Chapter 1
The Internet's Hidden Backbone: Where Digital Meets Physical
Ever wondered what happens when you send an email or upload a photo? In 2010, author Andrew Blum's internet connection failed because a squirrel had chewed through his cable in Brooklyn. This simple incident sparked a two-year global odyssey to discover what the internet physically is. While we think of the internet as an abstract cloud, Blum reveals it as a tangible network of cables, data centers, and exchange points spanning continents and oceans. His book "Tubes" arrived at a pivotal moment-just before Edward Snowden's 2013 revelations about government surveillance would forever change our relationship with tech companies. Once celebrated as a utopian information superhighway, the internet's physical infrastructure has since become both more secretive and, paradoxically, more visible, with Google commissioning artistic murals for once-unmarked data centers and Chanel even staging a fashion show with a data center theme. Blum's journey takes us behind locked doors to see what most of us never will: the actual, physical internet that powers our digital lives.
Chapter 2
Following the Wires: The Internet's Physical Geography
On that freezing winter day in Brooklyn when my internet connection began failing, the repairman traced the cable from my living room to a rusty switch box in the backyard. Looking up, he spotted a squirrel scampering along a wire toward a gray enclosure on a pole. This simple animal had temporarily defeated the revolutionary global network used by billions daily. This vulnerability revealed something profound: despite existing vividly on screens everywhere, the internet physically exists as tubes, wires, and machines in specific places.
This realization led me to Milwaukee, where I watched cartographer Markus Krisetya examine a test print of his latest creation: a map of the internet's undersea cables. These physical connections between continents appeared as rigid red and yellow lines striping across oceans, each representing cables mere inches in diameter but thousands of miles long. The irony wasn't lost on me that this digital map was being produced as a physical object in a city that knew something about how the world was made.
Through conversations with network engineers, I discovered the internet has layers of depth - multiple networks run through the same wires despite being owned by different organizations. Networks carry networks. One company might own the fiber-optic cables, another operates the light signals, and a third owns or rents the bandwidth. This overlapping meant abandoning the misleading "information highway" metaphor. The internet exists in three realms: the logical (how signals travel), the physical (machines and wires), and the geographical (the places reached).
In Milwaukee, network engineer Jon Auer showed me through a 1901 building that served as the city's internet hub. In the basement, a tangle of pipes and wires hung from the ceiling - steel conduits, orange plastic ducts, and dangling black threads carrying fiber-optic cables as pulses of light. Upstairs, his piece of the internet sat on a raised platform: two steel racks with a half-dozen machines nestled in cables. The key equipment was a black Cisco router that directed data along two paths: one yellow wire went to Cogent's equipment room upstairs, while another connected to Time Warner as backup. For his customers, this simple fork in the road was their source - two yellow cables leading to the world.
Chapter 3
The Birth of a Network: From ARPANET to Internet
The internet's history is surprisingly underwritten, with most serious book-length accounts published around 1999. Unlike other technologies with central founding figures like Edison or Morse, the internet lacks a single creator, and its distributed nature complicates any linear historical narrative. Rather than asking where the internet began conceptually, I sought something more concrete: where was its first physical box?
In summer 1969, the first Interface Message Processor (IMP) was installed at UCLA under Professor Leonard Kleinrock's supervision. On Labor Day weekend, graduate students gathered with champagne in Boelter Hall's courtyard to welcome the arrival of this modified Honeywell DDP-516 minicomputer - a 900-pound, $80,000 machine that represented the first physical piece of what would become the internet. The machine arrived from Bolt, Beranek and Newman in Cambridge, Massachusetts, which had a $1 million Department of Defense contract to build the experimental ARPANET.
