From Labor Replacement to Cost Collaps
For two years, Silicon Valley aggressively sold a seductive narrative: artificial intelligence would massively replace workers because it was more efficient and, above all, far cheaper. Executives, analysts, and investment funds repeated the mantra until it became dogma. Mass layoffs were justified under the “AI-first company” banner. Today, that promise collapses under irrefutable evidence.
Microsoft has begun restricting internal AI tool usage among its engineers due to skyrocketing bills. This is not a minor adjustment: the costs of inference, tokens, and cloud computing have surpassed, in multiple cases, the salaries of the human employees it was supposed to displace.
NVIDIA’s vice president, Bryan Catanzaro, put it bluntly: “For my team, the cost of compute is far beyond the costs of the employees.” A devastating statement from the company that benefits most from AI infrastructure.
This admission is not isolated. Companies like Uber have also seen AI budgets exhausted in months, revealing that automating human intelligence is not the same as automating repetitive industrial processes. AI generates value in specific tasks, but its energy consumption, GPUs, and dedicated infrastructure maintenance create a permanent economic burden.
AI More Expensive Than Humans: The Physical Problem No One Wanted to See
Stock market euphoria rewarded any story inflating expectations for years. Laying off staff and announcing aggressive AI adoption boosted valuations, regardless of real returns. However, physical reality — energy, chips, latency, and inference costs — has imposed limits that marketing cannot ignore.
Automation does not always reduce costs. In many contexts, maintaining humans remains more predictable and economical than sustaining armies of artificial agents running 24/7 with variable and growing bills. The business model based on scaling tokens and compute clashes with actual operating margins.
This dynamic exposes a central contradiction of the Fourth Industrial Revolution: the concentration of technological power in few hands does not guarantee generalized economic efficiency. On the contrary, it generates investment bubbles that distort the labor market and capital allocation.
Geopolitical and Structural Implications
While big tech promised to democratize productivity, the underlying infrastructure becomes prohibitive for most actors. This accelerates the concentration of capabilities among those controlling compute (NVIDIA, Microsoft, hyperscalers), creating even higher entry barriers.
European governments and regulators must watch carefully. Dependence on models whose operation is more expensive than human labor questions optimistic narratives about accelerated digital transformation. It is not enough to regulate ethical uses; understanding the real economic viability of mass adoption is essential.
The irony is brutal. The same industry that used fear of labor replacement as a corporate discipline tool now faces the limits of its own machinery. They tried to replace humans and ended up trapped in a more expensive, resource-hungry infrastructure.
Conclusion: Confronting Reality
AI is not useless, but neither is it the unlimited efficiency elixir that was sold. Its deployment requires economic rigor, not just hype. Companies ignoring the gap between narrative and physical costs risk not only margins but strategic credibility.
Europe and the UK have an opportunity here: avoid the trap of blind adoption and prioritize AI applications where real returns clearly exceed computational costs. The future of productivity is not built on illusions, but on precise calculation.
Sources
- Fortune, 2026. “Microsoft reports expose AI’s cost problem: The tech is more expensive than paying human employees”. https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/
- Axios, 2026. “AI can cost more than human workers now”. https://www.axios.com/2026/04/26/ai-cost-human-workers
- Entrepreneur, 2026. “Nvidia VP Says AI Costs ‘Far’ More Than Human Employees”. https://www.entrepreneur.com/business-news/nvidia-vp-says-it-costs-more-to-use-ai-than-to-hire-humans
- Publicación en X de @IAbyEsteban, 25 de mayo de 2026.
