Will AI trigger the end of net neutrality?
Since its inception, the internet has worked for one fundamental reason: net neutrality. This is the principle that internet carriers – broadband and mobile providers – must treat traffic equally, regardless of where it comes from. They cannot discriminate.
This idea, noble in its intentions, was what spurred the enormous growth of internet services over the last few decades – without it, things would look very different to what they do today.
Over the past decade, our relationship with the internet has changed dramatically, thanks in no small part to the arrival of streaming services and short-form video. Internet traffic has increased dramatically, and at the same time, has been concentrated among a smaller number of massive firms: Google, Netflix, Meta and the like. Still, these trillion-dollar companies have been accessing the internet on the same terms as you or me.
This is something that has irked the big telco operators. And they have not hidden their frustration.
Marc Allera, the chief executive of BT’s consumer business, wrote in a blog post in 2022: “Just a handful of companies can be responsible for a huge proportion of the traffic they carry.
“If I look across any other industry I can name, the more a resource is used, the bigger a contribution is made. Heavier vehicles are legally obliged to contribute more to road maintenance. Bigger parcels cost more to post.
“But under net neutrality rules, all traffic must be treated equally. That means a business can send any volume of traffic, at any time of day, in almost any way they want, and quote ‘net neutrality’ to deny any obligation to do different.”
Regulator holds the power
Media and telecoms regulator Ofcom launched a review into its net neutrality guidelines in 2022, in which it briefly explored the possibility of allowing telcos to charge particular high-usage businesses.
Ofcom ultimately came down against the idea, stating that “while there are potential benefits to a charging regime, we have not yet seen sufficient evidence that this is needed, and we believed that there is sufficient flexibility …ultimately, whether or not a charging regime should be introduced in the UK would be a decision for government and Parliament.”
A lot, however, has happened in the last few years. AI has gone mainstream and adoption has skyrocketed. The top hyperscalers are set to blow $750bn in capex this year alone, such is the thirst for data and compute of the latest models.
AI could also become much more localised, with developers building their own systems at home using open-source models, sucking in lots of extra data rather than bouncing requests over to centralised systems. All of this will put lots of extra strain on existing systems.
We could reach a point where the levels of traffic relating to AI will be orders of magnitude greater than anything else. In this context, does it really make sense for everyone to be accessing the internet on the same terms – when networks effectively become AI portals with a few other things on the fringes – or should we, to use Allera’s phraseology, start charging more to post the biggest parcels?
It would take a lot to unwind the coveted net neutrality rules, and there are a lot of vested interests in (and legitimate reasons for) protecting it. But the telco execs I speak to are adamant: something has got to give.