Nvidia exec says AI compute costs more than human workers, as Big Tech pours $740B into AI capex with no clear productivity data

Despite widespread tech layoffs that appear to signal a shift from human to AI labor, an Nvidia vice president of applied deep learning told Axios that for his team “the cost of compute is far beyond the costs of the employees,” a sentiment backed by a 2024 MIT study finding AI automation is only economically viable in 23% of roles where vision is a primary part of the work. Big Tech firms have nonetheless announced $740B in capital expenditures on AI in 2026, a 69% increase from 2025, even as AI software fees have risen 20-37% over the past year and there is no widespread data supporting the idea that AI is displacing jobs or improving productivity at scale. Analysts say the cost of AI inference for large language models is expected to drop more than 90% over the next four years, and that AI's economic viability will ultimately depend on the technology proving itself reliable enough to reduce hallucinations and human oversight requirements while also shifting from flat subscription pricing to usage-based models.