NVIDIA’s VP of Deep Learning believes AI computing spend will surpass human payroll costs

Artificial intelligence is developing rapidly, and the hardware and compute costs that companies face are rising significantly. Bryan Catanzaro, NVIDIA’s Vice President of Deep Learning, recently told the media that the compute costs of AI exceed employee salary expenses. In real-world applications, AI may not necessarily reduce labor costs as expected.

AI compute spending exceeds the cost of professional salaries

When NVIDIA’s Vice President of Deep Learning, Bryan Catanzaro, spoke to the media, he admitted that, for the technical team he leads, the computational cost required to run AI models is far higher than the salary expenses of hiring senior technical professionals. This observation challenges the view commonly held in the market that AI must inevitably save companies on operating costs. Although AI is capable of processing large volumes of data, the high-performance hardware, power supply, and maintenance costs behind it are extremely high—making the operating cost of machines in certain areas of technical development instead more burdensome than labor costs, resulting in a new pattern of capital-intensive spending.

Automation technology is not economically feasible for most jobs

Research data released in 2024 by the Massachusetts Institute of Technology (MIT) provides academic support for this cost observation. The study analyzed multiple tasks that rely on visual judgment, finding that AI automation is currently economically beneficial in only about 23% of jobs, while in the remaining 77% of work items, hiring human employees remains a more affordable and efficient choice. The research indicates that if AI is to reach performance levels comparable to humans, the required investment in both software and hardware is very substantial. For most small and mid-sized enterprises or certain industries, the financial threshold for full automation is still too high. In addition, errors caused by immature technology—such as the risk of database corruption reported by engineers—also increase hidden costs.

Tech companies face excessive pressure to reallocate budgets

Despite the staggering initial investment costs, large technology companies have not slowed their spending on AI. According to statistics, global tech leaders are expected to invest about $740 billion in AI-related infrastructure in 2024, representing significant growth compared with the previous year. However, this high-intensity investment also impacts corporate financial planning. Praveen Neppalli Naga, CTO of Uber, said that adopting AI coding tools raises overall R&D costs, forcing companies to re-examine their budget plans because actual spending often far exceeds the originally expected range. This shows that while AI technology improves efficiency, it also creates pressure on companies’ cash flow and resource allocation.

At the same time that companies are investing in AI, large-scale layoffs are also underway. Statistics show that since the beginning of this year, more than 92k tech workers have been laid off and become unemployed, with the pace of layoffs far exceeding prior years’ levels. This reflects an integration period as the tech industry balances financial restructuring and technical transformation. Although current compute costs are extremely high, as infrastructure matures and model operating efficiency improves, there is still room for the cost structure to decline. Whether AI will truly deliver economic benefits in the future depends on whether the technology can achieve more stable, scaled deployments without relying on extensive human supervision—not merely on cost considerations.

This article believes that NVIDIA’s deep learning vice president says that AI compute spending exceeds human labor salary costs, and first appeared on Chain News ABMedia.