Decades later, Kleinrock showed me the historic IMP #1 - the refrigerator-sized, beige steel cabinet that served as the first node of the ARPANET. "This is it-a beautiful machine! A magnificent machine!" he exclaimed, opening cabinets and twisting dials. Despite its historical significance, the IMP had nearly been discarded before Kleinrock rescued it, now sitting unceremoniously in a small conference room. He pulled out the original tan logbook recording the first successful connection between UCLA and Stanford Research Institute on October 29, 1969, with the simple handwritten note: "Talked to SRI host to host" - the internet's first breath. "In those early days, none of us had any idea what it would become," Kleinrock reflected. "What I missed was the social side-that my ninety-nine-year-old mother would be on the internet."
The internet's adolescence was marked by fragmentation before unity. By the early 1980s, numerous specialized networks had emerged - each with its own acronym. The watershed moment came on New Year's Day 1983, when ARPANET hosts switched from Network Control Protocol to TCP/IP - a transition so challenging that system administrators made "I SURVIVED THE TCP/IP TRANSITION" buttons. This standardization permanently established the internet's distributed structure and lack of central control.
With TCP/IP in place, explosive growth followed. From just 15 autonomous networks in 1982, there were over 400 by 1986. Connected computers grew from 2,000 in 1985 to 159,000 by 1989. The internet was becoming a place, with companies building their own backbones and metropolitan networks. Key hubs emerged in suburban Virginia, Silicon Valley, London's Docklands, Amsterdam, Frankfurt, and Tokyo's Otemachi district.
Chapter 4
The Meeting Places: Where Networks Connect
In 1993, the commercial internet faced a crucial problem - providers couldn't legally use the government-run NSFNET backbone for commercial traffic, yet needed to exchange data with each other. The solution emerged over lunch at the Tortilla Factory in Herndon, Virginia, where leaders from Sprint, PSI, and UUNET agreed to connect through MFS's proposed hub.
MFS created the Metropolitan Area Exchange (MAE-East) in Tysons Corner, Virginia, which quickly became the internet's central crossroads. "If you connected to MAE-East, you'd have the entire internet at your doorstep," explained Steve Feldman of MFS Datanet. Within a few years, MAE-East handled half of all global internet traffic - messages from London to Paris, Tokyo to Stockholm all passed through the fifth floor of 8100 Boone Boulevard.
By 1996, MAE-East was overcrowded as networks discovered they could improve performance by installing their actual routers and servers there. What began as a crossroads had become a depot. Feldman moved the operation to a humble plasterboard enclosure in a basement parking garage across the street. With internet traffic doubling yearly and perpetual congestion, a new model was needed.
That model emerged in Silicon Valley, where Jay Adelson and Brian Reid created the Palo Alto Internet Exchange (PAIX) in 1996. Reid's revolutionary idea was that networks should connect directly to each other rather than through a central switch. Digital Equipment Corporation provided a few million dollars and space in the basement of 529 Bryant Street, a 1920s telephone switching office. The business model changed too - instead of charging for traffic, they charged rent for cage space.
The PAIX quickly became the internet's most important building. Walking through the facility reveals the physical internet: hundreds of cages filled with routers, thick bundles of cables suspended from ceilings in "waterfalls," and over ten thousand cross-connections between networks. Each fiber-optic cable carries up to ten gigabits per second - enough to transmit ten thousand family pictures every second.
The building thrives on diversity, hosting networks from Microsoft and Google to global telecoms from Singapore to Qatar. These connections are both physical and social, often arranged through personal relationships between network engineers. The facility demonstrates the "network effect" - the more networks that joined, the more valuable it became.
Chapter 5
The Internet's Crossroads: Exchanges and Peering Points
The PAIX's success revealed both the viability of carrier-neutral exchanges and their limitations. When Digital was acquired by Compaq in 1998, Adelson saw opportunity. He and colleague Al Avery raised $12.5 million to create Equinix-a chain of "telecom hotels" where networks could interconnect neutrally and efficiently. Rather than building in congested Tysons Corner, they chose farmland in Ashburn, Loudoun County, Virginia, where officials eagerly offered rights-of-way for fiber connections. This decision would prove prescient, as Ashburn's location offered ideal conditions: proximity to major population centers, affordable land, and cooperative local government.
