Why SemiAnalysis believes the first $1 trillion in AI spending will not come from white-collar automation

A SemiAnalysis representative argued that replacing white-collar workers will not drive the first $1 trillion of AI spending: companies are slow to change job responsibilities, compensation and staffing, so adoption in those areas will take time. He estimated that the cost of a given level of AI capability falls by 100x to 1,000x a year, and said the fastest scaling will be in fields where additional intelligence can generate a quick return.

He named chip design, cybersecurity, financial services, life sciences, materials science and software engineering. In science and engineering, he said, much remains beyond the human frontier; in finance and cybersecurity, competitive rivalry sustains demand for intelligence. He predicted that OpenAI, Anthropic and other leading labs could approach $1 trillion in revenue within a few years. If fields requiring frontier models do not generate that demand, he said, the market could quickly saturate existing tasks while white-collar replacement diffuses too slowly—potentially making it harder for AI labs to meet their compute commitments.