Skip to content

China's Tech Industry: From Follower to Frontier

Abstract

In 1978, China had no commercial computing industry. It did have a computing tradition: a state institute founded in 1956, a first machine copied from Soviet blueprints in 1958, a first independent design in 1960, and a line of classified mainframes that did the arithmetic for the bomb. Its universities had been shuttered for a decade by the Cultural Revolution, its scientists scattered or silenced, its access to foreign technology restricted by Cold War export controls. By 2020, China produced more computer science research than any other country, hosted two of the world’s ten most valuable technology companies (Alibaba and Tencent), and controlled critical links in the global semiconductor supply chain. It got there through state strategy, mass engineering education, a protected domestic market of a billion people, and the absorption of foreign technology by licensing, joint ventures and, according to US prosecutors, theft; since 2018 American export controls have tried to slow it down.

Xia Peisu 1946
Xia Peisu, designer of the Model 107, photographed in 1946 at the Chongqing campus of National Chiao Tung University. Image: public domain, via Wikimedia Commons.

The Institute Years, 1956–1983

Punched cards reached China in 1932, when IBM machines went to Nanjing, probably for medical statistics; their fate after the Japanese capture of the city in December 1937 is unknown. Chinese computing proper began as a proposal in a mathematics institute. Hua Luogeng first proposed building a Chinese electronic computer in 1952, and in early 1953 a research team formed under him at the Academy of Sciences’ Institute of Mathematics. The State Council approved a preparatory committee for an Institute of Computing Technology (ICT) of the Chinese Academy of Sciences on 25 August 1956, the country’s first body devoted to computer research. Hua had recruited Xia Peisu, an electrical engineer with a 1950 Edinburgh doctorate, in 1952 to lead the computer work. Xia taught China’s first course in computer theory in 1956 and wrote its first systematic textbook. Three Soviet analog computers shown in Beijing in 1954 were the first computers in the country. The institute moved to Zhongguancun in Beijing in February 1958, where it still is.

The first machines were Soviet. In September 1957 the institute obtained the blueprints of the Soviet M-3, and on 1 August 1958 the copy, the Model 103, ran a four-instruction program (the Heinz Nixdorf MuseumsForum dates its completion to September 1958; 1 August is also Army Day, the source of its first name): about 30 operations a second on a magnetic drum, 1,500 to 2,000 once core memory was fitted. Built with State Factory 738, the Beijing Wired Electric Factory, it was first called the “August First” machine and later the DJS-1; about 49 were made between 1958 and 1966. The Model 104, a copy of the BESM-II with 4,200 tubes and a 39-bit floating-point word (the Heinz Nixdorf MuseumsForum gives about 5,000 tubes), was debugged in April 1959 and delivered that October, and ran China’s first home-written compiler in 1961. It did 10,000 operations a second, and the country’s nuclear and weather programmes queued for it.

The Soviet advisers were recalled in 1960, when the Sino-Soviet split ended technical aid, and the institute was on its own. In April 1960 Xia Peisu’s Model 107 became operational, the first general-purpose computer designed in China, which is why the China Computer Federation gave her its first lifetime achievement award in 2010 and Loongson’s architect Hu Weiwu, her student, named a processor after her. The Model 119 of 1964, the institute’s first self-developed large computer, played a part in the calculations for China’s first nuclear test in October 1964. The institute’s transistorised Model 109B followed in 1965, and Ci Yungui’s 441-B series gave China its first transistorised general-purpose computers in the 1960s. The Cultural Revolution then closed the universities for a decade; the institute kept working, and in 1972 delivered the 48-bit Model 111, the first computer built entirely from Chinese-made integrated circuits, followed by the 013 with 768K of core memory in 1976.

The line ended with two supercomputers. The 757 vector machine, certified in 1983, is the subject of the Dead End section below. The other, Yinhe-I (“Galaxy”), was built by Ci Yungui’s team at the National University of Defense Technology in Changsha and completed in 1983 at 100 million operations a second, the first Chinese machine in that class. Ci had been elected to the Academy in 1980; he died in Beijing on 21 July 1990. By then the institute’s commercial spin-offs had begun: Legend, later Lenovo, in 1984, then Dawning, Loongson and Cambricon. In 1978 China had no computer industry, but it had thirty years of designers who knew how to build a computer without buying one.

The Foundation: Reform and the Return of Science

Deng Xiaoping’s Reform and Opening Up policy (1978) restarted Chinese higher education after the Cultural Revolution’s decade of destruction. Universities reopened, the gaokao university entrance exam was restored, and science and engineering were declared the path to national modernization. The State Council’s 1986 863 Program (named for its launch date: March 1986) targeted seven high-technology areas, including computers and automation, with directed funding designed to close the gap with the West.

