Shulou(Shulou.com)06/02 Report--

Https://36kr.com/p/5185094.html

2019 may be the best time for mergers and acquisitions and industrial chain integration in the AI field.

By Sun ran

Technology entrepreneurs have realized that money is no longer so easy to get.

In 2019, valuation myths will still exist, but they will only focus on head companies or subdivided emerging markets such as NLP (Natural language processing). Some people pin their hopes on Science and Technology Innovation Board, but the warning of investors is: Science and Technology Innovation Board is still a big variable, do not bet your life, self-hematopoiesis is the reliable way out.

For most entrepreneurs, 2019 will be the year to spend carefully, find ways to get orders, and survive first.

A large number of AI companies face the same dilemma: whether to choose to be a "project company" or a "product company"? How to scale, not limited to doing small and beautiful customized business? Should we be a big B customer, a small B customer or a government business?

Entrepreneurs in different situations have different choices. But their consensus is to find a niche vertical market and make a benchmarking case first. after all, general-purpose AI products still have a long way to go.

In an exciting change, Tencent announced its move into the Internet industry in the second half of 2018. BAT officially bet on AI, which means more investment, traditional corporate customers will be more willing to pay, small and medium-sized entrepreneurs can also take advantage of Dongfeng to get a number of dividends.

So it's important to seize the moment. 2B business is different from 2C, which emphasizes the coordination and resource integration of upstream and downstream industrial chains. Giants need partners to do detailed work, and startups need giant endorsements to get big orders, which means more investments and acquisitions will occur in 2019.

It's hard to get the money, and technology entrepreneurs have to be better at accounting.

In the subdivision track that has run out of the head company, the Matthew effect is intensifying, and it will still be the head company with a certain business scale that will get the money in 2019. The opportunity for startups to get money exists in new niches, such as NLP (Natural language processing).

Many enterprises began to regard Science and Technology Innovation Board as a way to get money. However, the suggestion of investors is that Science and Technology Innovation Board does play a role in promoting enterprises, but it is only a way. For those companies whose business is not solid enough, but regard landing Kechuang as a top priority, there is a big gap between high valuation and business value, which may be a hidden danger in the long run.

Go out and ask founder Li Zhifei's focus this year is on innovation, revenue and efficiency. The latter means paying more attention to meticulous operations. From the end of last year to the beginning of this year, he quantified the operation of the entire company into a series of tables and models.

Not only go out to ask, many AI companies have begun to do internal management optimization. "in the past, we also paid great attention to this, but did not make a very quantitative model, and this year we can no longer invest as casually as we did in the past. On the one hand, we have to expand the scale of revenue, on the other hand, we have to support the efficiency of these revenue pairs, because in the end, the company has to make money." Li Zhifei said.

On the way of urgent innovation and benchmarking, many AI companies have encountered customers who do not have a strong willingness to pay, but also listed as a misjudgment of strategic projects. In 2019, these entrepreneurs are smarter and no longer willing to pay for useless work.

At the same time, this group of technology-trained AI entrepreneurs and CEO are also learning how to be good managers, not just good scientists.

Who has a chance in the cold winter? The first person to land on a large scale

For AI, commercialization is a proposition that has continued since last year.

The core of commercialization is production and landing. When the purse is tight, investors and customers can no longer be convinced by algorithms and demo alone, and whoever can do a good job landing first will be able to get out the fastest.

Compared with the previous two years, today's AI entrepreneurs will not talk about doing general-purpose AI, but choose to start from the vertical field.

In every field, entrepreneurs are doing benchmarking projects. The strategies of DuerOS and Tmall genie are to demonstrate the technical level with their own smart speakers as the benchmark project, and then attract third-party manufacturers to settle in one after another. On the other hand, when it comes to small companies, it is necessary to make reliable benchmarking projects with reliable customers.

Last year, entrepreneurs may not have counted investment in order to be a benchmark case, but that attitude changed in 2019. When the money is spent, there will be a hole in it, and entrepreneurs are becoming more and more cautious in paying for innovation, which is a year for them to test their business models. Not only the product quality itself, but also the production efficiency, cost, stability, the possibility of scale, and the willingness of enterprises to pay the bill have become the dimensions considered by CEO.

Because most investors no longer have the patience to pay for the sophisticated technology that landed more than a decade later.

Another common question is what kind of customer orders should be made for 2B (divided into big B and small B), 2G and 2C?

