As concerns grow that operating costs for artificial intelligence (AI) agents could exceed the average annual salary of U.S. workers, an analysis said barriers to adoption for companies are rising more than expected.

Cointelegraph, a blockchain media outlet, reported on Feb. 20 that tech investor Jason Calacanis (제이슨 칼라카니스) said on a recent podcast he spends about $300 a day to run Anthropic's Claude AI agent. He said the cost is steep even though the agent performs only about 10 to 20 percent of his total work, asking, "When do token fees exceed an employee's salary?" Tokens refer to billing units based on AI model usage.

Chamath Palihapitiya (차마트 팔리하피티야), CEO of Social Capital, raised a similar issue. He argued that for an AI model to replace one employee it must prove at least double the productivity to be economical, and stressed that companies need to set clear budget standards for AI use.

Tech investor Mark Cuban (마크 큐반) said this cost structure is a persuasive counterargument to claims that "AI will soon replace humans." He said that including token costs and maintenance and management expenses, running 8 Claude AI agents to take over one employee's work could cost more than $1,200 a day, and the actual cost could potentially double. He added that companies should consider whether AI will be more than twice as productive as humans, or whether ethical and qualitative risks must also be weighed.

The debate is drawing more attention as companies expand AI adoption while jobs in some occupations decline. A study Microsoft released in July found knowledge-based roles and customer service and sales are most likely to be affected by AI.

Views differ between policymakers and the industry. White House AI and crypto adviser David Sacks (데이비드 색스) previously said, "AI still needs user instructions and verification, so there are limits to creating full business value." By contrast, global consulting firm McKinsey forecast that AI agents could evolve into models that aim for end-to-end automation and operate without human intervention.

The cryptocurrency industry is also focusing on the potential use of AI agents. Jeremy Allaire (제레미 알레어), CEO of stablecoin issuer Circle, predicted that tens of billions of AI agents will use stablecoins for everyday payments within 5 years. Binance co-founder Changpeng Zhao (창펑 자오) also said AI agents could naturally combine with blockchain technology and mentioned the possibility that cryptocurrencies could become AI's base currency.

While there is no disagreement that AI agents have the potential to reshape workplaces in the long term, high operating costs are expected to be the biggest variable limiting corporate adoption in the short term.