Nvidia Executive Says AI Compute Costs Far Exceed Human Worker Pay

@rohanpaul_ai It is so sad that his company does not require intelligence

Fortune article: “For my team, the cost of compute is far beyond the costs of the employees”

- Bryan Catanzaro, vice president of applied deep learning at Nvidia.

An MIT study says AI automation was cheaper in only 23% of vision-heavy jobs, while humans still won on cost in 77%.

But big tech is still spending hugely because companies are buying a future cost curve, not today’s savings, with $740B in AI capital spending already tied to a 69% jump from 2025.

---

fortune. com/article/why-is-the-cost-of-ai-higher-than-human-workers-nvidia-executive/

@rohanpaul_ai It’s a massive upfront investment, but building the infrastructure now is what sets up the entire industry for the next decade.

@rohanpaul_ai If we read through the article written by Satya Nadella

This comparison is baseless as the real worth of AI can only be derived off by complementing Human Intelligence & AI

@rohanpaul_ai AI现在最大的问题是：说得比做的好听。真正能落地的项目，远比融资PPT里的少。 「Fortune article: “For my team, the cost of compute is far be...」

@rohanpaul_ai nvidia admitting shit is expensive as hell thats the real story

@rohanpaul_ai They should all be fired for incompetence or show the tokens spent on what .

With a OpenAi Pro subscription you replace every programmer alive . So how can programmers using it mess up so badly ?

I think management is burning tokens making funny Images cause Ai supervises now

@rohanpaul_ai The biggest bitchers about the cost/benefit of computer are the losers and soon to be chapter 11 companies

@rohanpaul_ai Interesting, wonder if this considers total compensation+benifits or just hr cost? Also humans get sick, show up late, get distracted, etc how do you account for this loss (i.e. increased cost)

@rohanpaul_ai Bryan's right. A single frontier model training run burns millions in GPU-hours. Inference cost scales with volume. The key difference: compute is a variable cost that grows with usage, headcount isn't. It's structural, not a spending mistake.

@rohanpaul_ai had claude code a volume filter last week that paid for itself in one session on perps. betting on that cost curve already.

@rohanpaul_ai The missing KPI is cost per accepted task. Compute alone is misleading: retries, review time, failed outputs, and cleanup can make a “cheap” agent more expensive than a human doing the narrow job once.

@rohanpaul_ai Came across this on the app and had to post a screenshot. Our AI systems rank this post as 3rd best on the internet right now. 🎉

@rohanpaul_ai $740B is not a bet on today’s ROI. It is a land grab before the cost curve makes everyone wish they moved earlier.

@rohanpaul_ai First we are told that the employers love AI and their employees hate it. Then we are told it's the other way around. The employers love it as long as they think it's making them money.

@rohanpaul_ai ai's real value isn't replacing humans but scaling what we can dream the future isn't cheaper labor it's unprecedented ambition

@rohanpaul_ai I understand the long-term vision of abundant intelligence.

But today, for many tasks, abundant human labor is still cheaper than abundant AI.

Until the economics change, adoption will be driven by value creation, not labor replacement.

@rohanpaul_ai Useful reminder. AI still needs financial discipline: compute cost, utilisation, workflow fit, and clear payback. Market pressure alone is not a business case.

@rohanpaul_ai I keep reading humans are cheaper. We’re still burning $740B on AI. They’re not buying efficiency. They’re buying our replacement. Ask yourself where your career lands in five years.