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

The corresponding business costs of ArcSoft include the costs of technical personnel responsible for product integration; the corresponding business costs of VeriSilicon include the costs of customized IP licensing and the personnel costs required for the customized IP combinations and technical support provided in this type of business.

(2) The gross profit margin of the cloud-based smart chip and acceleration card business is higher than the average level of comparable domestic companies. The main reason is that the company's cloud-based smart chips are general-purpose smart chips in the field of artificial intelligence and are priced higher than general dedicated functional chips.

The gross profit margin of cloud-based smart chips and acceleration card businesses is higher than Nvidia's overall gross profit margin, mainly because Nvidia's overall business includes products with relatively low gross profit margins, such as consumer graphics cards and terminal SoCs.

(3) The gross profit margin of the intelligent computing cluster system business is higher than the average gross profit margin of the corresponding business segments of comparable companies. The main reason is that the Company’s intelligent computing cluster system uses self-developed cloud-based intelligent chips and acceleration cards, as well as basic system software platforms. Therefore, the gross profit margin is higher than that of comparable companies in the same industry.

Unlike the significant scale effect in the traditional chip field, the AI ​​chip field presents a unique pattern: there are currently almost no companies listed on the market with AI chips as their main business, and most AI chip companies are still in the customized technical service stage, which is still some distance away from large-scale mass production.

Companies whose businesses involve AI chips can be roughly divided into three categories:

The first is large companies, such as Alibaba and Huawei;

The second is traditional chip companies, such as Allwinner Technology and Rockchip;

The third group is start-up AI chip companies, such as Cambrian and Horizon.

In general, although many companies have released AI chip products, both smart driving and smart security require algorithm development based on specific scenarios. Therefore, AI chips are highly customized based on scenarios and have not yet reached the point of large-scale commercial use.

Since Cambrian was founded only recently, it is not very meaningful to look at the overall revenue and growth rate. China's chip war has just begun.

On the other hand, let’s look at the performance of Megvii Technology.

Megvii believes that its self-developed Brain++ deep learning framework is the core of its innovation and provides important support for the algorithm training process. With artificial intelligence algorithms, Megvii has created innovative platform software and application software to meet the needs of end users in different vertical fields. Megvii's self-developed artificial intelligence and software capabilities can empower IoT networks (including cloud centers, edge servers and/or IoT devices).

Combined with the relevant financial data as of June 30, 2019 (2019H1), we can have a preliminary understanding of Megvii's operating status.

As the data foundation brought about by industrial digitalization becomes increasingly mature, the implementation of artificial intelligence in industries such as marketing, finance, digital government, retail, and healthcare continues to advance and has begun to bring significant benefits, but the progress of implementation has varied.

Judging from the development of artificial intelligence, smart cities and digital governments are relatively mature fields, and we will focus on them in our analysis.

In its prospectus, Megvii cited a research report from Frost & Sullivan, which stated that the market size of China's smart city management vertical is expected to increase from RMB 11.9 billion in 2018 to RMB 103.1 billion in 2023, based on revenue, at a compound annual growth rate of 54%.

Among them, the answers to the non-AI technology companies 1 and 2 that have the largest market shares are self-evident.

Let’s compare the profit margins of Megvii, Hikvision and Dahua in the field of urban Internet of Things.

During the reporting period, Megvii's gross profit margin fluctuated greatly. The gross profit margin of urban IoT solutions increased from 34.8% in 2017 to 64.8% in 2018, and then fell back in 2019.

The explanation given in the prospectus is that due to business growth, better economies of scale can be achieved and costs can be controlled, and the solutions are more popular with customers, so pricing can be increased.

However, Megvii has a limited operating history, and its current gross profit margin may not necessarily serve as an indicator of future gross profit margins.

In the field of urban security management, Megvii is in a race where giants such as Hikvision and Dahua have long been in the game. Its smart logistics also has to face the pursuit and blockade of emerging warehousing robot companies such as Geek+ and Kuaikan.

The wave of big data and artificial intelligence has been rising for several years, but there are only a few companies that have truly established a sustainable business model and are able to achieve self-sustaining growth.

Whether Megvii Technology can maintain this momentum in the future depends on whether it can cherish every moment.

03

What are the disruptive innovations in business models?

The vast majority of AIoT companies are based in the B2B enterprise sector. The product, technology, and sales thresholds in this field are relatively high, and the competition is less intense than that in the B2C consumer sector.

If AIoT companies remain in a homogeneous business model for a long time, although they may be slightly better in technological innovation, it will still be difficult for them to ultimately win out among many competitors, and the probability of project investment failure will be high.

So have AIoT unicorns brought disruptive innovation to business models?

Judging from Cambrian's prospectus, the company has always adopted a direct sales model and has a dedicated sales team to communicate with customers in a timely manner.

Under the direct sales model, the company directly participates in the customer's public bidding or business negotiations. After reaching an agreement, the company directly signs a sales contract with the customer.

Its competitor Rockchip adopts the sales model of "distribution as the main and direct sales as the supplement".

Under the distribution model, distributors purchase chips from Rockchip and sell them to device manufacturers or solution providers, and purchase components and sell them to end customers such as electronic product developers or enthusiasts.

Under the direct sales model, Rockchip sells chips directly to OEMs and solution providers, or provides professional technical services; and sells components to end users such as electronic product developers or enthusiasts.

According to iResearch’s research, the sales models of artificial intelligence companies can be roughly divided into four categories.

API call: Common among basic layer vendors and general layer vendors, which output their own technical capabilities in the form of API.

For example, SenseTime in the field of computer vision, Baidu Apollo platform in the field of autonomous driving, iFlytek in the field of speech recognition, etc., all export artificial intelligence technology to application manufacturers, who then complete the final step of product and solution packaging.

The advantage of this model is that it is lightweight and has strong scalable replication capabilities.

Product Subscription/License: mainly products for individual users in the form of robots, apps, etc., and products for Internet customers and small and medium-sized customers in traditional industries in the standard SaaS model.

For example, companies such as DJI and Squirrel AI mainly use this method to serve individual users.

"Product + Service" solution: mainly for medium and large customers in traditional industries. The application scenarios of such customers are relatively complex, and it is difficult for a single product to meet their needs, so a certain degree of customized services is required.

For example, companies such as Megvii Technology and Mininglamp Technology serve clients in the public security field and need to provide end-to-end solutions.

Pay by results: After AI is combined with business scenarios, charges are made based on the actual measurable business value it generates.

Artificial intelligence companies and their customers have more of a similar cooperation model, charging a certain fee based on business volume. Currently, there are some early applications in the fields of finance and intelligent customer service, where applications are relatively mature.

For example, intelligent customer service vendors measure their effectiveness based on how much labor costs they help corporate clients save, and can charge according to workload and number of seats.

If we take a detailed look at Megvii Technology's products, commercialization and operating models, they generally fall into the above-mentioned business models.

However, Megvii Technology’s exploration of business models does not seem to stop at the common ones mentioned above.

In the book "Intelligent Internet: New Thinking", I mentioned the ecological business model.

Under the instigation of building an "ecosystem", the typical practice is to try to operate the entire industrial chain, with upstream and downstream layout, vertical integration, and mutual matching to form an "ecosystem" for the entire industrial chain.

AIoT unicorns can play the role of ecosystem builders, laying out the entire industry chain of basic computing capabilities, data, general algorithms, frameworks and technologies, as well as application platforms and specific solutions, gathering a large number of developers and users, and promoting the formation of an ecological business model.