Breaking News: AI unicorn SenseTime raises $600 million in strategic round C financing, led by Alibaba

Today, SenseTime announced through its official platform that it has completed a $600 million Series C financing. This is the largest single round of financing in the field of artificial intelligence, following the $460 million Series C financing completed by Face++ in October last year.

It is reported that SenseTime's Series C financing was led by Alibaba Group, followed by Singapore's sovereign fund Temasek, Suning and other investment institutions and strategic partners.

In this regard, Xu Li, co-founder and CEO of SenseTime, said: "As a world-leading artificial intelligence platform company, SenseTime's Series C financing will further consolidate the company's leading position in the field of artificial intelligence: first, with SenseTime's original technology as the core, it will empower more industries; second, it will conduct in-depth cooperation with global leading partners to further expand its business territory; third, it will strengthen the upstream and downstream industrial chains and deepen SenseTime's layout in the artificial intelligence industry chain."

阿里巴巴集团执行副主席蔡崇信表示：“商汤科技积极拓展人工智能领域，尤其在深度学习和视觉计算方面，其科研能力让人印象深刻。阿里巴巴在人工智能领域的投入已为旗下业务带来显著效益，今后我们将继续在人工智能领域作出投资。我们期待与商汤科技的战略合作能够激发更多创新，为社会创造价值。”

Sun Weimin, Vice Chairman of Suning.com, said: "Smart retail is essentially the digitization of retail elements and the intelligence of retail operations. Therefore, artificial intelligence will play an increasingly important role in the field of smart retail in the future. As a leading domestic industry company, SenseTime focuses on original technology and builds an artificial intelligence business ecosystem, which coincides with Suning's smart retail strategic layout. After the cooperation between the two parties, Suning will apply artificial intelligence to the retail ecosystem and reversely empower artificial intelligence technology through scene Internet and retail big data, bringing a more intimate consumer experience to consumers."

According to an insider of SenseTime, after completing the C round of financing, the company will further strengthen its advantages in industries such as security, mobile phones, autonomous driving and interactive entertainment, accelerate the implementation of technologies in smart retail, finance, education and other fields, expand its business territory, and accelerate the development of artificial intelligence platforms.

As one of the leading investors in this round, at the end of last year, the market had heard that Alibaba had invested 1.5 billion yuan in SenseTime, but SenseTime’s official response was: no comment on market rumors. It was not until March this year that Alibaba and other parties officially finalized their Series C investment in SenseTime, and SenseTime is currently valued at more than 4.5 billion US dollars.

As a target of venture capital funds, SenseTime has been hotly pursued by capital in recent years, and SenseTime's targeted choices have also "risen with the tide." In an exclusive interview with Caixin, SenseTime President Xu Li said that whether it is a financial investor or a strategic investor, the key point of cooperation is whether it can bring "relatively big changes" to SenseTime's business.

For example, before Alibaba invested, the two parties had already had business contacts. In the future, SenseTime will become a partner in Alibaba's smart city project, mainly providing technical support in smart monitoring, smart transportation, and urban management, and forming a large product matrix on this basis;

For example, SenseTime is also a technology provider for Suning's unmanned stores and AI-based ordinary stores.

Another example is Temasek, which appears to be participating in the investment as a financial investor, but in fact, it has very close ties with China's exclusive technology companies and has strong global resources.

At present, SenseTime's technology is widely used and has been implemented in many industries and realized in scale. Previously, the company's business has achieved explosive growth in orders and revenue scale. More than 400 well-known institutions and enterprises such as China Mobile, UnionPay, Huawei, Xiaomi, OPPO, vivo, Weibo, iFlytek, and JD.com have become SenseTime's contracted customers.

The following is a transcript of Leifeng.com’s interview with SenseTime CEO Dr. Xu Li.

Dr. Xu Li, CEO of SenseTime, previously accepted an exclusive interview with Leiphone.com, in which he asked in detail about the application implementation of SenseTime technology, the homogenous response strategies faced by competitors, the evaluation criteria for entering a field, the losses of doing B2B instead of B2C, and other issues, showing us the inner side of this mysterious giant.

Summary of Xu Li’s views:

1. On the technical level, we prefer to position ourselves as competing with original companies such as Google and Facebook in terms of core algorithms and brainpower.

