Last weekend, Professor Xiaoou Tang, the founder of sensetime, passed away. As we reflect on the legacy of this pioneering figure, questions arise about the future trajectory of SenseTime, a once-prominent AI company facing turbulent times, especially in China’s evolving AI landscape. 

In this episode we delve into: 

  • The growth of SenseTime since 2014 to become one of the biggest independent AI company in China;
  • Grizzly reports accusations against SenseTime for engaging in revenue fabrication;
  • The current challenges faced by SenseTime – increasing market competition, tightening data regulations, on-premise deployment requirements, etc.;
  • Lessons to learn from SenseTime’s predicament – importance of product market fit, cautioning against unsustainable cost structures, and pitfalls of meeting growth expectations without a solid foundation.

Featured materials:

SenseTime Financial Results 

We Believe SenseTime is Artificially (Un)Intelligently Inflating Revenue Through Highly Questionable Revenue Round-Tripping Schemes, Grizzly reports.


Jianggan 00:00

Hello, this is Jianggan from Momentum Works. Welcome to the Impulso Podcast. 

Today is the 18th of December 2023 – I am recording this commentary about the development of artificial intelligence companies in China. 

Do check out our insights, reports and perspectives at 

Last weekend, Professor Sean Tang, the founder of Sensetime, one of the most notable AI companies from China, passed away. He was 55. 

The AI industry in China has been mourning Professor Tang’s passing, which came at a difficult time for SenseTime, the company he founded in 2014. 

Since IPO on 30 December 2021, SenseTime’s share price has dropped almost 80%. 

Last month, short seller Grizzly Research published a report called “We Believe SenseTime is Artificially (Un)Intelligently Inflating Revenue Through Highly Questionable Revenue Round-Tripping Schemes” – in which it made a number of accusations. We will come to that later. 

Professor Tang was one of the pioneers of deep learning and AI in China. An MIT graduate, he founded the Multimedia Laboratory at Chinese University of Hong Kong in 2001. In 2014, he built SenseTime. 

The Chinese name of SenseTime, Shang Tang, is that of the first king of Shang Dynasty, who ruled China in the 16th century BC. According to historical records, Tang was a wise king who oversaw a period of prosperity. 

Over the years, SenseTIme has grown to be one of the biggest if not the biggest independent AI companies in China, leaving its main competitors Megvii, Yitu and Cloudwalk behind. 

Between 2014 and 2020, SenseTime raised 11 rounds of funding, with a total amount of more than US$ 5 billion. Investors included Temasek, Softbank Vision Fund, Tiger Global, Alibaba and many other famous names. 

At its peak in 2021, SenseTime recorded an annual revenue of 4.7 billion yuan or about 660 million US dollars.  

However, the company has never made an annual profit. In 2022, SenseTime lost 6 billion yuan or roughly 850 million US dollars. 

What happened? 

Jianggan 03:06

The Grizzly Research included the following accusations: 

SenseTime falsified revenue through employment of revenue fabrication schemes;

  • In particular, SenseTime either directly or through intermediaries provides funds to customers that in turn are used to purchase goods from SenseTime that might never have been delivered;
  • Many government contracts were unprofitable (and seemingly unpaid);
  • SenseTime faces intense competition from larger and better funded peers such as Alibaba, Baidu, Huawei, Tencent, etc.;
  • To be fair, Grizzly Research has a chequered track record. In early September this year, it published a report accusing PDD Holdings – parent to Pinduoduo and Temu – of being a ‘dying fraudulent company’. PDD’s share price has thus risen almost 50%, and many informed observers, myself included, laughed while reading the report, which clearly showed lack of understanding of the company. 

In any case, we have been in discussions with some close friends in the AI sector in China over the years – and they all point to a comment problem: SenseTime has essentially evolved into a large contractor, rather than the AI software company it claims to be. 

With four business segments – Smart Business, Smart City, Smart Life and Smart Auto – SenseTime served a total of 162 customers in 2022. 

Jianggan  05:14

Facial recognition – the technology that gave SenseTime its initial ascent – has become a deep red ocean in China. Aside from SenseTime and its peers Megvii, Yitu and CloudWalk, many companies including Hikvision and Huawei are competing in the same arena. 

Because of the tightening data regulations in China, many large government and enterprise customers now require on premise deployment, further eroding the profit margins of companies such as SenseTime. 

Many of the customers do not pay on time or at all – at the end of 2022, SenseTime had a total receivable of 7.75 billion yuan or close to 1.1 billion US dollars – 65% of which is more than 1 year past due. 

We were told by a few friends that much of SenseTime’s revenue was from contracts it signed but outsourced completely to others to develop and deliver. We could not verify that – but if this is true, SenseTime’s huge and continuous losses are more understandable now. 

Jianggan  06:55

When Generative AI upended the whole AI sector in China, with companies big and small rushing into building Large Language Models and applications, SenseTime is in an even trickier situation. 

Its size and resources tied to custom developments with hundreds of customers mean that it could not easily pivot or manoeuvre. 

Considering that SenseTime was founded a year before OpenAI, it has definitely missed a lot. 

With Professor Tang’s passing, the future of SenseTime seems to be more questionable. 

The predicament of SenseTime has a lot of lessons to many. When a good startup founding team with good technology is looking to productise, capital will likely accelerate your growth trajectory, but it can also lead you astray. 

When you have not yet found a product market fit to scale, it is better not to try to meet the growth expectations of capital – scaling fast an unsustainable cost structure will eventually get to you. 

Professor Tang was well respected, the outpouring of grief from the entire ecosystem was a testament to that. May he rest in peace. 

Thanks for reading The Low Down (TLD), the blog by the team at Momentum Works. Got a different perspective or have a burning opinion to share? Let us know at [email protected].