The early domestic computing industry was defined by state-owned enterprises making IBM-compatible clones under license or reverse-engineered. Lenovo (then Legend, founded 1984 by Liu Chuanzhi at the Institute of Computing Technology) was an exception, a hybrid of state institute spinout and commercial enterprise that would eventually acquire IBM’s PC division in 2005, becoming the world’s largest PC maker.

The home market took a different route. China’s first eight-bit microcomputer, the DJS-050, appeared in 1977 and the IBM-compatible Great Wall 0520CH in 1985, but the machine most Chinese families met first was a copy of Nintendo’s Famicom. In 1989 the electronics firm Nihwa in Zhongshan, under its new manager Duan Yongping, began selling a Famicom clone as Xiaobawang (“Little Overlord”), far cheaper than the import, and renamed itself after it in 1991. From 1993 it added a keyboard and sold the clone as the Subor “learning computer”, with typing tutors, word processors and BASIC on a cartridge, advertised by Jackie Chan; the SB-486D of 1994 shipped with Dr. Mario until Nintendo objected and it was swapped for Chinese chess. Duan left in 1995 to found BBK, which made him one of the richest men in China, and a state ban on game consoles from 2000 to 2015 ended the line while leaving PC and online games untouched.

Huawei: The Infrastructure Champion

Huawei was founded in 1987 by Ren Zhengfei, a former People’s Liberation Army engineer, in Shenzhen with 21,000 yuan of capital. Its early business was reselling imported telephone switches. By the late 1990s it was developing its own switching equipment. By the 2000s it was underbidding Ericsson, Nokia, and Cisco for telecom contracts in developing countries, trading lower margins for market share.

Huawei’s rise was built on engineering capability, aggressive pricing, and (according to a 2020 US federal indictment) systematic appropriation of competitors’ technology through employee hiring, joint ventures, and theft. The company has denied the theft allegations. Whatever the mechanism, it became the world’s largest telecom equipment maker and, through its HiSilicon subsidiary, one of the few companies capable of designing its own smartphone chips at the leading process node.

The US placed Huawei on the Commerce Department’s Entity List in May 2019, and in 2020 extended the foreign direct product rule so that any chip made with American tools or design software, which meant every advanced foundry including TSMC, needed a US licence before it could be sold to Huawei. TSMC stopped shipping to Huawei in September 2020, and HiSilicon’s Kirin processors, which had rivalled Apple’s A-series chips, disappeared from new products. They came back in September 2023: the Mate 60 Pro shipped with a Kirin 9000S that TechInsights teardowns identified as SMIC’s second-generation 7nm (N+2) process, restoring 5G to Huawei’s flagship without any access to EUV lithography.

The Internet Giants: BAT and Beyond

China’s connection to the world’s networks began with an email. On September 20, 1987, the Institute for Computer Applications in Beijing sent the country’s first email over a CSNET link to the University of Karlsruhe, set up with Werner Zorn’s German team (see Germany’s IT Industry). Its text: “Across the Great Wall we can reach every corner in the world.” The phrase became famous in China; the permanent TCP/IP connection followed only on April 20, 1994.

China’s internet industry consolidated around three companies that became known as BAT: Baidu, Alibaba, and Tencent.

Baidu (Robin Li, 2000) dominated Chinese search through PageRank-equivalent algorithms adapted for Chinese-language content and, critically, through Google’s departure from mainland China in 2010 following disputes over censorship and state-sponsored hacking of Gmail accounts. Baidu became China’s default search engine and spent its advertising revenue on autonomous driving (the Apollo project, 2017) and AI.

Alibaba (Jack Ma, 1999) began as a B2B trade platform connecting Chinese manufacturers with foreign buyers. The pivot to consumer e-commerce through Taobao (2003), launched in direct competition with eBay’s entry into China, created the foundation of Chinese retail commerce. Alipay (2004), initially an escrow service to make up for the lack of consumer credit cards and trust between strangers, became one of the world’s largest mobile payment platforms, with about a billion users by 2020. Alibaba’s 2014 IPO on the New York Stock Exchange raised $25 billion, then the largest IPO in history.

Tencent (Ma Huateng, 1998) built its empire on messaging: OICQ (later QQ), which cloned ICQ’s model in 1999, followed by WeChat (2011), which evolved from messaging into a super-app integrating payments, social media, government services, food delivery, and ride-hailing within a single application. WeChat’s mini-programs made it a platform inside the phone’s own operating system, a model Western companies such as Facebook tried to copy with little success. Tencent also became the world’s largest video game company by revenue through investments and acquisitions: League of Legends publisher Riot Games (majority stake 2011, full ownership 2015), Clash of Clans publisher Supercell (2016), and a 40% stake in Epic Games (2012).

