Cambricon Brand Guidelines: The Beijing AI Chip Maker Named for the Cambrian Explosion

A laboratory idea named after an explosion of life.

Cambricon — 寒武纪, Hánwǔjì, in Chinese — designs processors for artificial intelligence and is based in Beijing. Its name comes from the Cambrian explosion, the burst of new life forms in the fossil record: Chinese Wikipedia says the name expresses the hope that AI will bring an era-changing shift, and English Wikipedia says the name was first given to a research project.

That project began at the Chinese Academy of Sciences. English Wikipedia says the brothers Chen Yunji and Chen Tianshi, who both earned computer-science doctorates at the University of Science and Technology of China, worked on computing technology at the academy, and that the Cambricon project was formed there in 2008 to develop a processor specialised for deep learning. Chinese Wikipedia adds that from 2014 the brothers’ group at the academy’s Institute of Computing Technology proposed an instruction set for deep-learning processors, later presented at the ISCA computer-architecture conference in 2016.

The company followed in 2016. Cambricon’s own website says it was founded that year to develop AI chips, and English Wikipedia dates the spin-out of the project as a business to March 2016, with the brothers as co-founders; it says Chen Yunji soon returned to academic work, while Chinese Wikipedia names Chen Tianshi as the founder. The same year, Chinese Wikipedia says, the company released the Cambricon-1A, which it calls the world’s first commercial processor dedicated to deep learning, and which was chosen among fifteen “world-leading” technology achievements at the third World Internet Conference in Wuzhen. Chinese Wikipedia also says that in 2016 Cambricon signed a strategic cooperation agreement with the Chinese computer company Sugon (中科曙光).

From phone processors to data-centre cards.

Cambricon’s first large customer was a phone maker. Chinese Wikipedia says that from 2017 the Kirin 970 and Kirin 980 processors designed by Huawei’s HiSilicon used Cambricon’s AI unit. The relationship did not last: English Wikipedia says Huawei ended the contract in 2018 in favour of HiSilicon’s own AI designs, while Chinese Wikipedia says that by 2024 Cambricon was no longer in Huawei’s supply chain. The trade magazine EE Times included Cambricon in its Silicon 60 list of start-ups to watch in 2017 and 2018, Chinese Wikipedia says.

The company then turned to servers and edge devices. Its website describes a range that spans cloud, edge and device — “云边端一体” in its own phrase — with training and inference combined. Accelerator cards are sold under the Siyuan (思元) name and MLU model numbers: the Siyuan 370 series, including the MLU370-S4, X4 and X8 cards; the earlier Siyuan 270 series; and the Siyuan 220 series of edge modules. The company says the Siyuan 370 is its third-generation cloud AI chip, built with chiplet technology on its MLUarch03 architecture, and that it announced the MLU370-X8 training card, which joins two Siyuan 370 chips over its MLU-Link connection, in March 2022.

Alongside the cards it lists processor designs licensed to device makers, the Cambricon-1M and Cambricon-1H, and its software platform, NeuWare, with a tool called MagicMind. Its website lists offices in Beijing, Shanghai, Shenzhen, Hefei, Xi’an, Kunshan and Nanjing.

Investors arrived early. CGTN, in July 2020, named Alibaba, iFlytek, Lenovo Ventures and the academy’s own venture-capital arm among them.

How the category presents itself.

AI chip companies sell something most of their customers will never see. The processor sits inside a server, the server sits in a rack, and the rack sits in a data hall far from the people who use the models running on it. So the industry’s visual language has grown up around making that hidden hardware visible: a whole silicon wafer, as wide as a dinner plate, held up on a keynote stage; a single chip pinched between finger and thumb for the cameras; a die shot recoloured in neon false colour so that its blocks of memory and compute read like a city map at night.

The second set of images is the data centre. Press photographs tend to show long rows of racks in dark halls, lit by the glow of status lights and cold aisles, with cabling arranged as neatly as a diagram. These pictures stand in for scale, which is what buyers of AI compute are being asked to imagine, and they appear again and again in launch announcements, partner news and investor material.

Keynote slides do much of the remaining work. Model names, token counts and benchmark charts appear beside the chip’s own product name, and the companies’ identities are usually kept spare to let those numbers and images lead: black or white grounds, a single accent colour, a sans-serif wordmark and very little ornament. Several firms have adopted coined product categories, names for their processors that set them apart from the graphics processor.

The shop window is the technical conference. Hot Chips, NeurIPS and the AI Infra Summit are among the events where architectures are presented, demonstrations run and booths compete for engineers’ attention. This guide looks at how each company in the group presents itself, and keeps to what can be checked about who founded it, who owns it and what it makes.

A listed company on the STAR Market.

On 20 July 2020 Cambricon’s shares began trading on the STAR Market of the Shanghai Stock Exchange, which CGTN called “Nasdaq-style.” The broadcaster reported that day that the shares closed well above their offer price, and quoted a fund manager who called Cambricon the first listed AI chip maker in China’s A-share market. The listing came four days after that of the chip manufacturer SMIC on the same board, the broadcaster noted, and less than five years after the company was founded.

In December 2022 the company was named in an American export-control rule. A rule from the US Department of Commerce’s Bureau of Industry and Security, effective 16 December 2022 and published in the Federal Register on 19 December, added thirty-six entities to the Entity List, among them Cambricon Technologies Corporation Limited, several Cambricon subsidiaries and the Chinese Academy of Sciences’ Institute of Computing Technology. The rule gave the agency’s reason as “acquiring and attempting to acquire U.S.-origin items in support of China’s military modernization,” and set a license requirement for all items subject to US export rules, with a policy of presumption of denial. Chinese Wikipedia says the listing limited the company’s access to advanced manufacturing processes.

Its website states its vision as helping machines “better understand and serve humans,” and English Wikipedia notes that it has often been compared with Nvidia because of its similar focus.

For a design guide, Cambricon shows how a scientific name — a geological era — can carry a chip brand from laboratory to stock market. Compare Nvidia, the company it is most often compared with, and Arm, which also licenses processor designs. For this guide we use a deep slate, an ocean blue and a fossil ochre with a sans, an unofficial interpretation; no product or logo is reproduced.

Keep Cambricon’s identity on hand — save it to a board in the Made Good inspiration library.