In China, 2G orders are the most attractive in terms of scale and sustainability. As a low-level technology, AI has a wide range of contacts with 2G, and the rise of four computer vision unicorns (Shantang, Xiangshi, Yuncong, Yitu) has taken advantage of the increased procurement of security products by the government. 2G gross margin is high, but the problem is that experience is difficult to replicate.

More people choose 2B, the same confusion is that it is difficult to jump out of the "project company" and spread out on a large scale. In addition, entrepreneurs find that big B customers need better sales experience, and people who strive to break through the difficulties of standardization and scale choose to be small B customers, but the hidden worry is that the threshold is not high and the future revenue scale is limited.

No matter which path you choose, the ability to undertake large projects, the stability and accuracy of project operation will be the key to distinguish AI in 2019.

The entry of giants and the emergence of industrial communities

2019 may be the best time for mergers and acquisitions and industrial chain integration in the AI field.

As early as a few years ago, some investors noticed that BAT began to do project mergers and acquisitions for financial systems and retail information companies. Now, with the official promotion of BAT and unicorn-level technical service providers, these groundwork have finally come to the front of the stage.

An old question has been thrown out again: will the 2B industry be like 2C, once the giant comes in, the winner takes all, or it is not big, and once it gets bigger, it will be acquired.

A common view is that winner-takes-all will not appear. The industry Internet chain is very long, involving many participants, no matter who wants to do it alone, it is bound to be inadequate. Investors also put forward a new concept-investment industry community.

To build an industrial community, the giants who have the ability to take the lead are often giants who master technology and C-end user data. And getting orders is the most tangible benefit of small and medium-sized entrepreneurs around the giants.

"A lot of startups don't have the resources to undertake larger projects and provide more terminal implementation capabilities and service capabilities, and now BAT comes in to help small companies do this, [their role] is a bit like an upgraded version of traditional integrators, but better for these small companies." Zou Yanshu, executive director of Huaxing New Economic Fund, said to 36 krypton.

For example, iFLYTEK sees its open platform as a sales channel, where B-end customers come to shop, and iFLYTEK is a bazaar that allows customers to recommend products from small startups and endorse their brands.

Stand into the giant ecology, more or less can get a piece of the pie. Some people regard this relationship as "attachment", others call it "synergy", and it also depends on the benefit distribution of the platform and the subdivision of technical service providers. In 2B business, if the platform wants to survive, it must do a good job in the benefit distribution mechanism of each link.

Companies that have obtained investment from giants may also reap orders tied up by some giants. Another kind of valuable capital collaboration will take place in small and medium-sized companies huddling for warmth. When it becomes harder to get money and two companies with their own achievements choose to merge, they are more likely to be the first to dominate the market, and this kind of integration will occur frequently in 2019.

The following are the complete views of ignoring Technology CEO Yinqi, going out and asking CEO Li Zhifei, Youxuan CTO Xiong Youjun, and Zou Yanshu, Executive Director of Huaxing New Economic Fund:

Zou Yanshu, Executive Director of Huaxing New Economic Fund:

AI was proposed in the fifties and sixties and experienced several peaks and troughs, and each trough appeared because something in its commercialization had not been verified as expected. 2018 is a time to adjust and verify the AI industry. By 2019, many people have asked whether the AI bubble burst, but I think it's all right. If you want to promote an industry, it is always possible to bring it out with a relatively large capital investment at first, as is the case with 2C in the past, as is the case with AI.

The economic situation is declining, so now is the time to verify whether AI is reliable and whether it can replace manpower to improve efficiency. In this process, if you can verify yourself and constantly consolidate your business model, you will go a step further. If it doesn't work, it will be eliminated.

Validating business applications can be divided into many dimensions, such as efficiency, cost, stability, scale, and so on. The most fundamental thing is to see whether the market is willing to pay for your technology or products. If a technology has always said that I am very advanced, but can not fall to the ground, can not be realized, then they are not the projects pursued by most investors. Of course, it does not rule out that there are some sophisticated technologies that require long-term investment in manpower, energy and financial resources, but there must be a prediction of whether there is a great long-term value, as well as the possibility of practical application. When the market is bad, investors may not continue to invest, so whether the company has a positive or close to positive cash flow is very important to us-first of all, you have to have the ability to make blood, you have to survive. Now I think 2B companies will pay special attention to this.

