Every carrier now offers an unlimited data plan and none of them mean it. That is not an accusation of dishonesty exactly. It is an observation that the word was chosen at a time when it could not be tested.
Back in May I argued that the limit on the mobile web was the tariff rather than the network. That is still true for most subscribers. What has changed since is that a small number of people now have handsets with real browsers and flat-rate plans, and their usage is not slightly higher than the average. It is a different order of magnitude, and it is arriving on infrastructure that was engineered for telephone calls.
A voice network is a beautifully understood thing. Each conversation takes a predictable slice of capacity for a predictable duration, and a century of statistics tells you how many simultaneous calls to plan for. You can size a cell site with confidence.
Data does not behave like that at all. One person on a train watching video consumes what a hundred callers would, in bursts, with no upper bound and no established distribution to plan against. And unlike a call, there is no natural end. A phone with push email is transacting all day whether or not anyone is looking at it.
Unlimited was safe to promise when the browsers were bad, because a bad browser is self-limiting. Nobody exceeded a threshold that nobody was near.
The management has already started, quietly. Look closely at a contract and you will find language about acceptable use, about protecting network integrity, about excessive consumption at the operator’s discretion. Tethering is prohibited on most plans, which is revealing, because the bytes are identical whether a laptop or a handset requested them. What is really being restricted is the class of user likely to generate a lot of them.
None of this appears in the advertising, and I doubt it is meant to deceive so much as it reflects an organisation that has not yet decided internally which promise it is making.
My expectation is that unlimited disappears within a few years and returns as tiers, and that the framing will be fairness. Why should a light user subsidise a heavy one. It is a good argument and it happens to be true, which will make it easy to deploy and hard to object to.
The awkward part is that the industry will be withdrawing a promise at precisely the moment customers start relying on it, and the people worst affected will be the ones who did what the marketing asked of them.