Let machines better understand and serve humans
Founded:2016, Beijing, China

Category: Artificial Intelligence

Primary office: Beijing, China

Core technical team: Beijing, China

Status: Public, listed in the Star Market (Shanghai Stock Exchange Science and Technology Innovation Board)

Employees: 858

Amount raised:

$592 million, $200 million of private funding over three rounds, and $392 million in its first public offering in the Star Market (SHG) in July 2020. Formerly backed by venture capital, the latest round of private funding was in June 2018.

OVERVIEW

  • Cambricon Technologies Corporation designs AI chips to improve machine learning
  • Strategy to scale company value. Ride the future trend of computing by migrating from selling authorization to use intellectual property of AI chips for intelligent terminals (i.e., mobile phones) to designing, building, and selling (1) AI chips designed for cloud and edge applications and solutions, and (2) intelligent computing cluster systems
  • Offers
    • Core processor chips for intelligent cloud servers, intelligent terminals, and intelligent robots to chip design manufacturers and server manufacturers
    • Intelligent computing cluster systems to organizations with data center construction needs, especially local governments
  • Distinct competences. Innovation in (1) cloud training chips, (2) cloud inference chips, (3) edge-end artificial intelligence chips, and (4) system-level projects
  • Origin and evolution. Founded by Tianshi Chen, recognized as a genius boy, in 2016.
    • Chen and the other core technical members came from the University of Science and Technology of China and Tsinghua University
    • In 2016, the company introduced Cambricon1, its first processor for terminals. The Cambricon1 was integrated into Hauwei’s Kirin 970 processor and shipped in the Mate 10, Huawei’s flagship mobile phone
    • In 2017, an upgraded version of the Cambricon1A was integrated into Huawei’s Kirin 980 processor and shipped in more than 100 million phones. Huawei was Cambricon’s top customer in 2017 and 2018 accounting for more than 97% of the company’s revenue in both years
    • In 2019, Huawei switched from being Cambricon’s top customer to being its competitor and Cabricon started contracting to deliver intelligent computing cluster systems and selling AI chips for cloud and edge environments
    • Cambricon went public on July 20, 2020.
  • What is company known for. Cambricon is known for being a pioneer in AI chip design and integration into large systems. The company’s stock is known for being the first AI chip stock traded at the SSE Star Market

PERFORMANCE METRICS

  • Market cap: $12.46 billion (CNY ~89 billion or 400.1 million shares at CNY 222.50 per share) as of August 10, 2020
  • Revenue: $1.12 million (CNY 8 million), $16.38 (CNY 117 million), and $62.16 (CNY 444 million) in 2017, 2018, and 2019 respectively
  • Revenue from authorization to use intellectual property of terminal intelligent processors accounted for almost 100% of revenue in 2017 and 2018 and 15.49% in 2019
  • Breakdown of the CNY 444 million revenue for 2019 was: CNY 79 million (17.8%) from sales of cloud intelligent chips and accelerators, CNY 296 million (66.7%) from intelligent computing clusters, and CNY 69 million (15.5%) from authorization to use intellectual property of terminal intelligent processors
  • Has yet to generate profits

 

ACHIEVEMENTS

  • Closed series B funding round at a valuation of $2.5 billion in 2018, two years after being founded
  • Started trading as a public company on July 2020, company was worth more than $12.46 billion as of August 10, 2020

Sells

  • Authorization for using intellectual property of processors for intelligent terminals.  The processors in the M1 series support personalized deep learning and can be used in multi-channel video real-time processing, autonomous driving, and other fields. Processors in the 1H series are used in key areas of intelligent processing such as computer vision, language recognition, and natural language processing. These processors are sold to manufacturers of terminals that gather and analyze data close to the cloud or the users.
  • Cloud AI accelerator cards. Accelerator cards in the Siyuan 100 and Siyuan 270 series support multiple types of neural network applications such as vision, speech, natural language processing, and traditional machine learning. The accelerator cards in the Siyuan 270 series also integrate ample video and image codec hardware units for visual applications. These accelerator cards are sold to enterprise data centre.
  • Edge AI accelerator card. Siyuan 220 M.2 edge AI accelerator card can be used to implement AI solutions for terminal devices and edge devices. It can be used in edge computing scenarios such as smart transportation, smart grid, smart manufacturing, smart finance, and drones. Supports diversified artificial intelligence applications such as vision, voice, natural language processing, and traditional machine learning, and realizes edge-end intelligent solutions for various services.
  • Intelligent computing cluster systems.  For clients with AI computing construction capabilities, Cambricon integrates the cloud intelligent chip accelerator cards into the existing computing cluster. The offer consists of three parts, (1) cloud intelligent chip accelerators (MLU 270, 100), (2) the basic system software platform (i.e., Cambricon Neuware), and (3) an intelligent computing cluster management system.

