Is this a unicorn that can take advantage of the momentum, or a gold-swallowing monster that can't walk into the spring breeze?
Outlook Oriental Weekly reporter Chen Rongxue editor Jin Mingdae.
On July 6, 2023, at the Shanghai World Expo Exhibition and Convention Center, the 2023 World Artificial Intelligence Conference officially opened.
The wind rises at the end of the green ping, and the waves become between the waves. With the wave of disruptive changes set off by OpenAI, AI (artificial intelligence) has once again come to the forefront of capital and the market.
According to the "2022-2027 Chinese Artificial Intelligence Industry Demand** and Development Trend Prospect Report" released by the China Commercial Industry Research Institute, as of November 24, 2023, there have been 531 investment events in the field of artificial intelligence in China, with an investment amount of 6604.8 billion yuan. In 2023, the market size of China's artificial intelligence industry will be 337 billion yuan, and it will reach 401.5 billion yuan in 2024.
On the one hand, ChatGPT (a large language model trained by OpenAI) has triggered an entrepreneurial boom, and on the other hand, the 2023 ** Economic Work Conference clearly stated that "leading the construction of a modern industrial system with scientific and technological innovation", "accelerating the development of artificial intelligence", "encouraging the development of venture capital and equity investment", the superposition of technology and policies has made artificial intelligence entrepreneurs and investors stand on the cusp of a new round.
Facing the wind, how risky? Where are the opportunities? How to fly over the barriers of high investment, high threshold, difficult implementation, and long cycle of AI enterprises? With these questions in mind, Oriental Outlook recently visited a number of investment institutions.
AI investment is growing against the trend.
In 2023, the financing of Tus-incubated enterprises will exceed 10 billion yuan, 7 billion yuan in 2021 and 9 billion yuan in 2022. Wang Jiwu, chairman of TusHoldings (hereinafter referred to as Tus), said in an interview with Oriental Outlook that investment in China's artificial intelligence field is booming.
China's AI looks at Zhongguancun, Zhongguancun looks at Haidian, and Haidian looks at Tsinghua Science and Technology Park. Sitting on the enlightenment of Tsinghua Science and Technology Park, it has incubated more than 20,000 high-tech enterprises, with more than 80 ** branches, a management scale of 40 billion yuan, and has invested in nearly 2,000 enterprises, most of which are related to artificial intelligence. According to the report data of the China Business Industry Research Institute, it can be seen that the Tus-sector accounts for the top 1 7 in terms of new investment in artificial intelligence in the country.
Different from the enlightenment of "bringing its own AI genes", Cornerstone Ventures, a PE (private equity**) cadre under BII Holdings, originally focused on rail transit infrastructure, but in the past year, it has also invested "nearly 2 3 businesses in artificial intelligence-related fields". Huang Libo, founding partner of Cornerstone Ventures, told Oriental Outlook Weekly.
In 2023, the complex and volatile global economic situation has brought a lot of pressure to the investment and financing market, and issues such as the double decline in fundraising and investment, valuation downgrades, IPO tightening, and difficult exit have become the focus of attention of many investors and enterprises.
Zheng Qingsheng, China Partner at Sequoia Capital, said: "This is the beginning of another technology cycle. "Against the backdrop of declining overall venture capital, the proportion of investment in artificial intelligence has grown strongly against the trend.
According to the KPMG report, the number of global AI venture capital cases and the proportion of financing amount in global venture capital have increased year by year, accounting for 18 in the first half of 20239%, a new high in recent years.
Unicorn? Gold Eater?
Looking back at 2023, many institutions shouted "all in AI", and the news of star founders devoting themselves to AI entrepreneurship or large-scale financing of AI star projects continued to come.
IT Orange Data said that at present, there are 102 unicorn companies in the field of artificial intelligence in Chinese, and 10 new unicorns will be promoted in 2023, including Zhipu AI, Baichuan Intelligence, Zero One Everything, Minimax Name Dream, etc. (excluding the "light years away" that was newly promoted as a unicorn but soon withdrew from the list due to acquisition). Among them, Baichuan Intelligent has 3 semi-annual financing500 million US dollars, with a valuation of more than 1 billion US dollars, setting a record for the fastest promotion of domestic large-scale model start-ups to unicorns.
