Claiming Efficiency, AI Costs Turn Out to Be More Expensive Than Employee Salaries

Jakarta, CNBC Indonesia - Many companies are starting to replace human employees with “machine employees” known as AI agents to cut costs. However, company bosses now have to bite their nails because the expenses incurred for AI agents are actually higher than paying employees. Based on a report from Axios cited by Futurism, several company leaders now admit to spending more money on “computing” than on employees. For example, Bryan Catanzaro, Vice President of Applied Deep Learning at Nvidia, said: “For my team, computing costs are far higher than employee salaries.” This ballooning cost problem is becoming more prominent as more companies rely on AI agents and AI devices. “AI now generates nearly 100 percent of the code at Anthropic,” said Boris Cherny from Claude. Similar stories are told by bosses at Google and Microsoft. Around 25 percent of the code at both companies is created by AI. Meta, the parent company of Facebook, Instagram, and WhatsApp, also evaluates employees based on how much they use AI. As a result, tech workers are now competing to “spend” AI tokens, known as the activity of “tokenmaxxing.” Tokens are the “quota” given to users to access AI services. According to Futurism, some people spend more than US$150,000 (Rp 2.5 billion) per month on AI token usage. “I seem to be spending more money on Claude tokens than my salary,” said Max Linder, a programmer in Stockholm. The Information reported that Uber’s entire 2026 AI budget has been exhausted on Claude token usage. For AI providers like Anthropic and OpenAI, this trend is highly profitable. OpenAI investors state that the company’s rising costs give it an edge because Codex is more efficient than Claude Code. On the other hand, Anthropic is capitalising on the situation by raising prices.