Recently, a study used four scenario models to quantify AI's impact on employment in the US— the results are a bit sobering.

Let's start with the baseline scenario. The peak unemployment rate rises by only 0.6 percentage points, roughly 1 million people unemployed, with a transition period of about 2 years. The key is that afterward, the 15% productivity boost brought by AI will create new job opportunities. It sounds somewhat acceptable.

But looking upward, what if companies accelerate deployment or AI's substitution capability exceeds expectations? The peak unemployment rate could rise to 1.2 percentage points, 2 million people. This is already starting to feel tense.

The worst-case scenario—rapid adoption combined with high substitution rate—is catastrophic. Unemployment could soar by 2.4 percentage points, 4 million people unemployed. The current unemployment rate is 4.5%, so this would jump to 6.9%, definitely approaching recession levels.

The signals are already flashing red. By 2025, the unemployment rate among young workers in the tech industry could surge by 3 percentage points. Jobs easily replaced by AI, such as administrative office roles and graphic design, are seeing hiring growth fall below historical trends. Silicon Valley's venture capitalists are saying 2026 will be a turning point—AI will shift from merely improving efficiency to directly replacing workers. Companies' budgets will also shift from hiring to automation.

However, one honest point: current analyses do not consider the scenario of Artificial General Intelligence (AGI). If AGI is truly achieved, the actual impact could far exceed the worst-case prediction of a 2.4 percentage point increase.