Channels

  • Direct sales model. Company has a dedicated sales team that engages with customers
  • Partnerships. Company has established strategic cooperative relationships with well-known companies in various industries in China, government organizations, and venture capital providers
  • Cambricon Developer Forum. Supports developers of AI applications, devices, and solutions. Provides (1) access to an open source deep-learning framework integrated with a device-cloud integrated platform; (2) a machine learning programming library and a new programming language; (3) drives and system tools; (4) courses in the theory of intelligent systems; (5) data package and an environment to experiment as well as Q&A; (6) forum where knowledge can be exchanged; (7) access to resources; (8) announcements; (9) industry news; and (10) posts that help understand the forefront of knowledge
  • Trade shows. Makes product announcements that attract attention such as the 21st High-Tech Fair held in Shenzhen in November 2019

Competencies

    1. R&D. Research and development of AI chips for cloud, edge, and terminal application scenarios. It is one of the few companies in the world that can provide cloud-edge integration, software and hardware collaboration, training and inference integration, a series of intelligent chip products and platform-based basic system software with a unified ecology. Technology
    2. Innovation ecosystem. Building an ecosystem anchored on a framework to design AI programs and applications

Distinct AI Features

Type

  • Fixed-point neural network operations and other

  AI use

  • Advances field of AI chip design. Carries out leading-edge R&D and advances and delivers a course on Intelligent Computing System theory
  • Enables resource integration between cloud and edge computing service providers and developers of AI applications and solutions. Maintains and continuously advances a machine learning programming library and a programming language (i.e., Bang) to develop efficient, cross-platform artificial intelligence applications; provides drives and system tools; and maintain a developers’ forum

Rate of return on customer’s investment to make AI work

Immediate:

  • Ultra-high energy efficiency for artificial intelligence systems to meet cloud requirements for servers and data centers (i.e., process a large amount of raw data, which has high requirements for computing power and data storage of basic hardware)
  • Integrate cloud intelligent chip accelerator cards into the existing computing cluster

Mid-term:

  • Cloud environment can easily be migrated from a service provider premises to the enterprise premises

Long term:

  • Have an important role of China’s semiconductor industry by reducing China’s dependencies on foreign AI chip designers, and enable direct competition with large companies worldwide

Databases

  • Unique databases used in cloud and edge environments complement the AI chips

Quantum Computing

  • N/A

Edge computing applications

  • AI acceleration cards can effectively make up for the disadvantage of the insufficient computing power of terminal devices and alleviate potential problems such as data security, privacy protection, bandwidth, and delay in cloud computing scenarios.

Resources

Assets

  • Framework to integrate cloud and edge AI programs and applications
  • Relationships with leading technology universities in China
  • Program to recruit talent from lead technology universities in China
  • Suppliers’ arrangements. Cambricon has arrangements with suppliers such as ARM, Synopsys and Cadence for intellectual property and electronic design automation authorization

Processes

  • AI chip design
  • AI chip integration into cluster systems

Priorities

  • Successfully enter the cloud market
  • Attain a leadership position in the AI chip design market in China and then worldwide

References

  1. Cambricon Technologies Corporation Limited. 2020. Prospectus. http://static.sse.com.cn/stock/information/c/202006/2c43af7f9bb946f1a874bc0ebbca88e2.pdf
  2. Cambricon Technologies Corporation Limited. 2020. Website http://www.cambricon.com/
  3. cnTechPost 2020. IPO prospectus unveils Cambricon’s mystery. March 27. https://cntechpost.com/2020/03/27/ipo-prospectus-unveils-cambricons-mystery/
  4. Crunchbase. 2020. Cambricon. https://www.crunchbase.com/organization/cambricon-technologies
  5. Lu, D. 2020. Cambricon: Staggering Growth, IPO and Post-Huawei Uncertainty. June 28. https://equalocean.com/analysis/2020062814151
  6. Lu, D. 2020. CSRC Greenlights Cambricon’s Star Market IPO. June 23. https://equalocean.com/news/2020062314137
  7. Platanov, I. 2020. Cambricon Launches Star Market IPO – Why We Are Bearish on It. July 20. https://equalocean.com/analysis/2020072014295
  8. Platanov, I. 2020. Chinese AI Chip Unicorn Cambricon Goes Public on Shanghai Tech Board. March 26. https://equalocean.com/news/2020032613828

Contributors

  • Tony Bailetti