For 2B companies, 30 million (verified that the company's products can be landed), 100 million (the ability to commercialize), 300 million (there are certain barriers and large-scale ability), each is a threshold. If I can achieve an income of more than 300 million, I will make a big adjustment to its valuation.

Not limited to AI, some of the more vertical To B projects are also starting to make money, such as smart customer service and HR management systems. However, some small To B companies with heavy offline services generally have a problem, that is, the cost of large-scale management will become higher and higher, and many companies will die in the process of large-scale expansion. Can do some company size can achieve 300-2 billion, but achieve 10-2 billion of the mass, there is a bottleneck-that is, whether the threshold is high enough, in the face of low-price spoilers do not have a strong enough moat.

The timing is very important if entrepreneurs want to rush out. The biggest customer in China's 2B business is actually 2G. For example, Security has spawned 360 and many computer vision companies. I think the next wave of opportunities is that the government may have a need for visualization of some businesses and data governance; in addition, natural language processing (NLP) may have the result of disrupting the industry and giving rise to new opportunities. But in areas that big companies like BAT have focused on, if the entrepreneurs in the same field are still on the scale of start-ups and tens of millions of dollars in revenue, it will be difficult to do so unless there is a big technological breakthrough.

In addition, in the process of landing, technology entrepreneurs should consider whether what they have done is most needed by the market at this stage, and then consider making some compromises in the process of production-- who has cooler technology? it is not seen that it is the core of distinguishing the future survival style of entrepreneurs, and the core is whether they can undertake large-scale business.

For example, if you serve hundreds of thousands of cameras at the same time, it is very important to maintain good stability and accuracy on this basis. Among them, if there is a person with greater ability to undertake, the customer will be more willing to give him more projects, constantly improve his ability, and will be able to rise in the future. For companies that can only accept small orders, the competition between their projects will only be fiercer.

Disregard Technology CEO Yin Qi:

This is a more focused year for AI, but focusing requires preparation and strength, focusing on an industry represents what you need to do deeper, and not all AI companies have such genes. We have an idea that if we do not start from the application to do the platform is rogue. If Microsoft can't do Windows without Office, Office is Microsoft's most profitable product. Unable to create real value for end customers, but saying that you are a platform company is a mistake that most AI companies may make when they start. We may have made that expression, but we want to be a solid company, so we've been trying to find a scene to fit in.

The extensive scene around science and technology is AI+ supply chain, which is subdivided into three layers: 1, intelligent manufacturing; 2, intelligent logistics; 3, intelligent retail.

I think Internet genes are the most important thing to have two genes: the first is the technology gene. After the information reform, almost all companies say that they are technology companies. Every technology company has its own core technology genes. For example, the technology genes of BAT may be applications, mobile Internet, and online data operation capabilities. We believe that the most important technology genes of industrial Internet are AI capabilities. We also have such AI capabilities, and genes may grow a lot of things in the future.

The second is commercial genes. If we look at the three waves, the first wave is the IT wave, the second wave is the Internet wave, and the third wave is the Internet of things wave, what we need is 1-2 type of compound talents. IT talent, very particular about customer service, very pragmatic, pay attention to do vertical, do heavy; the Internet field has a very new thinking, talking about alliance, using capital to accelerate and so on. One is heavy, the other is light. In the industrial Internet, we need a new group of people who have both light and heavy thinking. GLP, for example, used to be in real estate and the heaviest business, but it also has the lightest models of finance, investment and technology, so it stands out from the industry. So the companies that really do industrial Internet in the future will be that kind of gene in a sense.

The core competence of ignorance is the ability of algorithms, the understanding of the industry, and the ability to build such an operating system, as long as we insist on doing our things well, there will be a relatively broad space. The scene of IoT has just begun, and we haven't run in the first half. Even for giants like BAT in the past, it is very difficult for one person to do everything in such a burgeoning Internet of things industry. The Internet makes people feel like there is only one winner-take-all, but I think in the industrial Internet era in the future, business will be a more collaborative network, and the business model will determine how this ecology will be built in the end. it will be better for companies to cooperate in the industrial chain.

Of course, when the company is big, everyone will have some cross boundaries, which is not terrible, the most important thing is to know where their core is. Including giants like BAT, there are a lot of new business attempts, and the boundaries are changing all the time, but if you think clearly that the core competence is the strongest in this field, you can make friends with most people, really build barriers, and guarantee your business value.

