The AI industry has recently shown a stark contrast, with extreme situations on both sides. On one hand, the valuations of major AI companies have all exceeded the trillion-dollar mark, and large models are now widely adopted in programming. On the other hand, many companies that had hoped to cut labor costs through AI have been surprised to find that AI is now more expensive than hiring human programmers.

This unexpected situation has put many tech companies in an awkward position. AI programming tools, which were once expected to achieve "cost reduction and efficiency improvement," have become a huge bill for the company's finance department due to their incredible consumption speed.

Even Big Companies Have to Concede to High Budgets

The Chief Technology Officer of the US ride-hailing giant Uber recently openly admitted that the company had completely exhausted its 2026 Claude Code budget as early as April. In order to pay this excessive AI usage cost, the company even had to slow down its annual hiring plan, which sparked deep reflection among management about the blind pursuit of internal AIization.

Unsurprisingly, the financially strong tech giant Microsoft is also facing the same financial burning problem. Recently, Microsoft CEO Nadella issued an order to switch from Claude Code to its own GitHub Copilot for internal development work starting in June. The core purpose was to strictly control the budget and prevent the uncontrolled spending of large models.

24/7 Operation Becomes a "Money-Guzzling Beast"

Although AI programming tools have an advantage over humans in terms of writing speed, in practical application, human programmers who earn several thousand dollars a month are often more cost-effective than AI charged by Token. Especially when many development teams start deploying AI agents to perform continuous 7x24-hour operations, the rate of fund consumption becomes difficult to control, like water released from a dam.

In addition to the high economic cost, the quality issues of AI-generated code are also being criticized by the industry. Several technology experts pointed out that AI often "mass-produces garbage content," with a lot of uncontrollable defects hidden in the generated code. The subsequent review, testing, and deployment still rely on humans to clean up the mess. Replacing human programmers in the short term remains just an empty promise.