2. To see whether an AI company is doing basic research, there are two necessary conditions. The first is whether it has computing power (such as a large enough GPU cluster), and the second is whether it has invested in data labeling.

3. The key to originality is whether it can bring core competitiveness. If successful, there will be Matthew effect and agglomeration effect, which can bring more and more original benefits.

4. Technological breakthroughs create blue ocean markets; technological leadership brings about product differentiation; in the current AI field, leading algorithms are the key to creating product differentiation.

5. The artificial intelligence industry is a bit like a world of martial arts. The core is to practice internal skills, that is, deep learning brain. Once you have practiced well, you can skillfully use all kinds of weapons (to solve problems in various vertical applications).

6. AI is like building a car, the application is the wheel, and the core technology is the engine. What can bring about industry change lies in the breakthrough of the engine, but only when the wheels are installed can the car run fast.

7. When we are working on artificial intelligence, a very core point is to find industries that were originally manpower or labor-intensive, and use AI to replace manpower, thereby improving the efficiency of the industry.

8. SenseTime-driven means using core technologies to connect with upstream partners and help them expand downstream markets. At the same time, it provides downstream partners with a full-chain one-stop solution to maximize their value in application scenarios.

9. We are doing something that most Chinese companies are unwilling to do, which is to develop original technology. This is also the mission of our SenseTime Technology: to insist on originality and let artificial intelligence lead human progress.

The following is the interview transcript:

Looking at the official website, typical cases include smart finance, smart business, smart security, and Internet+. Are they ranked according to the priorities of your strategic layout?

Xu Li: The core criterion for us to select industries is whether the traditional paths in this industry have completed the closed loop. For example, in finance, commerce, and security, most of their applications are completed manually. Therefore, when we are doing artificial intelligence, a very core point is to find the original manpower or labor-intensive industries and replace them with manpower, so as to improve the efficiency of the industry and reduce costs. The ranking of the official website is not entirely based on the strategic direction, but on the size of the market. Now, the security market is gradually growing. In addition, finance, commerce, and the Internet each occupy a relatively important position.

The four sections mentioned above cover a wide range of areas. How do you solve the problem of having so many competitors?

Xu Li: The key to this question is whether scene data is more important or core breakthroughs are more important. Artificial intelligence is a bit like a martial arts world. Each school has its own martial arts, and the real core is to practice internal skills. The engine of deep learning is actually internal skills. Finance, business, the Internet, etc. are all based on the application of this internal skill.

In the past, each vertical application had to be supported by experts and was isolated from each other. But now, the model has changed, and a more unified and simpler deep learning brain can be used to solve the problems of various vertical applications at the same time. It is like when your internal strength is well practiced, you can skillfully use all kinds of weapons such as swords, guns, sticks, and swords.

At present, we focus on core technology and then radiate to various industries. When the technology evolves very fast, using technology to radiate applications in various vertical fields can bring breakthroughs, which is much better than the incremental polishing of current applications. Technological breakthroughs bring blue oceans, not red oceans. This is why we are at the forefront of the industry in the capital market, industrial cooperation customers, and retail.

Generally, when a large company enters a field, even if it is very hot, if the profit only accounts for a small part of its annual income, it may not bother to do it; small companies will do it if there is business. How do you evaluate the standard of whether to do it or not?

Xu Li: We believe in "long-term stable cash flow". The so-called business model, or the essence of entrepreneurs, is to pursue long-term stable cash flow. This has three guiding significances for our company's industry selection and product definition, which we call "SSD":

The first S is Scalability. We need an industry with volume that can generate cash flow. The second S is Sustainability. Only when there is no sudden demand can we achieve stability. The third D is Differentiation. Differentiation can bring long-term benefits. Even if we don’t have good revenue now, we can achieve future revenue through core competitiveness in the future.

It is not necessary for a company to have a good profit in the early stage, but it must have a clear long-term profit model. The industries we choose are all scalable. As long as my clients have a large business, my income will also be large. For example, if the online business is SaaS, if the volume of the APP can be increased, the calls will obviously increase, and the income will also increase.

You have frequently received huge amounts of financing to build an AI business ecosystem driven by SenseTime in the upstream and downstream of the industry. What problems will you encounter when opening up the upstream? What problems will you encounter when opening up the downstream?

