The AI paradox: computing now costs more than paying an employee
A teacher asks AI to generate an exam. Students solve it using the same AI. The teacher grades it with AI. The cycle closes with no human involvement. It's not a joke: it's happening in classrooms and corporate training courses, where generating, solving and grading exercises is handled by generative models. AI has arrived to free us from «stupid busywork», as one analyst puts it, but the time savings hide an economic contradiction that is starting to shake foundations.
Compute costs more than the employee
A Nvidia vice president openly acknowledged that, for his team, the cost of computing is significantly higher than the cost of employees. The statement, captured in a video (1:05, 9:25-9:50), challenges the dominant narrative: AI doesn't replace workers to cut costs, because the cost of GPUs and energy can exceed people's salaries. The pushback came quickly: if AI is doing more work than before, or if compute sold to customers is counted, the math changes. Either way, the myth of cheap mass replacement is wobbling.
Democratization (and the antiestéticar of the subscription trap)
While the big models are offered for free or at reduced prices, skepticism is growing. Some analysts warn: first they hook you with free access, then they shut off the tap and charge an unaffordable fee. The pace of releases — Claude, Gemini, new models every two weeks — creates saturation. And some wonder whether intellectual dependence isn't making us dumber: students who can no longer imagine studying without AI.
Real-world use: concrete savings, inflated expectations
Amid the noise, ordinary users find everyday utility: «I used to search Google, now I ask the AI». The average worker says it has taken away «a pile of work» for very little money. But the real business isn't in individual subscriptions — it's in selling computing capacity to companies. There lies the paradox: AI is cheap for the end user, but extremely expensive to operate at scale. And the bubble of AI-generated YouTube channels is already deflating: audiences detect the low quality.
AI does spare us absurd chores, yes. But the real cost of its infrastructure — higher than the salary of many employees — suggests the savings aren't where we were told. The unsettling figure: while companies pay more for computing than for people, the average worker is rubbing their hands over a tool that costs them four euros a month. Something doesn't add up in the final tally.
Related discussions in the forum