AI Agents May Face a Cost Barrier Before Displacing Human Workers
Crypto & Tech Desk
February 20, 2026
Rising Expenses Prompt Caution Among Tech Investors
A handful of high‑profile technology investors have sounded a warning that the economics of today’s AI agents could keep them from eclipsing human labor in the near term. The concern centers on the price of “tokens” – the usage credits required to run large‑language‑model (LLM) services such as Anthropic’s Claude – and the productivity gap that must be closed for the models to be cost‑competitive.
Jason Calacanis, the founder of several media ventures and a regular guest on the All‑In podcast, disclosed that he pays roughly $300 per day for a Claude‑based agent that he uses to help manage his portfolio of businesses. Even at that price the agent is only delivering about a tenth to a fifth of its full capability, translating into a yearly outlay that approaches $100 k per bot.
Chamath Palihapitiya, chief executive of Social Capital, echoed the sentiment. He noted that a deployment must perform at least twice as well as a human employee for the token cost to make sense. Palihapitiya suggested that firms may soon need to cap how much AI they consume to stay within budget constraints.
Mark Cuban, the serial entrepreneur and investor, amplified the argument on the platform X. He calculated that eight Claude agents would cost roughly $1,200 per day, which is double the expense of a single human worker performing the same tasks. Cuban asked whether the agents could demonstrably exceed human productivity by a comparable margin, and stressed that intangible factors – morale, ethical considerations, and the quality of decision‑making – are difficult to capture in a pure cost‑benefit analysis.
The Broader AI‑Job Landscape
The discussion arrives against a backdrop of ongoing debate about AI’s impact on employment. A July paper from Microsoft highlighted that knowledge‑intensive occupations, as well as roles in customer service and sales, appear most vulnerable to automation by advanced AI. Yet some policymakers, such as former White House AI adviser David Sacks, argue that the hype is overstated because current models still require extensive prompting, verification, and human oversight to generate real business value.
Consulting firm McKinsey & Company paints a more optimistic picture for the technology’s future, emphasizing that the next wave of AI agents could execute entire workflows without continuous human supervision, thereby unlocking efficiencies that go beyond simple task replacement.
Stablecoins as the “Fuel” for Agentic Economies
Parallel to the cost debate, the cryptocurrency community is exploring how digital assets could power AI agents at scale. Jeremy Allaire, chief executive of Circle, predicts that billions of AI agents will adopt stablecoins for everyday transactions within the next half‑decade. His forecast builds on the rapid growth of stablecoin transaction volume, which analysts expect to exceed $50 trillion annually by 2030.
Changpeng Zhao, co‑founder of Binance, has previously suggested that blockchain may become the natural interface for AI agents, given its ability to provide transparent, programmable money flows. The idea is already being tested: on Ethereum’s Layer‑2 network Base, the AIXBT bot (operating through the Virtuals Protocol) executes micropayments and trade actions on behalf of users. Meanwhile, the ASI Alliance on Fetch.ai orchestrates asset management and coordination tasks for participants in its ecosystem.
OpenAI recently unveiled a benchmark suite designed to gauge how well AI models can discover, patch, or even exploit vulnerabilities in smart contracts. The initiative underscores a growing recognition that AI agents will play a pivotal role both as defenders and attackers in the financial‑technology arena, where billions of dollars of assets are secured by code.
Analysis
-
Cost‑Productivity Parity Is the Current Bottleneck
The token pricing model of leading LLM providers makes it essential for enterprises to demonstrate a clear productivity uplift – generally a factor of two or more – before AI agents become financially viable replacements for human staff. -
Human Factors Remain Unquantifiable
Even if an AI agent could match output, qualitative concerns such as employee morale, ethical decision‑making, and corporate culture are difficult to evaluate purely in monetary terms. These considerations may slow adoption, especially in sectors where trust and brand reputation are paramount. -
Crypto Infrastructure Could Reduce Friction
By leveraging stablecoins and blockchain‑based payment rails, firms might lower transaction costs and improve transparency for AI‑driven services. However, this shift also raises regulatory and security questions that will need to be addressed before widespread adoption. - Regulatory Scrutiny Likely to Intensify
As AI agents become more embedded in financial operations, regulators may focus on the intersection of AI risk management and crypto compliance, particularly around anti‑money‑laundering (AML) and consumer protection standards.
Key Takeaways
- High token costs are currently a disincentive for replacing human workers with AI agents, unless the agents can deliver at least double the productivity of a comparable employee.
- Prominent investors such as Calacanis, Palihapitiya, and Cuban are publicly questioning the economic rationale behind large‑scale AI deployment in business operations.
- Academic and consulting research suggests that while certain job categories are vulnerable, the transition to fully autonomous AI agents remains contingent on technical and economic breakthroughs.
- Stablecoins and blockchain technologies are emerging as potential “native currencies” for AI agents, promising streamlined payments and new business models, but they also introduce additional regulatory considerations.
- The interplay between AI and crypto will likely shape the next wave of enterprise automation, with implications for cost structures, security, and the future of work.
Cointelegraph remains committed to independent, transparent journalism. Readers are encouraged to verify information independently. For more details, see our Editorial Policy.
Source: https://cointelegraph.com/news/ai-agents-wont-steal-jobs-if-theyre-too-expensive-to-run?utm_source=rss_feed&utm_medium=feed&utm_campaign=rss_partner_inbound
