Portfolio: China Everbright and Terminus

For Terminus, a Chinese internet-of-things (IoT) platform, the journey from start-up to unicorn status has taken less than four years. The company crossed the $1 billion threshold in August on closing the first phase of a Series C round worth RMB2 billion. Its investor roster features some of China’s top technology companies, including artificial intelligence (AI) specialists SenseTime and iFlytek, as well as private equity players IDG Capital and CITIC Private Equity. 

However, when Victor Ai, a managing director at China Everbright invested in Terminus in 2015, its product range was limited to smart door locks. Smart home technology was among the hottest themes that year, following Google’s $3.2 billion acquisition of Nest, a manufacturer of products ranging from smart speakers to smoke detectors, less than 12 months earlier. But Terminus’ copycat strategy wasn’t working: Intellectual property protections in China were not as strict as in the US and the consumer market was proving difficult to penetrate.

“We don’t talk about them today, but in 2014 and 2015 the top three venture capital firms in China each invested in more than 10 companies of the smart home technology category,” Ai tells AVCJ. “China’s 2C market was – and remains – saturated. There are so many apps chasing after you and offering free gifts, yet most of them never get traction. You might only use Didi Chuxing for ride-hailing and QQ Music for pop music.”

Ai wasn’t prepared to accept failure with Terminus and started work on a reboot. “I am more private equity-style than venture capital-style,” he observes. “If something is wrong, I want to make it right.”

To some extent, Terminus was like the first large project Ai led at China Everbright in 2009. As one of the founding partners of EBA investments – a cross-border real estate asset manager that now oversees assets worth more than RMB105 billion ($15 billion) – Ai and his team invested in an unfinished building in the city center of Chongqing, transforming it into a high-end shopping mall under the IMIX Park brand. It listed as a real estate investment trust (REIT) and helped establish EBA’s reputation.

Ever willing to get his hands dirty, Ai started out in the same fashion, dismissing Terminus’ entire founding team. He then brought in new talent from companies like Huawei Technologies and redirected the business to serve enterprises rather than individual consumers. The strategy was christened AIoT – AI plus IoT. All that remained of the original Terminus was the name, which is hardly a selling point. The Chinese name was deliberately chosen for its similarity to Tesla

“At that time, everyone wanted to imitate Elon Musk, that explains the origin of the name,” Ai explains. “I kept it. My goal was that when people talked about Terminus, they wouldn’t think of Tesla.”

Dual identity

Ai took on the CEO role and retains it to this day. It is an unusual arrangement, given he also leads China Everbright’s new economy-focused private equity division, which has $4 billion in assets under management. When he assumed control of Terminus, Ai’s main role was launching a new economy investment vehicle with colleagues who previously worked alongside him in real estate. 

He admits being drawn to Terminus, which in itself is intended to serve as an investment platform in the AIoT space as well as a service provider. “Who have been the most successful investors in the past 10 years in China? Tencent and Alibaba, not any VC or PE firms,” Ai says. “Why is this? Because the past decade was the mobile internet era. You needed to be an influential company within this industry to find the best deals ahead of other investors and generate the best returns.”

In this way, Terminus is being positioned as Everbright’s Tencent or Alibaba for the IoT era. The company will build and invest in its own ecosystem, taking advantage of synergies between its investees and using them as a catalyst for its own business.

There is already some cross-pollination with other parts of China Everbright’s new economy portfolio, and indeed with IMIX Park. The new economy fund has invested in 100 projects, including 20 unicorns such as SenseTime, online-to-offline services platform Meituan-Dianping, and video streaming service iQiyi. By building relationships with these companies, China Everbright understands their needs and looks to sign them up as customers or partners for Terminus. Meanwhile, IMIX Park uses Terminus’ smart building solutions to monitor and control electric devices, cutting 20% of its electricity bills.

What the unicorns and IMIX have in common is scale – which makes them typical Terminus clients. Four in every five orders the company receives are above RMB100 million. They come from the likes of China National Petroleum Corporation, Sinopec, China Mobile, China Unicom, Wanda Group, Greenland Holdings, and China Resources Group.

They account for a large part of the demand base that has turned China into the world’s largest IoT market. The country has 960 million devices connected via cellular networks, accounting for 64% of global cellular connections, according to the Global System for Mobile Communications Alliance (GSMA), a trade organization representing mobile operators. The market value is expected to be worth RMB373 billion this year, up from RMB299 billion in 2018.

Practical application

AIoT-based systems absorb information generated in offline scenarios – such as supermarkets – through sensors and cameras. The data are analyzed and run through algorithms that use them to direct future actions. In this way, supermarkets can learn as much about customers’ shopping behavior as e-commerce platforms. 

“I’m absolutely convinced that AIoT will become the backbone of the entire retail industry, replacing the very old ERP [enterprise resource planning] systems,” says Yong Du, founder of T11, a supermarket in Beijing that specializes in fresh produce. It is a China Everbright portfolio company and a Terminus customer. 

