(Image source: Unsplash)
At present, the popularity of ChatGPT is declining, and investors are worried about the profitability of the AI industry, prompting AI concept stocks in China and the United States to show signs of turning from prosperity to decline and "correction".
On April 24, China's "AIGC No. 1 Stock" Mobvoi (02438. HK) successfully landed on the Main Board of the Hong Kong Stock Exchange with a floor price of HK$3.8 per share. However, it suffered a break on the first day and opened down more than 21%. The decline began to narrow in the afternoon, closing at HK$3.68 per share on the first day, down 3.16%, and the latest market value reached HK$5.49 billion.
Earlier, the stock of US AI giant Nvidia (NASDAQ: NVDA) experienced a "roller coaster" market, plummeting nearly 15% in a week, and the company's total market value fell by nearly $300 billion (about 2.17 trillion yuan).
As the competition in the AIGC track of the "100 model war" has become fierce, the secondary market questions this field that burns money and is trapped in the problem of commercialization. At the same time, the "negative effects" of the secondary market are constantly spreading to the primary market, causing everyone to ask, what is happening to the AI industry?
The gap between China and the United States is widening, and the monetization ability of enterprises is being questioned
The AI-related capital markets in China and the United States are showing a "carnival" from a year ago to a "calm" situation today.
In the U.S. stock market, in just one week, U.S. AI-related stocks fell. In addition to Nvidia, Supermicro Computer (NASDAQ: SMCI) plunged nearly 20%, Meta fell more than 4%, Amazon fell more than 2%, and Tesla, Microsoft, Apple, and Google fell more than 1%.
In terms of domestic AI companies, iFLYTEK's share price has fallen about 7% in the past week. In the past 12 months, Cambrian (SHA:688256) shares have fallen 44.6%, and Baidu Hong Kong (HKG:9888) and Baidu US (NASDAQ:BIDU) have fallen 18.8% and 19.6%, respectively, over the past year.
If the U.S. stock market is still due to the "mysterious forecast" of supermicrocomputer and the sudden decline of the U.S. stock market, then the performance fundamentals of domestic AI companies are very worrying.
Take iFLYTEK as an example.
On April 22, iFLYTEK (002230) released the first quarter report of 2024, with operating income of 3.646 billion yuan in the first quarter of 2024, a year-on-year increase of 26.27%, a net loss of 300 million yuan, a year-on-year decrease of 418.99%, basic earnings per share of -0.13 yuan, and a weighted average return on equity ROE of -1.76%.
Earlier, iFLYTEK had released its 2023 annual report, and its performance was lower than expected. In 2023, the company will achieve revenue of 19.650 billion yuan, a year-on-year increase of 4.41%, lower than the company's original expected revenue of more than 20 billion yuan, and in 2023, the company will achieve a net profit attributable to the parent company of 657 million yuan, a year-on-year increase of 17.12%.
At the financial report meeting on April 23, the senior management of iFLYTEK also said frankly that the company's accounts receivable are now increasing quite rapidly due to the financial difficulties of the entire government, and the structure of the company's accounts receivable is mainly based on large and medium-sized government financial institutions and large partners. Although bad debts were accrued in 2023 according to strict accounting standards, the actual bad debts as a whole in the past calendar years were relatively low, and in 2020, the actual bad debt ratio was 0.1%, and then decreased year by year. The actual bad debt rate in 2023 is 0.01%, and it should be said that bad debts are controllable.
According to the financial report, in 2023, iFLYTEK's R&D investment will be 3.839 billion, accounting for 19.53% of the total revenue, an increase of more than 400 million year-on-year, of which the total R&D investment of the entire Xinghuo model will exceed more than 20000000000000000000000000000000000000000000000000000000000000000000000000000000
Obviously, poor performance and increased accounts receivable have become the reasons why iFLYTEK is difficult to generate profits.
It's not just iFLYTEK. According to the financial report of Mobvoi, the "first share of AIGC" on the Hong Kong stock market this morning, its AIGC products rely on a subscription-based model, but Mobvoi is currently facing rising pressure from the cost of acquiring new AIGC users. Among them, the customer acquisition cost of registered users has climbed from 1.4 yuan per person in 2021 to 13.5 yuan in 2023, and the cost of paying users has increased from 31.8 yuan per person in 2021 to 133.1 yuan in 2023.
Nowadays, Chinese AI companies have a special "ability": all large models surpass GPT-4, but no one really buys them on a large scale, without any high-tech barriers, and without any innovative technological differences.