According to Silicon Valley venture capitalist Erin Lee, a unicorn is defined as "a start-up valued at more than $1 billion." In this regard, Shen Quanhong, vice president of TusHoldings and chairman of the ** business department, explained: "Unicorns are piled up with capital, and the industry needs a banner. ”
Investors want to be the brave catcher behind the unicorn, but is this a unicorn that can take advantage of the momentum, or a gold-swallowing monster that can't walk into the spring breeze? This is a test for every investor.
The "lesson" of the AI Four Tigers is just around the corner.
SenseTime Technology (hereinafter referred to as SenseTime), which has the halo of "AI No. 1 Share", received 12 rounds of financing and blood transfusion in the seven years after its establishment in 2014, and landed in the capital market on December 30, 2021, with a market value of up to HK $320 billion. However, according to the prospectus and financial report, SenseTime (by mid-2023) has accumulated a loss of 46.8 billion yuan in the past five and a half years.
Yuncong Technology, which is also one of the four tigers, has not been able to prevent its losses from continuing to expand.
Shen Quanhong said that if there is no performance that matches the valuation, the valuation will become a stumbling block for the sustainable financing and development of the company.
The report from research firm Pitchbook, a leading provider of private equity and venture capital databases and data analytics in the United States, is glaring: since 2021, more than 400 unicorns around the world have not raised a new round of funding, and about 94% of tech unicorns are not profitable.
The four-legged robot of Unishu Technology, a member company of Sequoia, was unveiled at the opening ceremony of the "Sequoia Digital Intelligent Industry Incubation Center".
The Law of Twenty-Eight. At present, large models are the track with the highest attention in the field of artificial intelligence, and China has also become the country with the largest number of large models in the world.
At the Xili Lake Forum held in November 2023, Chairman Robin Li said that as of October 2023, China has released 238 large-scale models, a threefold increase in four months compared with 79 in June. Many industries, enterprises, and even cities want to train their own dedicated large models from scratch.
Looking back at the PC (personal computer) era and the mobile Internet era, various PC software was developed based on Windows (the operating system developed by Microsoft), and the only operating systems in the mobile era were Android and iOS. He believes that it is a great waste of social resources to constantly re-develop the basic model.
In this regard, Zhu Xiaohu, managing director of GSR Ventures, said that more than 200 large models will soon enter a convergence period, and there may only be one or two dozen large models left by the end of 2024.
There is a 28 rule in the investment world, and only two or three out of 10 projects are likely to succeed. For the top few who are optimistic about the track, investors tend to lay them out together. Shen Quanhong said that previously, Tus-Group had invested in SenseTime, Cambrian and Fourth Paradigm.
In the end, there may only be one or two large models in demand. Huang Libo said frankly, "We will not invest for the sake of investment, after all, the four AI tigers have not yet made a profit." ”
Therefore, under the boom of large models, Cornerstone Ventures chose to invest in "shovels" - front-end tools and model applications.
There are three criteria for my investment projects: first, there must be a good team, and technology is the first; Secondly, above the threshold of hard technology, there must be a future market space; Third, enterprises should have a good governance structure and equity structure to develop healthily in the long run. Huang Libo said that each investment manager of Cornerstone Ventures will look at hundreds of projects every year, and after screening, he will look at 70 or 80 projects a year.
In his eyes, the ideal project is to be able to go public and exit or apply for an IPO (initial public offering) in five years, with an annualized IRR (internal rate of return) of not less than 30%.
How high can the returns be?
How high can the return on investment be for "survivors" under the 28th rule?
The 100-fold return on investment in Jumei made Sequoia China's Shen Nanpeng the "godfather" of e-commerce; IDG Capital's Zhou Quan and Lin Dongliang have invested in Soufun.com for 11 years and returned 100 times.
Founded in 2005 in Tsinghua Science Park, GigaDevice was successfully listed with the incubation investment of Tus, and has grown into a giant in the chip industry in 11 years. The return of Tus is more than 200 times.
