After 2 years of being constantly told that AI would take jobs, replace people, and disrupt markets... the narrative is finally shifting.
Companies are now paying significantly more for AI just to handle routine tasks. Developers aren't immune either; Claude Pro plans are hitting weekly limits faster than ever on prompts that used to cost very little. GitHub Copilot has also become considerably more expensive following its change to a new billing model (Usage is now measured in AI credits — 1 credit equals $0.01 USD).
But why did it get so expensive all of a sudden?
AI runs on serious hardware. The way it uses 10 .md instruction files and fixes your bug in a single prompt; that isn't cheap to run. The compute costs behind every interaction are real and significant.
So why were we paying so little for so long?
AI companies needed people to use the product, get familiar with it, and let it become a core part of their workflow. Telling people to pay $100 upfront for unproven software was never going to work. So they did something smarter:
Step A: Raise substantial funding
Step B: Use it to subsidize the cost of AI, making it cheap or free
Step C: Accelerate adoption and build dependency
Step D: Roll pricing back to levels that actually sustain a business
It's a classic land-and-expand play, and it worked. But now we're in Step D.
So does AI become a premium tool only the privileged can fully use?
That question is becoming harder to dismiss. Even major players are feeling the pressure.
Bryan Catanzaro, Vice President of Applied Deep Learning at Nvidia, suggested that AI adoption isn't actually reducing labor costs and, in some cases, may exceed the cost of the employees it was meant to replace.
"For my team, the cost of compute is far beyond the costs of the employees," he told Axios.
Microsoft has reportedly begun canceling licenses for its engineers to use Anthropic's Claude, citing high costs — raising serious questions about the long-term economics of enterprise AI adoption.
And then there's what Sam Altman said during a virtual interview at a Commonwealth Bank of Australia conference:
"I don't think we're going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about."
"I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened. I now think I understand more about why it hasn't, and I'm obviously grateful, but that is an area where my intuitions were just off."
Altman went on to say that the human dimension of work is harder to replace than anticipated. That people genuinely care about interacting with each other, not just getting tasks done.
"We really do care about our interactions with people," he added — which, he said, "updated me to thinking that the jobs picture is likely to be very different than we thought."
The CEO of Linear put it well on X:
What actually happens next
My honest opinion: costs keep rising as companies don't want to fall into massive loss, and spending tokens carelessly becomes a real consideration. Developers will need to be more deliberate and write more of their own code again 😮💨.
Companies will start hiring again. But not for the roles that AI has genuinely absorbed: basic, repetitive, well-defined tasks.
The demand will be for people who can create, build, and maintain processes that produce results efficiently, with or without AI. Architect-level, senior, and even solid mid-level roles. People who think in systems, not just syntax.
The AI-replaces-everyone story was always too simple. What's emerging is more nuanced: AI compresses the value of shallow work and amplifies the value of deep work. If you can do the latter, you're more relevant than ever.
Thanks for reading 👍
Top comments (3)
PS: We're launching the new version of our recruitment platform built to solve a real problem on both sides of the hiring table — helping people find quality global remote roles, and helping recruiters identify candidates who are actually a fit, not fakes/liars.
If you're hiring or job searching, keep an eye on remoteworldwide.net.
This is exactly what happened with cloud computing. AWS subsidized everything, then quietly raised prices once everyone was locked in.
History repeating itself, it seems.
But just like cloud computing, people would still use AI like crazy
Honestly, the jobs apocalypse not happening is the boring outcome everyone rational expected. Hype sells. Nuance doesn't.
You buried the most important point. If compute costs more than employees, then business who plan to replace workers with AI fully need to re-strategize.