There's no such thing as free internet – we just don't understand how we pay for it 3/2.

This article is the second part of a series on the monetisation of network services. In the first part we looked at whether internet access is a fundamental human right. In this article we examine the evolution and infrastructure costs of online data systems. The third and final part will cover the rise of paid content and the paywall, and the business model of subscription services.
„The only free internet is the one where we don't understand how we're paying for it."
– wrote Stan Hanks, CFO of Columbia Ventures Corporation. Hanks experienced several of the internet's generational shifts first-hand. In his Quora post on the subject, also published by Forbes, he explained the infrastructure and operating costs that went with the development of each generation of the network. To interpret this correctly, we need to know the phases in which today's internet came about.
From the defence industry to a thirty-thousand monthly fee
In the 1960s, the United States used supercomputers whose performance seems laughable today but which were extremely expensive at the time. To cut development costs, the US government proposed not building new machines but sharing computations and data between them.
The project was entrusted to the US Department of Defense's division responsible for military and technological development, DARPA. By 1969, the ARPANET was born, and with it the concept of the packet-switched network, which made it possible to send small data packets over phone lines, from which the receiving station then rebuilt the original file.
ARPANET was initially used solely to send and receive such files; email and mailing lists only arrived later, in the 1980s. But even then it was clear what a gap network-based electronic communication filled, and from this one could infer how widely society would adopt it.

The costs of building and using the network were initially spread among few participants. Building the military network that operated as part of ARPANET, MILNET, involved staggering hardware costs. The routers and modems of the day, cutting-edge at the time, were very expensive: in 1985, a modem able to transmit 2400 modulated signals per second, i.e. 2400 baud, cost 500 dollars. Although MILNET's phone line infrastructure was built not by the network's operator, the Defense Communication Agency (Defense Communication Agency, DCA), but by the landline giant AT&T, using the leased lines still cost a fortune.
The defence ministry's military spending was funded by taxpayers, so it could afford to bill generously. But DARPA, considering expanding the network, knew few could afford the technology at full price, so it ensured member institutions joining the expansion initiative could access ARPANET at a reasonable price.
Member institutions paid for access individually, but to use the network they also needed a DARPA sponsor and support from a DARPA agency. Installation cost an average of 20,000 dollars, which the agency covered, but customers' extra requests – such as a new IMP slot or a modem – were charged extra. On top of that came annual running costs of about 30,000 dollars. (These are 1980s prices; 30,000 dollars then would be worth 95,000 today, i.e. over 28,000,000 forints.)
The first ARPANET derivative, the Computer Science Network (CSNET), launched in 1981, used the Tymnet X.25 protocol to carry IP packets. The high-bandwidth serial card installed on the host and the 9600-baud modem cost 20,000 dollars. Monthly costs were billed by data traffic; the fee for continuous, unlimited transmission sometimes exceeded 10,000 dollars.
The next generational leap, the launch of NSFNET, was initiated by the US National Science Foundation (NSF). Usage rights to NSFNET, which drew on DARPANET's earlier model, and the communications infrastructure were provided not by DARPA but by the NSF, using AT&T's LongLine network infrastructure. NSFNET was the final step before the internet became a commercial service: from 1995, after NSFNET was shut down, the internet became a service you could subscribe to, like a landline.
But – as Hanks pointed out – the internet wasn't free then or later. Beyond the 30,000-dollar entry fee, you needed DARPA's support, a team of experts working (at least) two shifts, and network build-out – not to mention the project's administrative costs. Had DARPA not borne most of the costs in the early days of networking, it would have cost member institutions another 50,000 dollars.
In Hungary, on this side of the Iron Curtain, we saw none of this. The internet was a forbidden COCOM-listed technology; even academic researchers only knew of the network by hearsay. But when the internet became widely available in Hungary from 1995, crowds gasped at the subscription fee of 30,000 forints a month charged for a modem connection and a few kilobytes per second. (For contrast: according to the Hungarian Central Statistical Office (KSH), the Hungarian minimum wage at the time was 12,200 forints a month.)
If it can be cheaper, can it be free?
Today, when multi-gigabyte mobile data packages are thrown at us for the price of a burger meal and the number of internet subscriptions exceeds the population, these prices seem incredible. But anyone who has half-followed hardware prices and subscription fees over the past decades can guess that the per-capita unit cost of computing power and internet access has fallen to a fraction of what it was in the nineties.
A freeconomics – this principle has accompanied the development of the web. Today we can use unlimited storage or infinite computing power – sometimes even for free. It seems that Moore's law – which says the unit cost of computing power halves every eighteen months – applies not only to hardware manufacturing but to network services too.
Based on price and performance trends, it seems logical to conclude, that if hardware performance and bandwidth approach infinity while the price converges on zero, sooner or later we'll get unlimited-speed internet and web hosting for free. The downward price trend is expected to continue, and growth can be guaranteed too. Billions worldwide live without internet access today, but as more and more people connect, internet penetration will in time reach a level where per-capita access costs become affordable for people in developing countries too.

