(Yicai) Oct. 8 -- Salaries for PhD graduates in artificial intelligence from China’s elite universities are climbing, with compensation hitting several million Chinese yuan (millions of US dollars) and more, amid surging demand for talent in the field.

PhD holders from top labs at Peking University and Tsinghua University who join AI projects at big technology firms can earn multi-million yuan annual packages (cash, stocks, and options), one AI startup’s founder told Yicai, adding that some AI master’s degree graduates can also command daily remuneration in the thousands of yuan.

TikTok owner ByteDance, whose Beijing headquarters is less than 10 kilometers from Peking University, is a major draw for its graduates. Yang Ling, an assistant professor in the International Machine Learning Research Center at Peking University, said many of its postgraduates have interned at ByteDance.

Some Peking University PhD students have already taken deep dives into ByteDance’s research projects during their studies, Yang said. “Many interns hold important roles within ByteDance’s research teams and may even participate directly in model training,” he noted.

A top-flight researcher can save a company from much futile model training, Yang said, explaining the reason for the hefty salaries. A single experiment can cost several million to tens of millions of yuan, making high annual salaries well worth the investment, he pointed out.

The AI startup’s founder echoed that sentiment, saying that an annual salary of several million yuan looks reasonable as long as the hired expert can help the company’s AI projects make chip use more efficient.

“Large models represent the most valuable track right now, and professionals in this sector are the most valuable talent of this era,” the founder said.

“Many PhDs from Peking University and Tsinghua University now go straight to [large tech firms] to take on core tasks, directly replacing the previous generation of tech talent,” Yang stated.

But even this talented cohort could be replaced by even more capable people in a couple of years’ time, according to Yang, who expects the AI talent shortage to ease over the next five years. Universities will produce more AI graduates, and model developers will cultivate in-house talent pools, while industrial expansion may hit a ceiling, and demand for talent will not keep expanding at the current rapid pace.

Yang said the demand for AI talent will become sharply polarized. “Only the very top tier — those capable of delivering genuinely high-value outputs -- will remain critical. The importance of everyone else will decline,” he said.

Yang cited AI coding platforms as an example. They are making the work of junior programmers dispensable. “Codex can handle the tasks these people used to perform,” he said, referring to the AI coding platform developed by OpenAI.

This means universities must adapt to meet the challenge. “Peking University is changing, and other universities are following suit,” the startup’s founder said. “Higher education institutions must integrate with industry. Without such integration, they will struggle to secure research funding, and students will fail to gain practical knowledge.”

Editor: Tang Shihua