The Subscription Gap
As AI subscriptions become more expensive, the richer may get better before getting richer.
TL;DR. AI is already levelling parts of academic work, especially for non-native English speakers and smaller teams. But a recent conference dinner made the economics hard to ignore: if the best tools become increasingly expensive, the same technology may also help existing advantages compound.
At a recent conference dinner, six of us sat down after a long day of talks. All established academics, no youngsters. Two colleagues had presented that afternoon. Their slides were sharp, and visually stronger than the average academic deck.
Over dinner, they told us how they had made them. One had used Claude, the other ChatGPT, for essentially the whole deck.
This was not said with embarrassment, but with genuine enthusiasm. A feeling I can relate to: the childish joy of playing with these new tools. They had found something that changed how they worked, and they wanted others to try it too.
Of course, it immediately turned out that everyone was using AI for much of their daily work: writing, coding, summarising, teaching preparation, slides, emails, administration. We were all paying for premium subscriptions. But I was the only one currently not paying for a $200 per month plan.
I have not stopped thinking about that dinner.
None of my dinner companions was a heavy coder. Not at all. Yet the difference between occasional AI use and sustained access to the most powerful systems is not marginal. Beyond a cooler model, you get better coding support, longer context windows, fewer usage limits, and the ability to work through a slide deck, grant proposal, lecture plan, or messy administrative document without constantly hitting the limits of the tool.
But this is not simply about productivity. AI is changing what colleagues may begin to expect from one another. Once stronger slides, cleaner code, sharper grant narratives, and more polished drafts become easier to produce, they will become normal. The baseline moves.
This moving baseline is where the subscription gap begins.
It is not only the difference between people who pay and people who do not. It is also about how much one pays, and what that buys. In academia, as elsewhere, small advantages accumulate. A better first draft here, a faster revision there, a clearer deck, a stronger proposal, a piece of code debugged in half the time. None of these moments is decisive on its own. Together, repeated often enough, they change the pace of work.
The richer may get better before getting richer.
As I have argued in different contexts, AI can reduce friction. It is a real levelling force for non-native English speakers. It can improve teaching materials, make coding more accessible, and support researchers without large teams. There is something genuinely democratising in that.
But as it lowers some barriers, it may raise others, especially if prices keep going up. The risk is not that academics stop thinking. The risk is that the academic profession changes unevenly and informally. Some people acquire a new layer of assistance. Expectations rise. Others compete against capacities they do not have, cannot afford, or have not had time to master.
The hope is that academic subscriptions will start to exist. Universities and funders may eventually treat advanced AI access as part of the infrastructure of knowledge production, rather than as a private consumer choice. But for now this is not really a governance hope. It is mostly a hope that big tech companies decide to be generous, or strategic, or both. As such, it is a small hope.
That is the change. The AI advantage is becoming ordinary. And ordinary advantages, in academia, rarely remain evenly distributed.