What If the Internet Had Stayed a Closed Military Project?

ARPANET, the internet's direct ancestor, was funded by the US Department of Defense's Advanced Research Projects Agency (ARPA) and went live in 1969, initially connecting four university research computers in California and Utah. Its purpose was explicitly research- and defense-oriented: a resilient, distributed communications network partly motivated by Cold War concerns about the vulnerability of centralized systems. Through the 1970s, it remained restricted mainly to universities, defense contractors, and government research labs, governed by an acceptable-use policy that explicitly barred commercial traffic.

The transition from closed research network to open civilian infrastructure happened through a specific sequence of decisions rather than any single moment. The National Science Foundation's NSFNET, built in the mid-1980s to link university computing centers, adopted TCP/IP — the protocol suite ARPANET had standardized on in 1983 — and gradually expanded access beyond the original defense-research community. Crucially, NSF's 1991 decision to lift its own commercial-use restrictions, followed by the network's privatization through the mid-1990s, opened the infrastructure to commercial internet service providers for the first time. Tim Berners-Lee's invention of the World Wide Web at CERN in 1989-91, released freely rather than licensed, gave the newly open network something ordinary people could actually use. By the mid-1990s, the shift from defense and research tool to mass consumer medium was essentially complete.

The most plausible version of this divergence isn't the internet never being invented — the underlying packet-switching research was valuable enough to defense and academic interests that some version of it was always going to exist — but a version where the network stays walled off from commercial and public use for considerably longer, perhaps another fifteen to twenty years past 1991. This could plausibly happen if NSF and the Pentagon had continued treating the network as sensitive infrastructure worth restricting, similar to how encryption technology itself was treated as a munition under US export law well into the 1990s, or if early commercial online services like CompuServe and AOL had successfully lobbied to keep the government-funded backbone closed to protect their own proprietary, closed dial-up networks from competition.

A narrower, equally plausible version imagines Tim Berners-Lee's employer, CERN, deciding to patent and license the Web's core technologies rather than releasing them royalty-free in 1993 — a decision Berners-Lee has said in interviews he considers one of the single most consequential of the entire internet's history, precisely because a licensing fee at that early stage would likely have pushed developers toward a competing, unlicensed alternative or simply slowed adoption enough to change everything that followed.

In the years immediately following a continued military and research restriction, the most direct effect falls on the early-1990s commercial online landscape: proprietary, closed dial-up services — America Online, CompuServe, Prodigy — continue operating as separate, non-interoperable walled gardens for considerably longer, since there's no open, common network for them to eventually connect to and be absorbed by. Each service remains its own self-contained internet-in-miniature, with its own email system that can't reach users on a competing service, its own content, and its own forums.

Academic and defense use of the ARPANET/NSFNET lineage continues largely unaffected in this scenario, since that was the network's original and sustained purpose — researchers, government agencies, and defense contractors keep exchanging data much as they did in reality, just without the eventual opening that let a home computer with a modem join the same network.

For ordinary households and small businesses through the 1990s, the practical experience of getting 'online' looks meaningfully different: rather than a single open internet reachable through any of several competing service providers, going online means subscribing to one specific proprietary service and largely staying within it, unable to freely reach content or people on a different service the way an open internet allows. Email addresses are tied to a specific provider in a much stickier, more consequential way than they are today, and switching providers means potentially losing contact with everyone who only had your old address.

Small businesses hoping to reach customers online face a similarly fragmented landscape — building a 'web page,' in the sense that became ordinary by the late 1990s, either doesn't exist as an option or requires choosing which proprietary network's ecosystem to build a presence within, closer to how a business today might have to choose which single social media platform to invest its limited marketing effort in, except for the entirety of what 'having an online presence' would mean.

At a global scale, the most significant consequence is economic and geopolitical: much of the last three decades of technology-sector growth, from the dot-com boom through the rise of search engines, social media, e-commerce, and cloud computing, depended on a single, open, interoperable global network that any developer or company could build on without permission or licensing fees. A world where that openness arrives fifteen to twenty years later plausibly delays the entire subsequent wave of internet-driven economic growth by a comparable span, while very plausibly changing who ends up dominating it — the specific American technology companies that built early advantages during the 1990s and 2000s land in a different competitive environment entirely if the open network arrives in, say, 2010 instead of the mid-1990s.

The delay would also very plausibly reshape which countries and companies end up setting the internet's technical and governance standards. Much of the internet's core technical governance — the IETF's protocol standards, ICANN's domain name system — was established during the 1990s by a relatively small, US-centric community operating in the network's early open era; a later opening, arriving into a more multipolar, economically distributed technology landscape, could plausibly have produced a considerably less US-centric set of founding institutions.