But even the first step, building the infrastructure, costs hundreds of millions if not billions of dollars, and network expansion, increasing bandwidth and price cuts are hampered by other factors too.
- It's hard to save on network-building costs. Online infrastructure costs have been spread more widely since internet users grew from a few dozen to several billion. But rising hardware performance and transfer speeds don't help if you can't save on digging trenches for cables, the wages of on-call developers and sysadmins, or the precious metals, plastic and silicon used in hardware manufacturing. That's why the cost of the internet won't – can't – be zero.
- There are no rival internet providers preserving backward conditions. Linked to the previous point: government bodies or private companies that own and operate network infrastructure don't lease the bandwidth needed to internet providers. This not only pushes up the price of existing subscriptions but also discourages service development.
- Where there are cartels, there's no price competition. This is what the United States, the world's leading superpower, struggles with: a quarter of its population has no access to internet at normal speeds or affordable rates. According to a 2015 study by the Center for Public Integrity, telecoms companies probably like the status quo; it looks as though service area boundaries were drawn so they don't have to compete with each other and can keep rates artificially high. Whether that's intentional isn't known, but in any case in the United States an internet subscription costs on average 3.5 times what you'd pay a French provider for the same bandwidth and data traffic.
Funke Opeke, CEO of the Nigerian mobile company MainOne, a few years ago told Wired, about the difficulties she faced working with Nitel, Nigeria's biggest mobile operator. Although almost everyone in the country had a phone, few had internet access, because providers couldn't reach thebackhaul) networks – the parts linking a broadband network's access and core networks.
To get around the problem, in 2008 Opeke founded MainOne and recruited investors to build an independent broadband network infrastructure. The result was the 7,000-kilometre undersea cable from Portugal to Nigeria, which boosted internet speeds in the region twentyfold. And it wasn't the only such initiative. Thanks to the undersea cable of another African provider, SEACOM, the monthly fee for a megabit connection fell in some countries from several thousand dollars to 50–100 dollars.
Bandwidth went up, prices came down: expanding infrastructure delivered as expected, but at a price: MainOne's cable network cost 240 million dollars and SEACOM's 650 million – not counting the cost of building and installing the backbone network.
If your cable's too short, extend it with a satellite
„Then let the internet come from the air! Beam it from solar drones, like Facebook's Aquila; from balloons, like Google in Project Loon; or from satellites, like OneWeb and Starlink do!" – anyone who thinks laying fibre-optic cable is outdated and expensive might interject. A fair point – but we can't yet know how cost-effective satellite data transmission will be in the long run.

Satellites in low Earth orbit, 160–2,000 km up, offer higher bandwidth and lower latency than other satellite services, but because they fly so low they can't cover as much area as geostationary satellites orbiting 35,000 km up. A stable wireless service therefore needs hundreds of satellites. Although such a fleet can offer more coverage than fixed cables, it costs roughly an order of magnitude more than cabling a continent.
OneWeb founder Greg Wyler told Businessweek he launched the venture with two billion in starting capital. Elon Musk, CEO of SpaceX, estimated the expected total cost of the Starlink project, which aims to connect the world with satellite internet, at ten billion dollars. They didn't say whether this includes installing ground stations and gateways or developing and maintaining the software running the system – but the service needs those too.
Satellite technology can be a promising alternative where building a wired backbone is impossible or difficult. But adoption is hampered because satellite data transmission isn't as mature and reliable as traditional wired networks, and there's no benchmark business model that could boost wider acceptance of the technology.
Ever cheaper, but ever worse
Users don't care whether they connect via a backbone or a satellite. What matters isn't the means but the goal: access to the information accumulated over the web's thirty-year history. Today far more people can do this, far more cheaply and quickly than ever before. But as the cost of internet access fell, the value of publicly available content went into a nosedive too.
On the internet we're wading knee-deep in rubbish. Digital noise keeps growing, opinion bubbles multiply, debates get shallower and angrier, and more and more people write ever bigger nonsense. Yet quality hasn't disappeared – it has just changed shape. Exclusive information, valuable content and reliable news increasingly end up behind payment barriers – paywalls. Because not only network infrastructure but also producing quality content costs money, and some think that money is best collected from the most appropriate party: the user.












