With less than five years of operating history, SenseTime has closed more than 10 rounds of financing by well-known investors from around the world, Tiger Global, Silver Lake, Fidelity, SoftBank, Qualcomm, Alibaba and Temasek Holdings. The valuation has skyrocketed after rounds of funding, leading to both positive perceptions and questions on the feasibility of SenseTime’s business model: The team consists of scientists and professors intent on publishing academic papers and participating in global AI competition, but it is weak on monetization, and its vaunted “AI for everything” strategy seems impractical at this stage.
Comparatively, other three Computer Vision startups in China, Megvii (originally known as Face++), CloudWalk, and Yitu Technology remain much more “low-profile” in public, though they have also drawn millions from investors. For example, Megvii, another Alibaba-backed startup and the biggest competitor to SenseTime, completed its Series D funding round last year, but chose to postpone the announcement until May 2019. With the intention of "cooling the market”, the company repeatedly emphasized that news of a new funding round did not carry the same weight as a major technological breakthrough. An AI business, just like companies in other sectors, should have its own business cycles.
Moreover, the rest of the three startups are narrowing their focus to merely a few industries, convinced that startups worth less than USD 10 billion should focus on specific verticals instead of doing everything. In addition to using facial recognition technology in public security and consumer-oriented mobile apps, Megvii mainly focuses on logistics; Yitu specializes in healthcare; CloudWalk has chosen the financial sector. Different from its peers, SenseTime has been exploring over 10 verticals and is “impatiently” looking to apply AI to more industries, including self-driving, surgery and diagnostics, augmented reality, chip technology, AI education and so on. During the 2019 Artificial Intelligence Summit on May 15, SenseTime released 11 products across five industry verticals, the largest number of new products unveiled at a single event since the company was established.
How SenseTime turned into the most expensive AI startup
Despite the mixed comments from different groups, we believe SenseTime’s rapid ascent into the world’s highest-valued AI startup is driven by the factors listed below,
1) The “best-in-kind” team. The company has grown to 3,000 employees now, doubling the number from a year ago. The team consists of PhD holders and college professors from MIT, Chinese University of Hong Kong, Tsinghua University and Peking University as well as tech experts from Microsoft, Google, Lenovo and Baidu. Everyone in the core development team has over 20 years of research experience. According to the founder TANG Xiaoou (汤晓欧), “There are less than 200 people with PhD on deep learning technology in China and SenseTime employs 120 of them. If SenseTime cannot offer the technology, the rest of the players have even smaller chance to do so.”
2) Greater government support. SenseTime has become the fifth National Open Innovation Platform for Next Generation Artificial Intelligence, after Alibaba, Baidu, Tencent and IFLYTEK. The Chinese government reportedly backs Baidu on autonomous driving, Alibaba Cloud on city brain, Tencent on healthcare medical images, IFLYTEK on Natural Language Processing (NLP) and SenseTime on Computer Vision technology. While the other four companies are tech giants with stable revenue streams, SenseTime is a new entrant to the list. We expect the policy to give SenseTime greater access to subsidies and increase the likelihood for it to land government contracts and hire more AI talents than its competitors.
3) The “AI for everything” plan underpinned by financial clout. Relative to its peers, SenseTime has the most funds available to deploy significant resources and support its aggressive expansion strategy. According to the company's vision, if it does not take initiatives now, the hotly contested market will be quickly taken over by emerging players in the next few years. Even if SenseTime might not have a substantial impact on certain areas, the company intends to build industry connections and brand awareness. When technology and solutions become more mature, corporate clients are more likely to choose SenseTime over vendors with less visibility.
4) Being a frontrunner on revenue and profitability. Though the specific revenue or net income figures are not disclosed, our industry researchers confirmed that in 2017, SenseTime posted revenue of around CNY 700 million and became profitable, just like Megvii. In 2018, sales reached CNY 7 billion, well exceeding all competitors'. Most of the revenue has come from government clients, with whom SenseTime could leverage its strong connections.
Among all sectors that SenseTime is currently exploring, we highlight education, Smart City and entertainment, areas we think SenseTime has competitive advantages in.
Governments are chasing AI education solutions
Recently, SenseTime introduced its internally designed AI textbooks, “Fundamentals of Artificial Intelligence”, and distributed them to around 40 high schools in Beijing, Shanghai, Shenzhen and Shanxi. To spark students’ interests, its platform “SenseStudy” offers over 200 hours of AI and programming courses. Students can experience and learn about facial recognition, gesture recognition, augmented reality and many other trending technologies.
It is the governments who are spending millions of dollars on SenseTime’s education solutions. The reason why the government introduced AI to K12 education was that while US has been promoting STEM education (Science, Technology, Engineering and Math) for decades, China suffers a shortage of engineers, physicians and data scientists. The edge in scientific and technological innovation has become only more important amid the current escalation of US-China trade war.
We believe that with SenseTime’s strategic partnership with the government-affiliated China Education Association, it is relatively easy for the business to secure additional orders and supply more schools with AI textbooks, training and lab equipment. In this sector, the path to monetization is clear and will likely be sustainable in the long run.
Smart City is a big pie with a small threat from new entrants
Visual surveillance systems installed in venues ranging from train stations to airports regularly receive technological updates. Besides public security, the scope of Smart City is even broader, including vehicle license plate recognition, traffic data identification, crowd flow analysis and so on. Cities and provinces are delegating Smart City projects to different vendors, who win contracts through competitive bidding.
While 2017/2018 was the time for AI vendors to tender bids for contracts, companies which have won deals are in the process of fulfilling their contractual obligations and providing services now. Since public security, which makes up a big portion of Smart City projects, involves considerable sensitive information, governments are unlikely to grant such projects to a provider without a trusted relationship. Therefore, in a market worth around USD 80 billion with double-digit growth rates, there are limited opportunities for new entrants.
SenseTime has developed particularly strong relationships with both domestic and foreign governments, demonstrated by 1) it won contracts with over 100 cities, nearly a third of all cities in China, and has installed over 100,000 smart cameras. 2) it has built research centers globally and has undertaken strategic cooperation with Southeast Asian countries, selling Smart City solutions to Thailand, Malaysia and Cambodia. It operates a heavier business model than its peers, who develop their businesses remotely or have yet to expand overseas.
AR brings higher user engagement to the entertainment sector
SenseTime helps mobile game apps, social video apps, and other user-generated content (UGC) and other video platforms blending innovative human-computer interaction with user-generated content via augmented reality (AR) technology. For example, Tencent's first AR mobile game "catching the demon” was launched last month; it adopted a cutting-edge mix of computer vision, depth detection and real-time object recognition, provided by SenseTime. On account of declining user engagement and rising user churn rates, mobile app providers are looking for ways to increase user “stickiness”. If an app operator starts to use AR technology to attract users, its competitors are likely to make the same move or it risks driving its users to other platforms and losing market share.
Downside risks
Under this business model, downside risks are easily identified. Trying to cover many sectors is likely to distract SenseTime from concentrating resources to develop industry expertise - the real moat - and may lead to a diminished role for the company in certain industries, such as autonomous driving, which requires years of development.
In addition to the government clients, SenseTime claims to have teamed up with over 700 corporations in China and overseas, such as Qualcomm, Nvidia, Huawei, Xiaomi and JD.com. However, the value that SenseTime creates for its enterprise clients may not directly be reflected in their financial statements. For example, Meitu, a leading photo-editing app which heavily uses SenseTime’s facial recognition technology to let users beautify their photos, failed to prevent its shares plummeting despite using AI and even blockchain technologies. It is challenging for SenseTime to drive profitability by serving only one or two industry leaders who have deeper pockets.