Go out and ask CEO Li Zhifei:

In 2019 we will seek change in three things: innovation, revenue and efficiency. I've been thinking about innovation, thinking about the innovation efficiency and team setup of companies like Google, Apple and Samsung, as well as the differences with domestic companies.

The innovation of AI can be divided into three categories: first, the innovation of the lowest core algorithm, focusing on the frontier areas of science and technology. Second, AI engineering innovation, to do things in academic papers need to do a lot of optimization, to do a good job of this step requires not only to understand the international cutting-edge algorithms, but also need to understand the needs of products and scenarios. The third is AI product innovation, the vast majority of Chinese companies may be in this area of innovation, because you do not have enough scientific research personnel, it is difficult to do international cutting-edge algorithm research, at this time, if there is a specific need, you can reverse what tools can solve the problem.

I think for Chinese companies, the second type of innovation is the most worthwhile to invest in, which can bring greater barriers to competition in the future.

In addition, we have made a lot of efforts on revenue analysis recently. To see what kind of revenue each business can achieve, and how to design a team to support it. Even make a quantifiable model of how much R & D is invested and what it is expected to be in the next 18 months.

Many companies are optimizing their internal management in 2019. In the past, many AI companies would invest unprofitable in order to do PR, or invest casually with interest, and many strategic projects would lose money and come to an end. But after this year, people will become more and more rational. On the one hand, we need to expand the scale of revenue, on the other hand, we need corresponding efficiency to support these revenue, and the real money has to be spent in the right place before it is sustainable.

Innovation should also pay attention to efficiency and innovate in a disciplined manner. In the past, one of the biggest difficulties I encountered was that everyone would come and say that we are strategic projects and there is a shortage of manpower, but we do not have a macro concept of which project is really necessary to increase investment. Now we use a quantitative way to look at the input and output of a project, every attempt to innovate will be very rational to Argue.

You must choose CTO Xiong Youjun:

In fact, there is a rule in the development of robots, that is, first To G, then To B, then To C, because the government plays a very important role in promoting this area, especially in the Chinese market, and there is a lot of demand. However, what kind of business the company does depends on the specific industry. before making a decision, we must make a market assessment to see if the industry is big enough and worth doing.

The must-have humanoid robot, Walker, has changed a lot from the version shown by CES in 2018: it used to be a biped robot, but this year it has added the upper body again, and now it is a complete humanoid robot. This means that it will have more space to interact with the physical world.

Some key issues have been broken through here. For example, the most difficult balancing technology in upright walking, as well as the hand-eye coordination technology of robots.

We have been investing money in developing the core technology of robots and doing some four-to-five-year or even more long-term research. Because in the future, robots may have other ways to drive, not necessarily electrode-driven, AI-driven, maybe electronically driven or bionic muscle start. These technologies are forward-looking, but they are also risky.

Walker is 1.45 meters tall, and we want it to be more intimate and natural when interacting with people like a child. If it is too high, there will be a sense of pressure, it is already a mechanical thing. If you are too short, you will not be able to see what is on the table, and your vision and interaction will be limited, and you will not be able to accomplish some tasks. Walker still has some time to go on the market, need to do some optimization, if you want to enter the family, the price will probably be the price of a mid-range car, 200000 yuan-300000 yuan. With the technological progress in the future, the price will be further lowered and really enter the family.

For example, the robot initiates services such as hanging schoolbags and clothes for you after users enter the door and bringing you umbrellas on rainy days. At present, we will plan them by ourselves. We hope that through AI technology, it will continue to learn these capabilities and provide proactive services in the future, but it will take a process.

The interaction between machine and human is divided into expression, posture and language. Now humanoid robots have been able to recognize other people's emotions such as surprise, joy and anger, and then make their own expressions in return. In addition, it also begins to recognize other people's age, gender, and family roles. On this basis, we can do targeted interaction in the future.

At present, there are still some universities and research institutes that can do development in the field of robots. Youbixuan is currently working with schools or partners to do some application development with robots, such as security inspection, smart home, shopping guide. In cooperation with commercial companies, the products are not yet on the market.

I can't give a definite time when the humanoid robot will enter the home, but judging from Walker's testing in the home scenario, it will be very fast. Humanoid robot in the early stage (slow development), because the surrounding industry is underdeveloped, artificial intelligence technology has not made much progress, but Walker is now only more than three years, has made great progress, and then the industry iteration will be accelerated.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.