How China is funding its AI ambitions without US capital

(Edited and refined by Grace Chong, with the assistance of AI translation.)

In 2026, ChangXin Memory Technologies (CXMT), China’s leading domestic memory and artificial intelligence (AI) computing chip company, successfully entered the capital markets, setting a new record for the market capitalisation of a listed company in China’s semiconductor and AI hardware sectors. This marks one of the most symbolic events in the recent capitalisation of China’s hard technology (hard tech) sector.

The rise and listing of CXMT was the result of a fully domestic capital system. Its core shareholders include the National Integrated Circuit Industry Investment Fund (also known as the “Big Fund”), industrial funds established by the Hefei local government, central state-owned enterprise (SOE) investment platforms, and private capital. Throughout its development, the company operated without the support of US venture capital firms or major overseas investors. 

Pre-2016: when US capital still ruled 

The fact that China’s highest-valued listed company was incubated by a domestic capital system is far more than an individual company’s listing success. It represents a milestone in the changing relationship between financial capital and technology industries amid the broader China-US competition in AI and semiconductor technology. 

From this point onwards, China’s AI industry is entering a mature stage, transitioning from a model dependent on overseas capital funding to one supported by a domestic, multi-tier capital system. The pace of industrial iteration and technological breakthroughs is expected to accelerate significantly, gradually narrowing the gap with the US and potentially surpassing it in certain AI sectors. 

Looking back at the decade-long technology competition cycle between China and the US from 2016 until now, the capital landscape supporting the two countries’ AI and semiconductor industries has undergone a complete reversal. 

Before 2016, the global AI and semiconductor industrial chains were characterised by a clear pattern of “US capital dominance and Chinese enterprise participation”. Chinese domestic hard tech companies relied heavily on funding from US venture capital and private equity firms. Domestic industrial funds were still in their early stages, and China lagged far behind the US in terms of capital scale, professional expertise and incubation capabilities.

Data shows that before 2016, overseas capital accounted for more than 65% of investment in Chinese semiconductor and AI startups. Nearly all leading AI algorithm companies and advanced chip design firms had investment from top US funds such as Sequoia Capital, Andreessen Horowitz (a16z) and Khosla Ventures. 

At the time, US technology capital, supported by its enormous financial resources, mature exit mechanisms and extensive industry networks, firmly controlled the pace of investment and direction of technological iteration in global AI hard tech.

2016: the turning point 

The year 2016 marked a watershed moment in the competition between Chinese and US technology capital. As global technology rivalry intensified and efforts to de-risk industrial chains accelerated, Chinese and American industrial chains began to undergo structural decoupling. Under political pressure and policy intervention, US capital gradually withdrew from China’s core AI and semiconductor sectors, while overseas investment in China’s hard tech industries declined sharply.

Statistics show that US annual investment in China’s AI and semiconductor sectors peaked at US$32 billion in 2016, accounting for 72% of China’s total domestic hard tech investment. After 2021, major US technology venture capital firms largely stopped investing in China’s core hard tech projects. Between 2024 and 2026, annual related investment in China averaged less than US$1.8 billion, representing a decline of more than 94% over the decade.

As US capital continued to withdraw, the Chinese government began building a four-level domestic capital support system consisting of central government funds, local government industrial funds, central SOE and SOE funds, and market-oriented private capital. This system gradually began filling the investment gap left by overseas capital.

Building a domestic capital foundation

At the national level, the three phases of China’s National Integrated Circuit Industry Investment Fund have an accumulated registered capital of 686.9 billion RMB (approximately US$101 billion). The third phase alone accounts for 344 billion RMB, exceeding the combined size of the first and second phases. Around 70% of these funds have been directed towards critical bottleneck areas, including AI computing chips, semiconductor equipment, and materials, thereby forming a core foundation of the industry.

At the local government level, chip- and AI-focused funds established by provinces and cities with hard tech clusters have accumulated more than 500 billion RMB. Through targeted industrial investment, cities including Hefei, Shanghai, Chengdu and Wuxi have incubated a number of industry leaders, including CXMT, Hygon Information Technology, and SJ Semiconductor. 

Together with innovation investment funds worth hundreds of billions of renminbi established by central SOEs and state-owned financial platforms, total state-owned investment in China’s hard tech sector has exceeded 1.3 trillion RMB.

When this investment system was first established, many overseas commentators criticised it for relying too heavily on government funding, arguing that it would fail to mobilise private capital and would be unsustainable over the long term.

Between 2016 and 2026, Chinese private capital invested a cumulative 520 billion RMB in semiconductor and AI sectors. These investments generated 128.6 billion RMB in realised cash-outs and 689 billion RMB in unrealised gains, producing an overall cash realisation rate of 15.7%.

During the same period, US AI and semiconductor venture capital achieved an overall cash realisation rate of 29.2%, including 37% in AI software and 27% in semiconductor hardware.

This difference in performance has contributed to the negative views held by numerous overseas commentators towards Chinese private capital participation in hard tech investment. They argue that Chinese private investors would not enter a sustainable technology investment cycle under government guidance. 