Xu Li: SenseTime is an upstream company in terms of technology, because we do not need any technical authorization from anyone. However, if we want to implement scenarios and products, we need to cooperate with partners, including purchasing customized equipment and infrastructure.

NVIDIA, Huawei or Sugon are our partners and upstream companies. They supply us with hardware and infrastructure. The key point to get through the upstream is not how much hardware is consumed, but how much business opportunities can be opened up. For example, we and NVIDIA are platform-level partners in the field of AI City, and we can push its solutions to our customers very well. So the key to get through the upstream is to tell them that I can help you build the ecosystem.

Usually, downstream partners have application scenarios. For example, one of our partners, Zhongke Jincai, has some applications in finance and smart city commercialization. It is a typical listed company of comprehensive service providers and integrators. When we cooperate with them, we provide them with a one-stop solution, and they can adjust it according to their application scenarios. The more complete the chain you provide them, the greater the value they can play.

So opening up the upstream is to tell them that our solution can help them expand the downstream market. Opening up the downstream is to combine differentiated products with their own scenarios, so that they can use their channels to quickly occupy the market.

In terms of company size, how do you define SenseTime now, a startup or a unicorn?

Xu Li: How to position a company depends mainly on who you consider as your competitor. At SenseTime, we are doing something that most Chinese companies don't do, and it is also our company's mission to insist on originality and let artificial intelligence lead human progress. Why is insisting on originality put here? In China, because the early supply formation did not provide enough protection for the most original technology, everyone is unwilling to do the most difficult and original part, but is willing to do the easier copycat part.

Our competitors in this position may be Google Labs and Facebook Labs. We regard those who truly make original core technologies as our competitors. In China, if no one else is doing this, once we succeed, we will be the only one. So in terms of size, we are a unicorn, but we realized very early that we are not competing with these companies in terms of application scenarios, because what we do is different. We prefer to position ourselves as competing with Google, Facebook and other original companies in terms of core algorithms and brainpower.

Most AI companies have begun to cover finance, security, transportation, and even smart cities based on their original businesses. Now that competitors have begun to become homogenized, what is your response strategy?

Xu Li: Our competitors are homogeneous, but our products and technologies are different.

Our content is different from others, because if everything is homogenized, then the only thing left is to compete on price-performance ratio, either good customer relationships or low prices. If neither is available, then there is no advantage. We believe that in the future, the product technology that SenseTime can provide will be very different from what can be provided in the market, including the product definition, which depends on the differentiation of products brought about by technological leadership.

Someone told me that artificial intelligence has entered the so-called second half. The first half is about technology, and the second half is about application. Technology has become homogenized, and the models are all the same. This is probably evolved from the second half of the Internet. People who say this don't really understand. For example, the generation from AlphaGo to Master has almost completely overturned the original algorithm, greatly reducing the reliance on the initial data of human chess records, and is even less like a "human". Basically, it can be at least three pieces ahead of the original. This is a terrifying thing. When you come up with such a product, homogenization is out of the question. In the financial field, if the error rate is one in a hundred million or one in a billion, it will be three to four orders of magnitude higher than the current error rate of one in a hundred thousand, and there will be no homogenization. The security, finance, and even smart cities we are involved in have very high requirements for security accuracy. It tests the core algorithm and the differences in the entire technology application scenario.

Therefore, algorithm leadership is the ticket to entry. In response to large scenarios and data, you can launch a good product for the business. Then this product comes from algorithm leadership, which is the key point of differentiation.

SenseTime is a B2B2C company, not direct B2C. Do you think there is any loss?

Xu Li: Most B2B models are B2B2C. Others are doing B2C, so its growth will promote the progress of the B side, and the influence of users on it can in turn tell me and promote our improvement. For example, one of our customers is Faceu, which rushed to the top of the AppStore in a very short time. In B2B, this sudden explosion is definitely not as big as B2C. This is a two-sided nature. To maintain security, it is definitely not very disruptive.

At present, SenseTime is taking a more stable path, because we are all scientists, and scientists like things with a higher probability of success. At least for now, we think B2B is suitable for our company.