Compared with the return of more than 200 times, we are more willing to tell the story of how we and Hefei ** contributed to the linkage development of GigaDevice and Hefei Changxin Storage, and solved the technical problem of "stuck neck". Shen Quanhong said that unlike the consumer sector, the development of technology companies is not linear. Losing money in the first 10 years, and breaking out in the 11th year (income), there are too many such things, and you can't use the development of the first 10 years to improve the future benefits.
This is a great test of whether investors can accompany them for a long period of time, and whether they have the ability to identify and tolerate faults. That's why Tus-built more than 80 different types**. He said.
Science and technology investment is not to be capricious if you have money, and it is not to burn money to have a future. A national-level person in charge said that in order to tap AI companies with growth value, he asked the team to deepen front-line research and be an "expert investor".
Everyone is optimistic about the large-scale model track, and we have also conducted research, but for each project, the team background, technical path, and application scenarios are different, and it is difficult for investors to measure its advantages and disadvantages. The person in charge said that in comparison, there are more innovative enterprises and investment opportunities in vertical industry models and industry applications.
Regarding the profitability problem of high-tech enterprises represented by AI, many respondents said, "The difficulty lies in getting out of the laboratory and landing applications." ”
People who do R&D like to do technology, and a lot of innovation is actually innovation for the sake of interest, but the market needs to be applied, how to get through? Shen Quanhong said that TusPark has built more than 300 science and technology parks around the world for more than 20 years, in order to provide real front-line scenarios and demand data for technology development.
For example, in the unmanned driving track, Tus-discovered the demand for urban sanitation through scientific and technological innovation network information feedback, and invested in incubating Xiantu Intelligence. Now, Xiantu Intelligent's unmanned sanitation has gone to sea to Saudi Arabia.
The whole Middle East welcomes such AI tools. Shen Quanhong said, "Technology companies need not only investment, but also platforms and bridges, and need a third hand to integrate resources and support innovation." ”
January 19, 2023, Nanjing, Jiangxinzhou Jiangdao Zhi Cube Industrial Park.
Risk of cyclical mismatch.
At present, many investment institutions clearly shout "invest early, invest small, and invest in technology", but the earlier you invest, the higher the risk, which is also the general consensus of investors.
Everyone talks about high risk and high reward. I agree that investing in AI companies is a high return, but I don't agree that it is a high risk. Wang Jiwu told the reporter of "Outlook Oriental Weekly", "Tus-has analyzed the big data of enterprises in the past 30 years, and found that the profit cycle of high-tech enterprises generally takes 10 to 15 years. Therefore, the risk of investing in AI companies lies not in the advancement of technology, but in the mismatch of investment cycles. ”
The ** cycle of most investment institutions is set at about five years, which is generally shorter than the growth cycle of AI companies. Huang Libo said that because of the background of state-owned capital such as Beijing Investment Holdings, the cycle is set at seven to ten years.
He said frankly that there is relatively little old money in China that is willing to invest in a long cycle, everyone is used to making quick money, and the willingness of private capital to participate is not high.
The above-mentioned national ** person in charge said that in addition to capital investment, it is more important for strategic investors represented by national ** to play a strategic guiding role and guide funds to invest in strategic emerging industries.
Whether we can use new models to transform traditional industries and whether we can solve the problem of 'stuck neck' is the primary consideration of our investment. He said.
For some strategic highlands that need to be vigorously miraculous projects, you might as well smash them. Wang Jiwu said that some places have been explored, such as the establishment of Jiangsu Industrial Technology Research Institute and Suzhou Laboratory, to support the transformation of scientific and technological achievements with the strength of the whole province.
As for market-oriented AI companies, how to solve the problems of high R&D investment, high professional threshold, difficult transformation and implementation, and mismatch of investment cycle? Wang Jiwu's suggestion is: "Can we learn from Silicon Valley Bank's approach and set up a specialized bank for entrepreneurs and venture capitalists?" ”
Click on the title below to read all the articles in this special section.
2024, Three Questions on Artificial Intelligence" special series.