The Bandwidth Thieves Who Built Silicon Valley: A Financial History of Stolen T1 Lines

There's a version of Silicon Valley's origin story that involves garages, vision, and the courage to dream big. That version is technically accurate and almost entirely incomplete. The fuller version involves a lot of network closets that probably shouldn't have been unlocked, corporate T1 lines being quietly shared with people who had no contractual right to them, and a cultural consensus that this was not theft so much as infrastructure optimization.

The early internet ran on borrowed bandwidth. Not metaphorically. Literally. And the moral framework that permitted this — that celebrated it, even — left fingerprints on Silicon Valley's risk culture that are still visible today.

What a T1 Line Actually Was

For anyone who came of age after broadband was a commodity, it's worth establishing what was actually at stake. A T1 line in 1995 delivered 1.544 megabits per second of dedicated bandwidth and cost somewhere between $1,000 and $1,500 per month, plus installation, plus whatever the local telco felt like charging for the local loop. For a startup operating out of a spare bedroom or a rented desk in a shared office, that was an impossible number.

Universities had T1 lines. Large corporations had T1 lines. Government contractors had T1 lines. These institutions had negotiated their connections years earlier, when the internet was primarily an academic tool, and their contracts were written for a world where the only people using the connection were employees and students with legitimate institutional affiliations.

That world did not survive contact with the mid-1990s internet boom.

The Colo Closet Economy

Colocation facilities — colo, in the vernacular — were the physical infrastructure that made the bandwidth economy possible. A colo facility was, in its simplest form, a room with power, cooling, and network connectivity where you could park a server. The legitimate version of this business is still thriving. The 1990s version had a somewhat looser interpretation of who was allowed to connect to what.

Former network administrator Dave K., who worked at a mid-sized ISP in the late 1990s and asked us not to use his last name for reasons that remain entirely obvious, described the environment with cheerful frankness. "There were definitely situations where a rack would go in, and the paperwork for the uplink would be... informal. Someone knew someone. The connection existed. Questions were not encouraged."

This wasn't always straightforward theft. The spectrum ran from genuinely criminal unauthorized access all the way to elaborate arrangements where a university employee would quietly extend the institution's connectivity to a friend's startup, technically violating their employment agreement and the university's network policies but not necessarily any criminal statute. The IRC networks that facilitated these arrangements — private channels on EFnet and DALnet where sysadmins and early entrepreneurs overlapped — were the deal-making infrastructure for an entire gray economy.

IRC as the Original AngelList

The deal flow happened in channels. Not always explicitly — nobody posted "seeking stolen T1 line, will trade server space" in a channel topic — but the relationships that made these arrangements possible were built and maintained in IRC.

The community of people who administered university networks, corporate server rooms, and early ISPs was small and heavily interconnected. They all knew each other from the same channels, the same Usenet groups, the same DEF CON hallways. When someone needed bandwidth and someone else had access to bandwidth that wasn't being fully utilized, the conversation happened in a query window. The terms were set. The handshake was digital.

Kev, a former IRC operator who ran a small hosting operation in the late 1990s out of what he described as "a very understanding university's network," put it this way: "The bandwidth existed. It was going to exist whether we used it or not. The university wasn't losing anything. We were just using capacity that would otherwise be wasted."

This argument — that unused capacity isn't really property in any meaningful sense — was so pervasive in the early internet community that it barely registered as rationalization. It was just the obvious truth, as far as most participants were concerned. The fact that it was also exactly the argument you'd want to make if you were stealing something did not cause widespread reflection.

The Startups That Benefited

Not every company built on borrowed bandwidth was a fly-by-night operation. Some of them grew. Some of them raised real money. Some of them went public.

The paper trail here is, for obvious reasons, thin. Companies that went public in the late 1990s were not filing disclosures that said "early infrastructure was provided through informal arrangements with university network administrators who were technically violating their employment agreements." The S-1 filings of the dot-com era are full of confident language about scalable infrastructure and strategic partnerships. The channels those partnerships were built in are not mentioned.

But the pattern was widely understood within the industry. Investors who came up through the technical community knew that early-stage companies often had infrastructure situations that wouldn't survive close scrutiny. They funded them anyway, because the alternative was funding nothing, and because the moral consensus of the era held that this was acceptable bootstrapping rather than fraud.

"Nobody thought of it as stolen," said one former venture capitalist, who invested in several early-stage internet companies and agreed to speak generally about the era. "The whole internet was built on university infrastructure. The ARPAnet was government money. The idea that there was some clean separation between legitimate and illegitimate use of that infrastructure — that wasn't how anybody thought about it."

The Long Shadow

This history matters because it shaped a specific cultural inheritance. Silicon Valley's famous tolerance for regulatory gray areas, its instinct to move fast and ask permission later, its tendency to frame rule-breaking as innovation — these aren't personality quirks. They're the residue of an origin story where breaking the rules was genuinely necessary and genuinely worked.

The problem is that the rules being broken in 1996 were mostly network usage policies enforced by university IT departments. The rules being broken in 2016 were labor laws, financial regulations, and consumer protection statutes. The cultural DNA that said "it's fine, we're just using capacity that would otherwise go to waste" scaled into industries where the costs of that philosophy were borne by workers and consumers rather than network administrators.

The T1 line that launched a startup in 1997 was, in most cases, a genuinely low-stakes appropriation. Nobody got hurt. The internet got built. The people who did it weren't villains.

But the mindset they normalized — that infrastructure is a commons to be used by whoever can access it, that rules are suggestions that apply to people without the technical skill to route around them, that the ends of building something cool justify the means of borrowing what you need — that mindset didn't stay in the network closet. It went public.