In fact, the difference in cash realisation rate was not mainly due to insufficient returns from domestic projects. Instead, it was largely the result of structural factors: China’s hard tech sector had limited ways for investors to exit, longer periods during which investments could not be sold, and greater uncertainty over reductions in state shareholdings. All these factors made it harder for private investors to turn their investments into cash.

The eventual listing model represented by CXMT has activated a positive cycle for Chinese private technology capital. It has helped address the longstanding challenges facing private capital in hard tech industries: difficulty exiting investments, slow capital circulation, and reluctance among private investors to commit funds to technology sectors. By improving the efficiency and likelihood of returns for domestic private capital, this model is encouraging more social capital to enter the AI and semiconductor industries, creating a positive cycle for industrial development.

Previously, the market was concerned that local government industrial funds and state-owned investment platforms, which had maintained large holdings in core technology companies over long periods, would conduct large-scale shareholding reductions after companies went public. Such selling pressure was expected to suppress valuations, reduce private investors’ profit potential, weaken incentives for private capital to exit, and limit the flow of capital into hard tech investment.

However, leading companies such as CXMT and Hygon Information Technology have demonstrated a different pattern. Local state-owned capital has maintained long-term holdings, conducted very limited reductions, and remained focused on developing industries rather than pursuing short-term financial returns.

Based on data from the complete 2016-2026 cycle, the exit and return structures of China’s three major categories of state-owned funds show significant differences. The National Integrated Circuit Industry Investment Fund has demonstrated the highest level of marketisation. It has achieved cumulative cash exits of 37.2 billion RMB, unrealised gains of 59.5 billion RMB, and a cash realisation rate of 38.5%, balancing industrial cultivation with capital rotation. 

Meanwhile, central SOE and SOE funds have achieved cumulative cash exits of 4.308 billion RMB, with unrealised gains of 105.638 billion RMB and a cash realisation rate of 3.9%. 

Local government industrial funds have taken a fundamentally different approach. They have achieved cumulative cash exits of only 2.886 billion RMB, while accumulating unrealised gains of 972.624 billion RMB. Their overall cash realisation rate is just 0.3%, effectively maintaining full holdings in core leading enterprises.

The primary objectives of local state-owned capital are industrial development, employment creation, and greater independence and control across industrial chains, rather than short-term financial gains. This characteristic has provided private capital with greater space for valuation growth and profit realisation. 

CXMT’s listing marks the normalisation of local state-owned capital’s “zero reduction, long-term holding” approach, effectively removing the primary concern of private investors. 

China’s current private capital cash realisation rate of 15.7% is now on a continued upward trajectory. As more domestic hard tech companies enter the capital markets and exit mechanisms continue to improve, private investors will find it increasingly easier to realise returns, with greater stability in investment outcomes.

From capital accumulation to technological breakthroughs

Compared with the US market, which relies heavily on merger and acquisition exits, short-term capital incentives and rapid technological iterations, China’s model of “state-owned capital supporting industrial stability and private capital driving market circulation” seeks to balance industrial security with market vitality.

The substantial unrealised gains are continuously recycled into industrial reinvestment, providing a steady stream of funding for core sectors such as AI large language models, computing chips, advanced packaging, and semiconductor equipment, thereby driving tangible technological breakthroughs. 

Over the past two years, supported by abundant, stable and long-term domestic capital supply, China’s AI and semiconductor industries have achieved a series of major breakthroughs. The density of breakthroughs, speed of commercial deployment, and scale of industrialisation have all reached historical highs.

In the field of AI computing chips, Hygon Information Technology has leveraged support from Chengdu state-owned capital and domestic private investment to achieve large-scale production of high-end CPUs and AI inference chips. The company has successfully challenged the monopoly of US high-performance computing chips and supported the large-scale deployment of domestic large language models. 

Domestic GPU companies such as MetaX Integrated Circuits and Biren Technology have completed multiple rounds of domestic financing and launched independently controlled high-end AI training chips. Their performance has gradually approached that of leading international products.

In semiconductor equipment and materials, companies including NAURA Technology Group and Advanced Micro-Fabrication Equipment have benefited from continued support from the National Integrated Circuit Industry Investment Fund. Their progress in critical technologies such as etching and thin film deposition has significantly improved domestic production capabilities, gradually replacing semiconductor equipment imported from the US and Japan.

The next phase of China-US AI competition

From the perspective of great power competition, the China-US AI race is, in financial terms, a contest between the resilience of their capital systems and their capacity for industrial incubation. 

As more domestic hard tech companies list on the capital markets, local state-owned capital maintains its long-term commitment to industrial value, and private capital continues to step up investment in strategic sectors, China’s AI industry will continue to strengthen both its capacity to generate capital internally and its efficiency in achieving technological breakthroughs. 

Over the coming years, the gap between the Chinese and US AI industries is expected to narrow rapidly. As the relative strengths of the Chinese and US capital systems continue to shift, the global landscape of AI competition is also set to undergo a profound reshaping.

Related: Can AI’s sky-high valuations survive disclosure? | From rule-taker to rule-maker: China’s AI challenge to US dominance