Ashburn transformed from rural farmland into what internet professionals now consider a major global hub, mentioned alongside London and Tokyo. The area became known as "Data Center Alley," hosting over 70% of the world's internet traffic. Equinix's unmarked complex sits behind an Embassy Suites hotel, deliberately nondescript but rapidly expanding. The facility's understated exterior belies its critical importance to global communications. Inside, the environment resembles Vegas-timeless black ceilings, spotlights, and amenities for the "server huggers" who spend their days tending to equipment. These technicians work in shifts around the clock, maintaining the vital infrastructure that keeps the internet running.
The data center interior feels like stepping into a machine-cold to offset equipment heat and deafeningly loud from cooling fans that maintain precise temperature control. Blue spotlights illuminate mesh cages with handprint scanners, creating both drama and practical privacy. Each cage houses servers from different companies, maintaining strict security protocols. Above, a complex layered ceiling infrastructure resembles an archaeological record-yellow fiber-optic cables in plastic raceways, copper "whalebone" racks, power systems, and finally the critical "innerduct" carrying massive fiber bundles from major carriers. This layered infrastructure tells the story of internet evolution, with newer technology built atop legacy systems.
After seeing Equinix's massive data center in Ashburn, I described my experience to my eight-year-old niece, telling her I had seen "the internet" or at least a significant piece of it. Unlike skeptical adults who struggle to grasp the internet's physical reality, she accepted this naturally. Having never known a world without the internet, the idea that it exists as something tangible made perfect sense to her. This generational difference in understanding reflects how our perception of technology evolves with each new generation.
I deliberately capitalize "Internet" as a proper noun, even as convention has shifted away from this practice. My exploration of its physical infrastructure reveals something singular and tangible, however amorphous. The Internet exists in specific places - not just behind locked doors in unmarked buildings, but everywhere in wires and towers. From suburban fiber-optic cables to rural cell towers, from undersea cables to satellite links, this vast network forms a single, interconnected system that deserves proper noun status. The physical Internet represents one of humanity's greatest collaborative achievements, a network that spans continents and connects billions of people.
Chapter 6
Global Connections: The Internet's Exchange Points
The answer to where the internet's most important "boxes" exist lies within Internet exchanges (IXs). These serve as critical meeting points where networks connect, either as physical buildings or as organizations facilitating connections across multiple locations. Their fundamental rationale is straightforward: get data packets to their destination as directly and cheaply as possible by increasing available paths. These exchanges function like massive digital Grand Central Stations, routing millions of data packets every second through optimal pathways.
The largest Internet exchanges are sophisticated professional operations with dedicated marketing departments and specialized engineering teams, intensely competitive with one another. The three biggest are European: DE-CIX in Frankfurt, which handles over 11 terabits per second during peak times, AMS-IX in Amsterdam, serving more than 875 networks, and LINX in London, connecting over 950 networks from 80 countries. Moscow's MSK-IX has been rapidly gaining ground, particularly as a crucial hub for Eastern European traffic. These exchanges actively post live traffic graphs and network member tallies on their websites, creating a horse-race atmosphere as they compete for prominence and market share in the interconnection industry.
The concentrated nature of these Internet exchanges raises significant security questions. After touring sensitive infrastructure - including reinforced concrete walls, biometric scanners, and military-grade security systems - I experienced momentary paranoia about knowing too much. During an interview, an executive directly challenged me: "Are we creating through this book a road map for terrorists?" Despite initial concerns about investigating these facilities, I discovered something surprising: network operators were consistently open, even philosophical about sharing information. This transparency stems from two factors: well-designed networks have multiple redundancies and backup systems, and the Internet is fundamentally a public infrastructure built on shared protocols and standards.
