Unicorns in the field of AIoT are experiencing a wave of listing

Typical application layer scenarios include smart homes and smart connected cars .

Representative companies in the smart home field include Tuya Smart, Orvibo, Broadlink, and Lumi, all of which have received multiple rounds of financing.

Representative companies of intelligent connected vehicles include Pony.ai, Roadstar.ai, WeRide, iDrive, and UISEE.

Most domestic AI companies start from the general layer or the application layer, and with the accumulation of technology and the expansion of business, the boundary between the two layers becomes increasingly blurred. But overall, AIoT startups have two development paths:

Take an application scenario (such as smart home) as a breakthrough point: obtain data by connecting to the internal system of enterprise customers or self-built scenario portals such as sensors, continuously train models and optimize algorithms based on multi-dimensional data, find the best solution in a certain scenario problem, and then replicate it to similar scenarios in other industries;

Use a general technology (such as machine vision) as a breakthrough point: dig deep into algorithms and underlying frameworks. Especially when machine learning is accepted by the industry, drive the training model from the bottom up, which can not only improve the universality and computing efficiency of the solution in different scenarios, but also ultimately improve the actual application effect.

The former has a deep understanding of the scenarios, making it easier to reach customers and accumulate data, so their products are more easily accepted by customers and have stronger monetization capabilities.

The latter hopes to leverage the advantages of algorithms and underlying frameworks to efficiently reach more industries, empower industry partners, and obtain data through open cooperation. It may not be able to directly reach customers in the process, so its coverage is wide but its monetization capabilities are relatively weak.

Overall, according to MMC Ventures' research, the ratio of AIoT startups based on application scenarios and general technologies is about 9:1, and most companies enter the market through vertical application scenarios.

Based on the customer groups, AIoT companies can be divided into two types: those providing services to businesses (to B services) and those providing services to end consumers (to C services).

The toB service is designed for enterprise users, focusing on business value. The functional logic and information architecture are relatively cumbersome and strict, which is reflected in how to enable enterprises to protect their rights and information, improve work efficiency, and bring economic benefits. The key points of the business model include: improving efficiency, reducing fees, and pricing power. The essence of a toB company is to help a certain link in the entire industry chain improve efficiency or reduce costs to gain profits.

Consumer-oriented toC services are designed for ordinary users, and such products are often provided to users free of charge. The focus is on user experience, and the logical process is kept as simple as possible, which is reflected in how to make it more convenient for users to use the product and make them more willing to use the product actively.

Yao Xin, an investment partner at BlueRun Ventures, believes that people often have a misconception: "If it is to B, the most advanced technology should be used, while if it is to C, cheaper technology should be used. But in fact, to C has higher requirements for technology, and it often requires excess technology, that is, technology that has matured to the point where it is very cheap today."

The artificial intelligence industry is still in its early stages of growth, so there are relatively more to B companies, while to C companies are in a more difficult situation. Therefore, a new technology that has just been put into use is not suitable for starting a business for the C end.

According to statistics, this is indeed the case. In the field of artificial intelligence, the ratio of B2C to B2B companies is 1:9, and the vast majority of companies enter from the B side.

In terms of industry maturity, over the past year, artificial intelligence has gradually entered the trough of the technology maturity curve, the bubble of technology hype has burst, and the focus of industry attention has begun to shift to how artificial intelligence can be implemented in the industry.

According to Gartner's research, the proportion of companies planning to implement artificial intelligence projects will increase from 1/25 to 1/3 in three years. In its latest report, Market Guide: Chinese Artificial Intelligence Startups, Gartner believes that China's artificial intelligence (AI) start-ups are playing an increasingly important role in the development of various AI applications.

Therefore, some research institutions believe that the adoption of artificial intelligence technology has crossed the chasm, expanding from innovators and early adopters to the early majority, and will have a profound impact on businesses, consumers and society.

02

How is the market performance of AIoT unicorns?

Back four years ago, 2017 was called the first year of the AI ​​industry.

This year, Cambricon received strategic financing from Qualcomm Ventures and Alibaba, of which Alibaba’s investment amounted to 1.5 billion yuan, setting a record for the highest single financing amount in the AI ​​field at that time.

