In March 2026, Unitree Robotics, a humanoid robot company, closed a 2.5 billion yuan funding round. The lead investor was the National AI Industry Investment Fund, a wholly owned vehicle of the Big Fund Phase III. Other participants included state-backed institutions, according to public filings. There was no Sequoia. No Andreessen Horowitz. No SoftBank.
Unitree fits a pattern that has become the norm in Chinese AI financing.
A September 2026 Rhodium Group report, Examining China's AI Financing, put numbers to what many in the industry had suspected. Announced equity deals in China's AI sector reached 282 billion yuan (approximately $40 billion) through mid-August 2026 — more than double recent yearly lows. Government guidance funds accounted for about 25 percent of that total. That figure was down from 30 percent in 2024 and 2025, but still far above the pre-pandemic average of 13 percent.
In the subsectors that matter most — AI chips and servers — the state's role is even more concentrated. According to Rhodium, state-affiliated sources accounted for more than 60 percent of equity investment in Chinese AI chips and servers. Of that total, 47 percent came from government guidance funds and investment platforms. Another 14 percent came from banks, most of them state-owned.
The mechanism is not a single fund. It is a layered matrix of national funds, provincial vehicles, state-owned enterprises, and privately managed funds that convert policy priorities into equity bets. As the South China Morning Post noted, state-affiliated investors supplied more than 90 percent of committed capital in China's private-equity market in 2025 — up from just under 79 percent in 2021.
China's AI industry is largely state-funded.
The Capital Stack
The Rhodium report estimated that China's top hyperscalers — Alibaba, Tencent, Baidu, and ByteDance — together with the three state telecom carriers, will spend 932 billion yuan (approximately $139 billion) on AI infrastructure in 2026. That is a 103 percent increase over 2025. The projection for 2027 exceeds 1.2 trillion yuan.
That spending is not financed the way American AI capex is financed. U.S. hyperscalers rely on debt markets, issuing bonds and tapping credit facilities. Chinese AI companies rely more heavily on equity issuance and bank loans. In the first eight months of 2026, Chinese AI companies raised 179 billion yuan through IPOs and private placements — more than double the entire 2025 total of 90 billion yuan. Alibaba and Tencent issued convertible bonds.
The distinction matters because equity financing is more cyclical than debt. Rhodium partner Logan Wright noted that China's leading AI labs "will be heavily dependent upon a favourable climate in the equity market." His assessment: "historically that's not an easy bet in China."
The cash flow pressure is already visible. In the first half of 2026, the combined free cash flow of Alibaba, Tencent, and Baidu turned negative 16 billion yuan — down from a positive 17 billion yuan a year earlier.
The Big Fund Phase III
The anchor of China's state AI financing architecture is the National Integrated Circuit Industry Investment Fund, known as the Big Fund. Its third phase, launched in May 2024 with registered capital of 344 billion yuan, is the largest of the three phases and exceeds the combined size of Phase I and Phase II.
Phase III marked a strategic pivot. Phase I and Phase II took a broad approach, investing across wafer fabrication, chip design, packaging, and equipment to rapidly expand domestic capacity. Phase III narrowed the focus. Approximately 70 percent of its capital is directed toward semiconductor equipment and materials — the upstream bottlenecks that determine whether China can manufacture advanced chips at scale. The remaining 30 percent is allocated to advanced packaging and AI memory technologies, including HBM.
Two core platforms manage the deployment. Huaxin Dingxin focuses on chip design IP and advanced packaging. Guotou Jixin invests in semiconductor equipment, lithography materials, and high-purity consumables. The fund also established a dedicated National AI Industry Investment Fund with 60.06 billion yuan in capital, specifically targeting AI compute startups.
That AI fund has already made notable investments. In March 2026, it led a 2.5 billion yuan round in Unitree Robotics — the first time the Big Fund invested in embodied intelligence. In July, it participated in DeepSeek's first external funding round, contributing approximately 980 million yuan. In August, it injected 1.4 billion yuan into Kuaishou's Kling AI video unit.
The investments are not passive. The National AI Industry Investment Fund was reported to have received an exemption from equity lock-up restrictions when it invested in DeepSeek — a signal that the state is willing to structure deals flexibly when strategic priorities are at stake.
The 1 Trillion Yuan Venture Fund
In December 2025, China launched the National Venture Capital Guidance Fund, a state-backed vehicle with an initial central government contribution of approximately 100 billion yuan and a target of mobilizing 1 trillion yuan in total capital.