At the last NVIDIA GTC conference, you mentioned in your speech that "China has a lot of applications, but relatively few are doing basic research. We are relatively persistent in going in one direction. In fact, at the beginning of SenseTime, our investors warned us that if we keep doing original work, it would be a loss-making thing. But I think this is a path that we must stick to, because if you are optimistic about a certain direction, you must get to the front, and then you will be able to see the world from afar." You mentioned that if we keep doing original work, it would be a loss-making thing. Have you found the ideal "original" input-output ratio now?

Xu Li: Originality mainly depends on whether it can bring core competitiveness. For example, when it came to Xiaomi's photo album, many companies were negotiating cooperation with it. Xiaomi put forward many requirements, but for us, it was just to adjust a module, but others didn't know which module to adjust, so they had to test all of them. In the end, only we could solve this problem. So it's very interesting that originality can find a premium in China, and can take orders in this industry through original technology, and find a premium with the breakthrough of technology itself. This is why we are exploring the possibility of building an AI platform company in China.

So once the original technology is done, you will have no competitors, and you can use it to make a premium, to make a subversion, to form an industry barrier. The core is to insist on being original, and to be profitable through originality. Whether the first step can be successful is critical. After success, it can produce Matthew and agglomeration effects, bringing more and more original benefits. We are not doing artificial intelligence because of the artificial intelligence craze. Many algorithm breakthroughs are actually a process of accumulation. So we told investors about our accumulation. After reading it, some investors immediately decided to invest in us exclusively. Originality is the point that impressed them. After they found such a good accumulation of technology, they began to turn around and said, "At this point, the industry has an opportunity for a breakthrough. If it is to succeed, this must be a point."

How do you make good use of the huge GPU computing cluster?

Xu Li: Because we started deep learning very early, we had to build this supercomputing platform from scratch, from hardware connection to hardware algorithm platform deployment. Our first large-scale cluster platform was around the end of 2015, and it achieved relatively good results. We have now connected several clusters, and the utilization rate of GPUs is very high. So it can be seen that our company is a computing resource company.

We believe that GPU is the key to building the productivity of the brain. Building an artificial intelligence brain is the same as designing a human brain. It requires a lot of iterations, which actually requires computing resources. Therefore, the investment of core resources in this regard can fully show whether you are a company that really does basic research. Investors asked me: How do you know whether an AI company is doing basic research? I said look at two questions. The first is whether there is a large enough GPU cluster. If there is no computing power, how is the basic research calculated? The second is to see if there is data annotation. If there is no data annotation, how does the machine learn? If there is no computing resources and no data annotation, it is obvious that you are just using other people's algorithm results to run forward. From this perspective, investors can easily find out whether a company is really doing core basic research.

You just said that AI is rarely mentioned this year, and it is directly called AI+, which is the application of landing. But you still think that algorithms are very important at the summit. In a previous interview with Hua Xiansheng of Alibaba, he said that AI algorithms are not everything, and special attention should be paid to landing. What do you think?

Xu Li: Implementation is very important. The breakthrough in AI is actually closely related to the implementation of products. It can quickly bring about an explosion in the industry, which I very much agree with.

But what is the driving force? For example, in the case of a car, the application is the wheel. Without wheels, the car cannot be driven. First, the wheels must be installed, and the car can drive very fast. But in the end, whether it can drive far or not actually depends on the engine, which is very critical. What can bring about changes in the industry now is still the breakthrough of the engine, but at the same time, we must land the product as quickly as possible before we can roll it out. The fastest way to land a product is to do it ourselves or to find a partner. All products cannot be completed by one person. A large system is also called a product. The system involves integrators, camera manufacturers, and component manufacturers. In the process of packaging them into a product, it involves the cooperation of multiple companies, which can quickly understand the strengths and weaknesses of each other's technology, learn from each other's strengths and weaknesses, and form a framework. At this level, SenseTime has done relatively well, and we have formed such a large ecosystem.

What are the 3 things you must do every day at work?

Xu Li: There are mainly a few things, both internal and external. What I have to do every day is to deal with emails (laughs) and sort out the tasks of the entire company. This is necessary. Then there is the sorting out of business and the sorting of priorities. AI is changing the world, and the entire industry and business atmosphere are constantly changing. So we need to sort out strategic directions and priorities frequently. But the only thing that remains unchanged is our core competitiveness of technological breakthroughs. The third is how to optimize internal management efficiency and the entire system. Ambitious goals also require refined management. This is also the main problem that we must face every day.