Has China’s AI bubble burst?

Shortly after the 20th Party Congress opened in Beijing on October 16, Yìn Jī 印奇, the founder and CEO of Megvii Technology 旷视科技, a company that develops artificial intelligence (AI) technology for businesses and the public sector and one of China’s “Four Little AI Dragons,” stated in an interview that the Congress “pointed out the direction for China to grasp the core AI technologies in our own hands.” AI, according to Yin, is a driving force for a new technological revolution and industrial transformation, and Megvii will continue to contribute to national development with self-reliance to promote Chinese-style modernization.

Yin did not elaborate in this interview about the existential challenges facing not only China’s AI industry but also his own company. Megvii was blacklisted in 2019 by the U.S. for its involvement in human rights abuses, but it has even bigger problems closer to home, such as the difficulty of raising money. Back in March 2021, Megvii submitted an application to list on the Science and Technology Innovation Board of the Shanghai Stock Exchange. But the company’s application is still in progress after multiple enquiries by the stock exchange’s review committee.

Megvii’s IPO application also suffered a setback when the company’s chief scientist, Sūn Jiān 孙坚, who was mentioned 32 times in the prospectus, died suddenly in June 2022. But the ongoing concerns with Megvii’s application reportedly focus on two main problems: the company’s shareholding structure — in particular it’s relationship with Ant Group 蚂蚁集团, it’s largest shareholder; and the same problem facing all of China’s AI companies: Profitability.

Investor exuberance has ended

After a long period of rapid growth, from 2021, the capital markets have cooled on the AI industry: According to a Chinese investment fund manager, AI investment and financing transactions in China increased from 16.52 billion yuan ($2.28 billion) in 2011 to 399.64 billion yuan ($55.30 billion) in 2021. Over the same period, the average amount of a single AI investment increased from 150 million yuan ($20.75 million) to 353 million yuan ($48.84 million).

But from the fourth quarter of 2021 to the second quarter of 2022, however, AI investment in China has dropped by about 70%. According to the fund manager, the market has not necessarily given up on AI technology, it has just become more discerning, i.e., investors are paying more attention to the actual value proposition of AI.

Thus it’s no longer sufficient simply to have the best AI technology, what matters now is how to make use of that technology. As one computer science professor from Tsinghua University put it this week, “If smart AI workers were looking down from the top of the building in the past, now they have to enter the factory, roll up their sleeves and work for others.” In other words, the days of easy money are over, the AI industry now needs to focus on solid application scenarios that actually make money.

A vicious cycle: More revenue, less profit

The problem with China’s AI companies comes down to a simple but insoluble formula: Revenues are up but so are research and development costs, and profits are falling as companies struggle to transform AI technology into marketable products.

4Paradigm 第四范式, a machine learning company that develops platform products focused on computing power and operating systems used in various industries such as energy, finance, and telecommunications, is a case in point:

In August 2021, the company submitted an application to list on the Hong Kong Stock Exchange, but it too has been unsuccessful so far. It’s not difficult to see why. According to the company’s prospectus, from 2019 to 2021 and the first half of 2022, revenue was 460 million yuan ($63.65 million), 942 million yuan ($130.35 million), 2.01 billion yuan ($279.25 million), and 1.05 billion yuan ($146.40 million), respectively. However, over this period, 4Paradigm reported a net loss of no less than 3.85 billion yuan ($533.32 million). Most of the company’s expenses were devoted to marketing and advertising, and paying high salaries, a common practice in the AI industry.

For many other AI companies, the demands of research and development are overwhelming. From 2019 to 2021, AI Speech 思必驰技术, for example, a company that provides software and hardware products for smart home appliances, cars, and consumer electronics, reported revenue of 115 million yuan ($15.91 million), 237 million yuan ($32.79 million), and 307 million yuan ($42.48 million), respectively. But over the same period, the company declared a net loss of 822 million yuan ($113.74 million), because over the three years, research and development costs accounted for 173.35%, 86.26% and 93.25%, respectively, of total revenue.

The takeaway: A difficult road ahead

High research and development costs and big net losses are common phenomena in China’s AI industry. AI companies are now under great pressure to launch new products on the market, but full commercialization relies on a long and costly research and development cycle, and even if the products are launched on the market, this may still not ensure the expected returns, and the company may continue to suffer losses. For many AI companies, this ultimately results in a vicious cycle of high costs, high losses, diminishing prospects in the capital markets, and ever greater pressure to commercialize with diminishing funds.