The fund's structure reflects a recognition that traditional Chinese venture funds operate on 5-7 year cycles — too short for semiconductor and AI development. The National Venture Capital Guidance Fund set a 20-year lifespan: 10 years of investment, 10 years of exit. Its mandate is explicit: "invest early, invest small, invest for the long term, invest in hard technology."
The priority sectors are AI, quantum computing, semiconductors, biopharmaceuticals, aerospace, and 6G communications. The fund is designed to fill the gap that private venture capital has vacated. As Western VCs and domestic private wealth retreated from Chinese hard tech, the state expanded.
Why This Structure Exists
The state-led financing model reflects the specific economics of AI infrastructure in China.
Chinese AI companies compete on price, not margin. DeepSeek's V4-Flash model ran more than 100 times cheaper than a leading Anthropic offering as of early August 2026. On OpenRouter, DeepSeek, Z.AI, Tencent, and Alibaba together claimed 45 percent of transactions in the week ending September 14. But profitability remains distant. Z.AI, valued at $40 billion after its Hong Kong listing, spent 2.1 billion yuan on research in the first half while generating less than half that in revenue. DeepSeek's gross margin in July was approximately 45 percent, compared with Anthropic's 65 percent. Zhipu AI and MiniMax reported first-half gross margins of 26 percent and 18 percent.
Price competition that compresses margin requires patient capital. Private VC funds that need to return money to limited partners within a decade cannot sustain a portfolio of companies operating at those margins. State funds with 20-year horizons and policy mandates can.
The valuation-to-revenue multiples tell the same story. Rhodium estimated DeepSeek's multiple at 163 and Moonshot's at 50, compared with 34 for OpenAI and 21 for Anthropic. The Chinese companies are priced on strategic optionality rather than current earnings.
What the Numbers Mean
China's AI industry has a financing structure that looks different from Silicon Valley's. The capital is largely state-directed, deployed through layered funds with long horizons, and concentrated in the infrastructure layer — chips, servers, and compute — rather than in model developers.
That structure has delivered scale. China's AI infrastructure spending will reach approximately $139 billion in 2026. It has delivered speed: the national intelligent compute scale reached 2,185 EFLOPS by June 2026, a 177 percent year-over-year increase.
It has also created dependencies. Rhodium's Logan Wright noted that the financing gap means China's frontier labs "will find it harder to expand sustainably." The state can fund chips and servers. It cannot guarantee that the model developers who use those chips will generate enough revenue to justify their valuations.
The 60 percent state share in chip and server equity reflects a choice to fund the layer of the AI stack that determines long-term capacity, while accepting lower returns in the layer that determines short-term profitability.
Whether that choice pays off depends less on the model developers than on the equity market that has financed them. If it remains open long enough for the returns to materialize, the structure works. If it closes, the state will be the last investor standing.
Sources: Rhodium Group "Examining China's AI Financing" (September 17, 2026) via WebProNews (September 20, 2026); X-TechCon (September 21, 2026); TechFocus24 (September 21, 2026); SCMP (July 25, 2026); Sedaily (July 14, 2026); TrendForce (August 7, 2026); SECCW (August 5, 2026); FT Chinese (August 19, 2026); 163.com (March 2, 2026; September 1, 2026); Sohu (July 16, 2026); Economic Daily (December 26, 2025); NDRC (January 6, 2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.
Forecasts from third-party analysts can change with market conditions.
Cost and pricing examples are point-in-time estimates; actual rates vary.
Country and company comparisons rely on public reporting, not operational data.
This sector moves fast; timelines and deal terms may be updated later.
Company deals and regulatory rulings may evolve; verify current status.
AI infrastructure is changing quickly; claims can become outdated soon.
Sources
- Rhodium Group "Examining China's AI Financing" (September 17, 2026) via WebProNews (September 20, 2026)
- X-TechCon (September 21, 2026)
- TechFocus24 (September 21, 2026)
- SCMP (July 25, 2026)
- Sedaily (July 14, 2026)
- TrendForce (August 7, 2026)
- SECCW (August 5, 2026)
- FT Chinese (August 19, 2026)
- 163.com (March 2, 2026; September 1, 2026)
- Sohu (July 16, 2026)
- Economic Daily (December 26, 2025)
- NDRC (January 6, 2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.; Forecasts from third-party analysts can change with market conditions.; Cost and pricing examples are point-in-time estimates; actual rates vary.; Country and company comparisons rely on public reporting, not operational data.; This sector moves fast; timelines and deal terms may be updated later.; Company deals and regulatory rulings may evolve; verify current status.; AI infrastructure is changing quickly; claims can become outdated soon.