The connections between networks, known as "peering," are complex agreements to interconnect two networks that are ostensibly peers of equal size and status. The peering process has a Talmudic quality, with intricate laws and precedents requiring years of study to understand properly. These arrangements involve detailed negotiations about traffic ratios, connection speeds, and technical specifications. The stakes are enormous: peering allows information to flow freely and cheaply across the Internet; without it, services like YouTube, Netflix, and social media platforms might require subscription fees to cover increased data transit costs.
Peering relationships are both technical and deeply personal, often built through face-to-face meetings at industry conferences and maintained through years of professional interaction. The connection between Google and Comcast, for example, exists not just in physical links at places like Ashburn, but in the relationship between peering coordinators who negotiate terms and troubleshoot issues. Peering can turn ugly when major networks "de-peer," literally pulling the plug on connections. In 2008, Sprint stopped peering with Cogent for three days, causing 3.3% of global Internet addresses to "partition" - a dramatic demonstration of how these business relationships directly affect Internet accessibility. Similar disputes between Level 3 and Comcast, and Cogent and France Telecom, have highlighted the fragile nature of these crucial interconnections.
Chapter 7
Under the Sea: The Internet's Submarine Cables
The underwater telecommunications cables form the fundamental medium of our global Internet village, connecting continents through simple yet fantastically complex tubes carrying light across ocean floors. When South Africa gained access to the SEACOM cable in 2009, Andrew Alston's university network went from a meager 250-megabit connection to ten gigabits at the same price, demonstrating how these physical connections directly impact a nation's digital capabilities.
In an unmarked New Jersey office park, Tata Communications manages what was once the Tyco Global Network - 40,000 miles of fiber-optic cable spanning three continents. Simon Cooper, Tata's roaming executive, explained their strategy to buck industry trends by connecting the "global south" - less-connected regions in Africa and South Asia. Tata has built what Cooper casually describes as "a belt around the world," stringing cables from the US to Japan to Singapore to Chennai, then from Mumbai through Suez to Marseille, overland to London, and across the Atlantic back to New Jersey - literally creating a beam of light encircling the planet.
I journeyed to Porthcurno, Cornwall - the spiritual home of undersea cables near England's westernmost point. For 150 years, this small cove has been a crucial landing spot for underwater communications cables and a training ground for the industry. Porthcurno nestles in a valley ending at a spectacular turquoise-watered beach beneath high cliffs. The Falmouth, Gibraltar, and Malta Telegraph Company landed its first cable here in 1870, choosing it over busier ports to avoid anchor damage. By 1900, it had become the hub of a global telegraph network linking India, the Americas, South Africa, and Australia.
Walking through Porthcurno, I discovered how the physical Internet manifested in this rural landscape - manholes labeled "ductile," humming equipment compounds behind wooden fences, and giant rhubarb plants seemingly nourished by the light passing beneath them. That evening, Skyping with my wife in New York, I ran a traceroute and realized our conversation was passing directly beneath the B&B where I stayed, through cables connecting Europe and America.
Chapter 8
Where Data Lives: The Cloud's Physical Home
The Dalles, Oregon has always been a geographical chokepoint where infrastructure must adapt to nature's demands. Named from the French word for "flagstone" referring to rocks in the Columbia River, it marks where the mighty river narrows before cutting through the Cascade Range. This natural bottleneck has made it a crucial crossroads throughout history.
I came to The Dalles seeking one of the Internet's most important data repositories - the de facto capital of a region devoted to storing our online selves. What we poetically call the "cloud" is actually consuming 2 percent of global electricity, growing 12 percent annually. A large data center might occupy 500,000 square feet demanding fifty megawatts - enough to power a small city.
The question of where data sleeps remains remarkably open-ended. While smaller data centers prioritize convenience, larger facilities face complex location decisions that ultimately aim to minimize the cost of keeping hard drives spinning and cool. Engineers compete to design buildings with the lowest "power usage effectiveness" (PUE), with location being a crucial variable.
