Nvidia exec says AI is more expensive than human workers as compute costs explode across tech

AI Cost Reality Challenges Automation Economics

The economics of artificial intelligence are proving more complex than anticipated, with compute costs now surpassing labor expenses at some of the world's leading tech companies. Bryan Catanzaro, Nvidia's vice president of applied deep learning, recently told Axios that "for my team, the cost of compute is far beyond the costs of the employees" 1

1

. This revelation comes at a time when companies are simultaneously laying off thousands of human workers while pouring billions into AI infrastructure, raising questions about the true viability of AI automation.

The shift is particularly visible at Nvidia, where the company selling the tools for the AI revolution is itself experiencing the cost pressures firsthand 2

2

. Nvidia CEO Jensen Huang has suggested that a $500k salaried engineer should spend at least $250k worth of tokens per year, positioning AI spending as a measure of productivity 1

1

. This perspective reflects a broader belief among tech leaders that current AI operating costs represent an investment in future automation rather than a red flag.

Token-Based Pricing Creates Budget Shocks

The financial impact of token-based pricing has caught many organizations off guard. Uber's CTO Praveen Neppalli Naga revealed he has "gone back to the drawing board because the budget I thought I would need is blown away already" 1

1

. The ride-hailing giant has already burned through its entire 2026 AI budget, largely due to heavy use of coding models like Anthropic's Claude Code 2

2

.

The continuous nature of AI agents working on coding tasks, automation workflows, and other scheduled operations requires a constant flow of spending that many finance teams failed to anticipate. Amos Bar-Joseph, CEO of Swan AI, posted on LinkedIn about a $113k bill from Anthropic for a four-person team, which translates to roughly $28k per person per month—likely exceeding individual monthly salaries 1

1

. These figures illustrate how AI spending can quickly surpass labor expenses when large language models are deployed at scale.

Corporate America Faces Rising AI Operating Costs

The trend extends beyond AI-native companies. Big Tech firms have announced $740 billion in capital expenditures this year, a 69% increase from 2025, according to Morgan Stanley 5

5

. AI expenditures may reach $5.2 trillion by 2028, with $1.6 trillion from data center spending and $3.3 trillion from IT equipment, McKinsey data shows . Meanwhile, fees for AI software have increased by 20% to 37% over the past year, according to spending management firm Tropic 5

5

.

Token costs have become particularly contentious, with some engineers engaging in what's been dubbed "tokenmaxxing"—using millions of tokens daily and racking up monthly bills exceeding $150,000 4

4

. Software engineer Max Linder told The New York Times he "probably spends more than my salary on Claude" 4

4

. These usage patterns have turned AI spending into a metered, recurring operating cost that rises with every request, making it harder to forecast than standard software licenses 2

2

.

Human Workers Remain More Cost-Effective

Despite massive AI spending, research indicates that AI is more expensive than human workers in most scenarios. A 2024 MIT study found that in 77% of cases where vision is a primary part of work, it was cheaper for humans to continue their jobs rather than implement AI automation 1

1

. Keith Lee, an AI and finance professor at the Swiss Institute of Artificial Intelligence's Gordon School of Business, describes the situation as "a short-term mismatch" driven by hardware and energy costs raising operating expenses for AI providers 5

5

.

The cost structure is further complicated by AI companies' flat subscription models, which often fail to cover operating costs for heavy users, leading to recent price increases from providers like Anthropic 5

5

. Brad Owens, vice president of digital labor strategy at Asymbl, notes that "the tone is shifting a bit more into what is the true value of a worker... human or digital" 2

2

.

AI-Induced Job Displacement Continues Despite Higher Costs

Paradoxically, layoffs continue even as AI proves more expensive than the human workers being replaced. There have been more than 92,000 layoffs in tech in 2026 so far across nearly 100 companies, far outpacing last year's 120,000 total, according to Layoffs.fyi 5

5

. Meta announced plans to lay off 8,000 employees while scrapping 6,000 open positions, citing the need to "run the company more efficiently" 5

5

.

Some executives view high AI spending positively, seeing it as evidence that employees are driving innovation and working toward large-scale automation. Uber's Naga reported that 11% of the company's live code updates are now written by AI agents, with a vision to "transform from software engineering to agent software engineering" 1

1

. Anthropic's head of Claude Code claimed that "pretty much 100 percent" of the company's code is now AI-generated, while Google and Microsoft report around a quarter of their code comes from AI 4

4

.

Future Outlook Depends on Cost Reductions and Productivity Gains

Whether current AI spending represents a temporary expense alongside salaries or a failed investment remains uncertain. Gartner predicts that inference costs for large language models with 1 trillion parameters will plummet by more than 90% over the next four years 5

5

. If these cost reductions materialize alongside improvements in reliability and reduced need for human oversight, the economics could shift dramatically in favor of AI automation.

However, recent studies show that the vast majority of companies rushing to implement AI without a strong plan experience massive losses on those initiatives 1

1

. Federal Reserve data indicates that only about 18% of companies had adopted AI tools as of late 2025, though this represents 68% growth in the adoption rate since September 2025 5

5

. As Lee notes, "It's not just about AI becoming cheaper than humans. It's about becoming both cheaper and more predictable at scale" 5

5

.