Professor Xue Lan, Dean of Schwarzman College at Tsinghua University and Dean of the Institute for Artificial Intelligence in International Governance, said that China has more than 130 large models, and although there is a lot of progress in terms of quantity alone, there are still many problems with China's large models. He mentioned that many large models are built by "shelling" and assembly, and the computing power is also "stuck".
About half of respondents to a survey by the Boston Consulting Group (BCG) (all executives) said they don't expect significant productivity gains from generative AI, and they are concerned about errors and data leaks with generative AI tools.
On the contrary, AI technology companies with a wide range of business scenarios and strong application capabilities will have much better revenue and operating performance, and many businesses have also achieved significant growth on the basis of large models.
A senior industry insider revealed to Titanium Media App that if a large model trained based on massive Internet public data wants to be applied to a vertical industry or business scenario, it needs a large amount of industry or business scenario data in the field to fine-tune.
At the same time, the effectiveness of the application of large models is closely related to the depth of the understanding of the business logic and key factors of the vertical industry or subdivision of the large model talents, which also requires the close cooperation of the large model talents and business experts in the field and mutual understanding of their respective language systems.
Tier 1 investment is declining, and the market is turning around
According to the 2024 Artificial Intelligence Index Report released by the Institute for Human-Centered Artificial Intelligence (HAI) at Stanford University in the United States, which is co-led by Li Feifei, AI investment will decline overall in 2023, but the amount of investment in generative AI will soar to $25.2 billion, an increase of nearly eight times from 2022.
Image source: HAI Artificial Intelligence Index Report 2024
It can be seen from the amount of financing of generative AI that the investment pattern in the AI field is undergoing profound changes, and the players of the basic model are basically a foregone conclusion, and huge investments are concentrated in a few leading companies.
For example, in the first quarter of 2023, Microsoft invested $10 billion in OpenAI, and in the third quarter of 2023, Amazon and Google invested $4 billion and $2 billion, respectively, in Anthropic.
Large-scale domestic investment is concentrated in Zhipu AI, MiniMax, Baichuan Intelligence, etc., and it is almost impossible to complete large-scale financing in others. In the future, as the leading companies consolidate their positions, investment in billion-dollar projects will gradually slow down.
In addition, the "2024 Artificial Intelligence Index Report" shows that in 2023, the investment in the AI industry in the United States will reach $67.2 billion, which is 8.7 times that of China, the second largest investor ($7.8 billion). In the 10 years from 2013 to 2023, the United States invested a total of $335.2 billion in the AI industry, compared with $103.7 billion in China. In addition, the gap between China and the United States is even more pronounced in private investment in generative AI. In 2023, the U.S. invested a total of $22.46 billion in generative AI, while China invested only $650 million.
Zhou Zhifeng, a partner at Qiming Venture Partners, recently said: "In the wave of mobile Internet, new energy vehicles and Internet lending, the amount of investment in the primary market between China and the United States is almost equal, or China is 80% of the United States. But in 2023, the total investment in China's AI market is only about 12% of that of the United States. This is very alarming, and there is still a lot of room for investment catch-up in China's AI market. ”
Behind the trend of investment in the AI field to concentrate on leading enterprises, especially those in the United States, is the affirmation of capital's algorithm talents and model capabilities of leading enterprises.
The R&D and breakthrough of large models first come from professional large model algorithm talents, and the number of talents in China and the United States are in the first echelon, but the specific data reveals a big gap between the two.
According to the "World's Most Influential Artificial Intelligence Scholars" report released in 2023, the number of people selected in the United States is as high as 1,079, accounting for 54.4% of the global total. China also had 280 people on the list, but the number of AI scholars in the United States is nearly four times that of China.
In addition, it is worth noting that as many as 70% of the 217 top AI researchers in the United States are from countries other than the United States, of which 50 are Chinese, followed by India and the United Kingdom. When OpenAI released GTP-4, there were 33 Chinese employees in its public contributor list.
It can be seen that China does not lack top AI talents, but the United States has a large number of top AI research institutions and universities, such as Stanford University and Carnegie Mellon University, which continue to attract talents from all over the world to develop. How to attract and retain talents in terms of policy and environment is a key to future development.
The R&D capabilities of non-leading enterprise models are insufficient, and many need to innovate based on open source software, but many open source protocols are not allowed to be used for commercial purposes, so the open source models that can be selected are very limited, and usually their parameter scale and model performance are still far from real commercialization.
On the whole, domestic and foreign AI technology companies are facing challenges such as computing power, large model research and development capabilities, and scenario application capabilities to varying degrees, while domestic enterprises will have greater challenges in computing power and models.
(This article was first published in Titanium Media APP, author|Wang Jian, editor|Lin Zhijia)