In the prospectus, Cambrian mentioned:

"Since its establishment in March 2016, Cambricon has achieved the industrialization of technology output. Cambricon 1A and Cambricon 1H are used in the flagship smartphone chips of a world-renowned Chinese technology company and have been integrated into more than 100 million smartphones and other smart terminal devices."

“As of February 29, 2020, Cambrian has 65 domestic and foreign patents that have been authorized, 120 PCT patent applications, and a total of 1,474 domestic and foreign patents that are being applied for.”

Megvii Technology's performance is also very impressive:

“Since the beginning of 2017, we have won 22 world championships in multiple international AI competitions and have performed outstandingly in multiple computer vision projects. We provide customers with full-stack solutions including algorithms, software and AI-enabled IoT devices, and have taken a leading position in multiple industries.”

“In 2017 and 2019, we were included in two of MIT Technology Review’s most recent “50 Smartest Companies” lists. ”

However, whether a company can achieve healthy and sustainable profits is the real skill of the company. How much of such impressive technical strength and scientific research data can be converted into actual revenue and profit? Compared with "traditional companies", what advantages do AIoT unicorns have in terms of financial indicators?

Let’s look at the Cambrian period first.

The company has recently been widely criticized for its high customer concentration.

This has been explained in the prospectus: in 2017, 2018 and 2019, the sales amount of the company's top five customers accounted for 100%, 99.95% and 95.44% of the operating income respectively, and the customer concentration is very high.

According to the analysis of Bu Rixin, founder of Chuangdao Investment Consulting, the problem is more serious than imagined.

First, the prospectus mentioned that in 2017 and 2018, the company's sales to its largest customer "Company A" accounted for the highest proportion. After obtaining authorization from the company, Company A integrated the Cambrian terminal intelligent processor IP into its flagship smartphone chip.

Obviously, this "Company A" refers to the major customer Huawei.

In 2017 and 2018, it can be said that Cambrian's main source of revenue came from sales to Huawei, which accounted for 98.34% and 97.63% of its total revenue. This was its only hope.

However, there is a high probability that Huawei, a major customer, will turn to independent research and development.

So Cambrian turned its attention to "Company B".

In 2019, Cambrian's sales to "Company B" were comparable to Huawei's, accounting for 14.38%. The document disclosed that "Company B" was a related party of the company, and Cambrian provided acceleration cards and R&D trial products to this customer. Analysts believe that "Company B" is most likely Lenovo or Sugon.

Finally, and to no surprise to the industry, Cambrian's revenue mainly comes from projects with several local government backgrounds, and there are factors of a relationship or strategic nature involved, rather than pure commercial needs.

In 2019, Cambrian reached relevant cooperation on intelligent computing cluster systems with Xi'an Fengdong Yixiang Technology Service Co., Ltd., Shanghai Brain Science and Brain-Inspired Research Center and Zhuhai Hengqin New Area Management Committee Commerce Bureau, and achieved sales revenue of RMB 296.1815 million, accounting for 66.72% of the main business revenue.

The prospectus shows that Cambrian achieved operating revenues of 7.84 million yuan, 120 million yuan, and 440 million yuan in 2017-2019, with year-on-year growth rates of 1,392% and 279.4%, respectively.

In 2017, 2018 and 2019, Cambrian's comprehensive gross profit margin was 99.96%, 99.90% and 68.19% respectively, showing a downward trend. The main reason was that in 2019, Cambrian expanded its cloud-based smart chips and acceleration cards, and intelligent computing cluster system businesses. The gross profit margins of these two businesses were lower than those of terminal smart processor IPs.

In the prospectus, Cambrian compared the gross profit margins of its main products with those of competing products from comparable companies.

During the reporting period, Cambrian's gross profit margin was higher than that of comparable companies. The explanation given in the prospectus is:

(1) The gross profit margin of the terminal intelligent processor IP business is higher than the average level of comparable companies. This is mainly because the technologies involved in the company's R&D projects are cutting-edge technologies. The software tools and hardware platforms purchased for chip R&D and design are not targeted at specific customer projects and can be used by the company in multiple projects and multiple links. The R&D expenses related to the terminal intelligent processor IP are included in the current R&D expenses.