There are two different approaches to storing and processing data captured offline: cloud computing, whereby information is transferred to a cloud-based centralized processing warehouse; and edge computing, a distributed model that involves processing near the edge of the network – where the data are generated – or in terminal devices. Terminus focuses on the latter because cloud computing requires heavy investment and it is already dominated by the likes of Alibaba Cloud, Qiniu Cloud and QingCloud. 

“Edge computing resolves three major problems: security, bandwidth and real-time feedback,” Feng Liu, a vice president at Terminus who previously led the smart city products line at telecom equipment and systems manufacturer ZTE Corporation, tells AVCJ. “It is complementary to cloud computing.” 

The security problem edge computing resolves concerns the reluctance of many enterprises to transfer data to cloud computing warehouses for fear their privacy will be violated. As for bandwidth and feedback, eliminating the need to transfer video-based data helps reduce bandwidth costs and cuts down on the lag time between capture and feedback. For example, if facial recognition functions are performed locally, visitors can pass through immediately instead of waiting several seconds for processing.

“When firefighters or policemen wear visual equipment or other sensors, the data must be processed on the terminal; autonomous driving cannot use cloud computing either,” Liu explains. “In general, the two approaches work synergistically – cloud computing is used for big data analysis and for the design of algorithm models that can be used in edge computing solutions.”

Terminus’ core product is the Poseidon series: hardware with built-in algorithms, including accelerators, servers and edge-computing units. They serve as the brains behind front-end devices such as cameras or sensors, enabling them to aggregate and process data locally. In T11’s store, Poseidon recognizes customers and categorizes them in two ways: as visitors, frequent customers and VIPs; and as promotional, hesitant and decisive. Every purchase generates dozens of data sets, which are used to increase sales efficiency per square meter.

“In the past, we relied on the store manager’s experience to arrange products on shelves,” one store manager explains. “Now, we use data analysis to decide where and how to place products, ” Jessica Sun, a vice president at Terminus, notes that retail is a low-profit industry and industry participants are sensitive to the cost of data mining. One of the advantages of Poseidon is that it can be installed as an upgrade in existing surveillance cameras.

Community counts

Terminus has 650 staff, 65% of whom are technicians, and five research centers. With established expertise in edge computing, the company has expanded into multiple segments, such as community security, building energy management and fire alarm management. It claims 8,400 projects nationwide, covering more than 10 million people with 420,000 connected data nodes across 673 million square meters. Sales have doubled every year since its inception.

“In the internet era, it’s winner takes all, it’s a horizontal aggregation. This is the nature of internet – you have to be connected,” says Ai. “In the IoT era, it’s a vertical aggregation. No technology company can survive relying on one or two core products, they must expand into different areas.”

This is already apparent in the strategic evolution of China’s leading AI companies. SenseTime started with facial recognition algorithms and has since developed applications across eight industries with dozens of product lines, ranging from healthcare platform to driving assistance. Megvii Technology’s launch product was cloud-based computer vision open platform Face++ but it now primarily provides identity authentication solutions for personal devices, smart cities and supply chains

The companies also integrate each other’s technologies to offer a comprehensive solution. “This is a bit like making a mobile phone. What is the most important thing? Product definition. iPhone users don’t care who makes the lens or who offers the facial recognition solution. If the final product definition is done by iPhone, the product is an iPhone. We are doing something similar. We integrate best technology in the market to serve different scenarios,” says Ai.

Of the segments in which Terminus has a presence, smart community services is the largest revenue-generator. The company ranks number one in this area by national market share. Typical solutions involve camera-based entry systems for apartment complexes, with Poseidon responsible for authentication. Edge computing is most effective in relatively small spaces, for example, a supermarket of 5,000 square meters; in large-scale scenarios – such as airports and train stations – cloud-based systems can be better.

The company’s latest offering is a security surveillance robot called Titan, which was launched earlier this month in Beijing. The robot can store 300,000 faces, equivalent in size to China’s entire library of most wanted fugitives. If a suspect appears, the robot can instantly inform local police. “If the data is on the cloud, it takes several seconds to give feedback, but with our robot, it’s a millisecond-level response,” Sun says. The robot can also patrol power stations, monitoring temperatures of high voltage circuits and substations. It can sense temperature changes and initiate alarms in real-time.

Smart trash cans are another new development – and a timely one given China’s ecological protection drive. The move to ban landfill imports earlier this year has been accompanied by a big push on domestic recycling, with companies like Tencent Holdings developing apps that help residents sort trash.

Residents operate Terminus’ smart trash cans, now available in Shanghai and Wuzhen, by scanning a code on the side. The trash is compressed and weighed, and users collect green points based on the weight. These can be redeemed for gifts. Garbage data is collected in real-time and fed into a unified community management platform that aggregates information on access control, public services, electricity supply controls and fire protection. An updated version, scheduled for launch by the end of the year, will be able to automatically distinguish wet and dry garbage.   

• Tweet  
• Facebook  
• LinkedIn  
• Google plus  
• Save this article  
• Send to  

• Topics
• Greater China
• Technology
• Early-stage
• China
• China Everbright
• artificial intelligence