ByteDance (Zhang Yiming, 2012) broke into the BAT market with recommendation algorithms first and content second: Toutiao (news aggregation) and Douyin/TikTok (short video, 2016/2017). TikTok was one of the first Chinese consumer products to compete globally with American platforms, and it reached a billion monthly active users in September 2021.

The Great Firewall: Architecture and Industrial Policy

The Golden Shield Project (colloquially the “Great Firewall”) operates through several technical mechanisms working in combination. At the network layer, China’s internet is connected to the global internet through a small number of international exchange points controlled by the three state-owned telecommunications carriers (China Telecom, China Unicom, China Mobile). All traffic to and from foreign destinations passes through these choke points. The filtering uses deep packet inspection (DPI) to identify and block traffic based on destination IP, DNS queries, and URL patterns, not just domain names but specific URLs and query strings. SNI (Server Name Indication) in TLS connections is inspected to block HTTPS traffic to blocked domains. The system also uses DNS poisoning to return false IP addresses for blocked domains, forcing users toward domestic alternatives even when circumvention is not attempted. VPNs and Tor are blocked through traffic pattern analysis: encrypted connections to known VPN endpoints are identified and disrupted, though technically sophisticated users can circumvent these measures through obfuscated protocols. Enforcement varies with the political calendar. The stated rationale is national security and social stability. The economic effect is a domestic market protected from foreign competition, allowing Chinese companies to scale to hundreds of millions of users before facing international rivals. How much of Chinese tech success the protection explains is disputed, but it let Chinese companies grow on a huge home market before they met foreign competitors.

The Semiconductor Ambition

China consumes more semiconductors than any other country (around half of world chip sales go to China, much of it for devices assembled there and exported), and makes only a small fraction of them domestically. Reducing this dependence became a national priority after the ZTE sanctions of 2018 (the US temporarily banned ZTE from purchasing US components, nearly destroying the company) and the Huawei restrictions of 2019.

SMIC (Semiconductor Manufacturing International Corporation, founded 2000) is China’s leading foundry. As of 2024 it can manufacture at roughly the 7nm node, several years behind TSMC. The gap is structural: the most advanced chips need EUV lithography machines, made only by ASML in the Netherlands and withheld from China under Dutch and US export controls, so SMIC reaches 7nm with older deep-UV machines and multiple patterning, at lower yield and higher cost.

The Chinese government’s response has been the “Made in China 2025” industrial policy (2015), directing hundreds of billions of yuan into semiconductor R&D, and three rounds of the state-backed National Integrated Circuit Industry Investment Fund (the “Big Fund”, 2014, 2019 and 2024), together worth close to $100 billion. The Big Fund has invested in SMIC, Yangtze Memory Technologies (NAND flash), and dozens of fabless chip designers. YMTC surprised the industry in 2022 when its 232-layer 3D NAND flash was found shipping in a retail SSD (the first 200-plus-layer NAND seen in production, briefly ahead of Samsung, Micron, and SK Hynix). In December 2022 the US added YMTC to its Entity List.

The Decoupling

The 2018–2024 period saw accelerating US-China technology decoupling. The Trump administration added Huawei, SMIC, and dozens of other Chinese tech companies to export restriction lists. The Biden administration’s October 2022 advanced chip export controls went further: banning not just sales of advanced chips to China but sales of chip-making equipment and (controversially) restricting US citizens and permanent residents from working in China’s advanced semiconductor industry. The CHIPS and Science Act (2022) added $52 billion in US domestic semiconductor subsidies.

China’s response has been accelerated domestic substitution across the technology stack, from operating systems (replacing Windows in government systems with Linux-based Kylin OS) to database software (replacing Oracle with domestic alternatives) to chips.

Dead End: The Siloed Supercomputer

An illustration of China’s early state-directed computing trajectory is the Model 757, the country’s first large-scale vector computer, developed at the CAS Institute of Computing Technology and granted state certification in November 1983. The machine ran vector operations at 10 million per second and scalar operations at 2.8 million, and it was a genuine engineering achievement, taking the Chinese Academy of Sciences’ special prize in 1983 and the first-class National Science and Technology Progress Award in 1985. But it had been commissioned by the State Council and Central Military Commission in 1975 for the Ninth Institute, China’s nuclear-weapons design establishment, and that is where it went. Its classified, mission-specific role meant its results circulated narrowly, its engineers could not collaborate internationally, and its architecture could not be improved through the open feedback loops that made Western computing innovation cumulative. It had capable engineers and adequate resources, but almost no knowledge transfer, the reverse of Silicon Valley’s model. The Reform era opened Chinese science to international publication and collaboration, imperfectly and under political constraints.

📚 Sources