In 2023, who is making money from AI?

Image credit: Generated by Unbounded AI tools

Source: Qian Hao Channel

AI is splitting the world in half.

With the advent of artificial intelligence applications such as ChatGPT, breakthroughs have changed many industries and labor methods. The creative mental work that was once considered irreplaceable is ushering in decomposition and reconstruction, and "inspiration" has lifted the veil of mystery. In the fields of games, music, and design, AI has already begun to occupy a place. According to media reports, some companies that are deploying AI in large quantities are expected to save 200,000 yuan a year.

AI has also brought a huge wealth effect to the primary and secondary markets. From suppliers of computing power and chips, to developers of large models, and even related concept stocks at the application layer, they have all experienced skyrocketing revenues or leaps in valuation. Brooke Day, technology portfolio manager at Goldman Sachs, said: "I've been investing in technology for over 30 years, and artificial intelligence is one of the most exciting developments I've seen.

Such duality is exactly the characteristic of AI as an industry outlet: it can burn money crazily and improve efficiency; it not only challenges traditional industries, but also creates new wealth myths; Send it to the wind.

For us, how to seize the AI outlet? Before considering countermeasures, we must first think about a question, why does AI suddenly break out?

A Brief History of AI

AI is like a milestone on the highway, "When you see it on the horizon, it's here in no time."

The origin of large models can be traced back to the 1980s. In the following decades, with the development of computer performance, the depth of neural network models and the size of training data slowly increased. By 2006, the deep belief network model was proposed, which opened up ideas for deep learning models. In 2017, the Google team developed the Transformer architecture. The scale of deep learning parameters reached hundreds of millions, and the real large model was born. At this time, it has been less than 5 years since the birth of GPT-3.

Generally speaking, the development characteristic of the AI industry is "slow accumulation, sudden explosion". Unlike new energy and the Internet, which require a long period of infrastructure construction and product popularization, AI has the advantages of easy deployment of infrastructure such as chips and cloud services and low marginal cost, so the core of its development is technological progress, as long as the technical parameters reach the threshold. , then its application scale will expand rapidly.

**Also, ChatGPT is far from the whole content of artificial intelligence. **In foreign countries, Meta (Facebook) and Google are closely following the pace of Microsoft; in China, from Baidu's "Wen Xin Yi Yan", to Ali's "Tong Yi Qian Wen", from Tencent's "Hun Yuan" to Huawei's " Pangu". Since March this year, more than 20 companies have entered the large-scale model track, including general-purpose large-scale models that benchmark against GPT, and vertical large-scale models that focus on vertical industries and are trained on the basis of open-source large-scale models.

In addition to the "technical layer" industry involving algorithm research and model building, the artificial intelligence industry chain also has two parts: the basic layer and the application layer. As the name suggests, the base layer is the infrastructure that mainly provides hardware devices and software services, mainly involving AI chips, cloud computing and other fields; while the application layer mainly covers application scenarios that require artificial intelligence technology, including finance, automobiles, and medical care. The latter is a company that focuses on various specific scenarios.

It can be seen that AI has passed the initial stage of focusing on the development of general-purpose large models, and has moved towards a new stage in which the basic layer pursues low-cost and high computing power, the technical layer moves from general-purpose to vertical, and the application layer realizes landing value in different industries. For huge industrial opportunities.

**However, just because the opportunities are so huge, there are elusive opportunities hidden under the quiet water. **

**When we invest in AI, what are we investing in? **

**Like many outlets, there are difficulties and more obstacles behind AI investment. **

As mentioned earlier, the core driver of AI is not popularization, but technology. Therefore, once the technical level is not as advanced as expected, or cannot support actual application scenarios, the development of the AI industry will become a "beautiful-looking" illusion.