In the late 1990s, the Bonneville Power Administration began mounting fiber-optic cables along its transmission lines, creating a network spanning the Northwest that converged in The Dalles. This challenging installation, often requiring helicopters to string cable on towers in rough terrain, yielded far more fiber than the power company needed internally. Despite telecommunications companies' objections, BPA began leasing this excess fiber - creating a robust, protected communications system that proved irresistible to data center developers.
Around 2000, despite being a historic crossroads with major fiber backbones running through it, The Dalles lacked high-speed internet access. The town's response was bold: "We'll do it ourselves." In 2002, they chartered the Quality Life Broadband Network ("Q-Life") as an independent utility, constructing a seventeen-mile fiber loop connecting city hall to the BPA's Big Eddy substation.
In 2004, just after Q-Life's completion, a young, sloppily dressed man named Chris Sacca arrived in The Dalles representing the suspiciously generic "Design LLC." He sought shovel-ready sites in enterprise zones with tax incentives, requiring astronomical quantities of power. The Dalles offered thirty acres near a decommissioned aluminum smelter. Negotiations centered on power and taxes, not land costs. The secret was eventually revealed: Design LLC was Google.
Chapter 9
The Human Internet: People Behind the Machines
Despite Google's supposed new openness about its facilities, my visit to their data center proved disappointingly opaque. The campus resembled a prison with its security lights, beige buildings, and perimeter fence. A sign reading "VOLDEMORT INDUSTRIES" in gothic script-referencing Harry Potter's "He-Who-Must-Not-Be-Named"-hinted at the continuing culture of secrecy.
The contrast between Google's secretive data center and the Bonneville Dam was striking. The dam, though a fortress with armed guards, welcomed visitors with a gift shop, exhibits about its construction and ecology, and viewing areas for salmon swimming upstream. This classic American roadside attraction blended big government with big landscape, technological triumph with environmental awareness.
Leaving The Dalles, I drove south toward Prineville, where Facebook had built its first ground-up data center. Unlike Google's secretive approach, Facebook arrived with fanfare - "WELCOME FACEBOOK" signs adorned store windows downtown. The data center occupied a butte overlooking town, marked by white-and-orange fiber markers leading to its entrance. Unlike Google's aesthetically loose buildings, Facebook's facility was surprisingly beautiful - long and low like a Greek temple on the landscape, a crisp concrete slab topped with corrugated steel.
Inside Facebook's facility, I met Ken Patchett, the data center manager who'd previously held the same job at Google's Dalles facility. Unlike the silent Google employees, Patchett was uncensored - an enormous extrovert with a booming voice. Patchett led me through Facebook's Prineville facility, starting in a glass-walled lobby decorated with historical photos of local residents. We entered massive data halls with cavernous ceilings where thousands of servers lined up in neat rows, each with fluttering blue lights representing terabytes of data. "This is the cloud," Patchett declared emphatically. "If you blew the 'cloud' away, you know what would be there? This. The cloud is a building. It works like a factory."
Back home after my journey across the physical Internet, I reflected on how strange it felt to use my laptop normally again - checking email, reading blogs - after having seen the actual infrastructure behind these everyday actions. The Internet's seamlessness made it easy to forget the physical reality I'd witnessed.
Despite science fiction writer Bruce Sterling's claim that with broadband, one's "position on the planet's surface is arbitrary," I'd discovered we're not merely connected but rooted. My own connection to the Internet began with a cable from my apartment that joined others in a metal enclosure on my Brooklyn street - node 8M48 on Cablevision's network. From there, fiber optic cables led to a local "head-end" building, then to the main facility in Hicksville, Long Island, before connecting to familiar exchange points: 60 Hudson Street, 111 Eighth Avenue, and Equinix data centers.
What I ultimately understood was that the Internet isn't a physical world or a virtual world, but a human world. The Internet's physical infrastructure has many centers, but from a certain vantage point there is really only one: You. Me. The lowercase i. Wherever I am, and wherever you are.
