascale era. After taking his post in 1999, Google’s Hölzle noticed the high electric bills. At fifteen cents per kilowatt-hour, power dominated his calculus of costs. “A power company could give away PCs and make a substantial profit selling power,” he said. At The Dalles, the huge protuberances on the roof are not giant disk drives but cooling towers. Pipes painted in Google’s signature colors snake through the warehouses beneath them, water-cooling the air.

  Hydropower is a limited and localized resource, while nuclear power promises centuries of nearly limitless energy th
at can be produced almost anywhere. China plans to build as many as forty newfangled nuclear plants; the next wave of data centers may well be in Shenzhen.

  For now, though, Google has attained one of the holy grails of computer science: a scalable massively parallel architecture that can accommodate diverse software while poring through petabytes of big data. Its petascale search machine in place, Google then faced the question: What else could it do? Google’s answer: just about anything. Thus the company’s expanding portfolio of Web services: delivering ads (AdSense, AdWords), maps (Google Maps), videos (YouTube), scheduling (Google Calendar), documents (Google Docs), transactions (Google Checkout), translations (Google Translate), email (Gmail), and productivity software (Writely), to name a few. The other heavyweights have tried to follow suit.

  Our CPUs—those of our PCs, amplified by billions of smartphones—are both more powerful and less employed than ever. Google and the others suck into their proprietary clouds more and more of the duties once delegated to the CPU. Optical networks, which move data over vast distances without degradation, allow computing to migrate to wherever power is cheapest. The new computing architecture thus scales across the earth’s surface. As I write in 2018, what is called the “cross-section bandwidth” of the internal networks spanning Google’s data centers has reached petabytes per second—a multiple of the total bandwidth of the entire Internet that Google searches and sorts, mines and monetizes. And it will never be enough.

  The Googleplex at the center of the sphere will soon dwarf the Internet itself. Introducing Google’s networking technology leader, Amin Vahdat, in October 2015, the magazine of the Association for Computing Machinery declared, “Everything about Google is at scale, of course—a market cap of legendary proportions, an unrivaled talent pool, enough intellectual property to keep armies of attorneys in Guccis for life, and—oh yeah—a private Wide Area Network (WAN) bigger than you can imagine that also happens to grow faster than the Internet.”