GF Securities believes that it is difficult for the domestic capital market to influence and determine the future trend of AI industry investment, which is mainly affected by two external factors. On the one hand, the industrialization practice breakthrough of domestic large-scale model platforms, on the other hand, is the overseas generative AI business If the realization of globalized technology, especially the commercialization of industrial applications fails to meet expectations, the investment logic of the entire AI industry will be shaken.

The essence of these two points is the accuracy and cost of AI content, which must reach a level that can be commercialized in specific scenarios.

On the other hand, because of the characteristics of AI development that stops moving and breaks out, it puts forward higher requirements for investors' forward-looking. Obviously, different investors will have different opinions on how to separate the false from the numerous AI investment advice. The failure to form an investment consensus means that there will be a "continuous rhythm" in the field of AI in the capital market.

Huaan Securities pointed out that the market valuation market hoped to be the first to reflect the possible "exceeding expectations", but different investors have different tolerances for time lags in performance realization. Adjustment.

**The consensus among investment institutions at this stage is that AI will be the main line of investment this year and for quite some time in the future, and people's daily life will be the primary object of AI change. **Of course, under such trend recognition, different capital markets have different investment priorities.

In the primary market, because the investment peak in 2021 has just passed, the attitude of investment institutions is relatively cautious. Some investors said, "The AI track has reached a high valuation before, but the business model is not clear, so in It's going to be cold for a while."

Therefore, even with the explosion of large-scale models and the heat up on the track, institutions still tend to seek "certainty", such as computing power chips, low-level large-scale models with independent controllability, and large-scale model companies in vertical industries, etc. Both "infrastructure" and "scenario application" are the directions that institutions focus on. Data show that since 2023, there have been 254 domestic AI investment events, with investment and financing amounting to 100.6 billion yuan. Among them, computer vision and images, intelligent robots, intelligent driving and other subdivisions are highly popular.

In the secondary market, the domestic capital market also has a large number of companies involved in the AI field. For example, the 50 index sample companies selected by the China Securities Artificial Intelligence Theme Index from the Shanghai and Shenzhen markets cover the basic layer, technology layer, and application layer of the artificial intelligence industry chain, including household appliances, automobiles, etc. that involve artificial intelligence extension application scenarios. industry. Most of these companies are concentrated in the two major industries of computer and electronics, with a combined weight of 90%. It can be seen that most investors in the secondary market still regard the computer and electronics industry as the driving force for the development of AI.

Most of the AI-related companies come from the computer and electronics fields, but companies that are not from the electronics and computer fields are related to AI, and how to bet on companies with potential is obviously the most concerned issue in the secondary market.

**The most stable investment strategy is not to bet. **

**In the future of AI, how many golden houses are there? **

Lu Qi, former CEO of Baidu, believes that from 2022 to 2023, the new paradigm has ushered in a new inflection point, "model" knowledge is ubiquitous, and the technological innovation represented by OpenAI makes the trinity of action, model and information, transforming towards general intelligence . "This inflection point of large models will affect all people and blue-collar workers in the service economy, because they are models. Unless you have unique insights, there are large models of services you are engaged in today."

**In the secondary market, compared with betting on a certain company in a certain track alone, sector index funds may be more likely to obtain growth dividends in subdivided tracks. **

As a technology that has brought tangible efficiency improvements to the industry and society, and is still rapidly popularizing and differentiated, AI has become an unstoppable macro trend, and its suspense comes from the specific development path. Astute investors are good at keeping up with trends, but they are often not good at grasping the implementation of specific products and services of a company. Therefore, compared to buying shares of individual companies, "investment trend" is a better choice, which can not only enjoy the opportunities of industrial growth, but also avoid the pressure of stock selection within the sector.

The Wind data also proves this point: in the ten years from June 29, 2012 to February 3, 2023, the China Securities Artificial Intelligence Theme Index, its annualized rate of return exceeded 10%, and its annualized volatility It was 32%, and the Sharpe ratio reached 0.46. At present, there are also 3 ETF products on the market that track the China Securities Artificial Intelligence Theme Index.