China's Strategic Power Subsidies for AI Chips: Investing in the Next Stage of Semiconductor Self-Reliance

China's push for semiconductor self-reliance has entered a critical phase, driven by a combination of geopolitical pressures and strategic foresight. As U.S. export controls tighten access to advanced GPUs and design tools, Beijing has doubled down on its AI chip ambitions, allocating billions to domestic manufacturing, R&D, and corporate partnerships. This is not just about catching up-it's about reshaping the global AI landscape.

The Strategic Push: From Subsidies to 3D Integration

According to a MERICS report, China's AI self-reliance strategy hinges on overcoming bottlenecks in lithography, packaging, and software ecosystems. The government's National Integrated Circuit Industry Investment Fund (Big Fund III) exemplifies this focus. In Q3 2025, Big Fund III committed RMB 450 million to Piotech Jianke, a subsidiary of Piotech, to accelerate 3D integration technology-a critical area for advanced packaging, according to a TrendForce report. This investment, part of a RMB 71 billion fund, signals a shift from broad-based to targeted support, prioritizing segments where China lags, such as 3D chip stacking and high-bandwidth memory, the TrendForce report notes.

The move is strategic: 3D integration could bypass traditional 2D scaling limitations, enabling more efficient AI chip architectures. Piotech, already a leader in thin-film deposition equipment, is leveraging its expertise to expand into this domain, a step that could reduce reliance on foreign tools like those from ASML or TSMC, the TrendForce piece suggests.

Challenges and the Road to Self-Reliance

Despite these strides, challenges persist. Huawei's AI chips, such as the 910B and 910C, still trail U.S. counterparts like Nvidia's A100 in performance, largely due to restricted access to advanced packaging and memory technologies, the MERICS report explains. Software is another hurdle: Huawei's CANN framework lacks the maturity of CUDA, deterring developers from adopting its ecosystem, the report adds.

However, China is addressing these gaps through education and incremental innovation. Universities are scaling AI programs to build a talent pipeline, while local labs are optimizing techniques like parameter-efficient fine-tuning to maximize existing hardware, the MERICS analysis finds. Meanwhile, clean energy projects are being expanded to power AI infrastructure, ensuring scalability for data centers and model training, according to the MERICS report.

Investment Opportunities: Where to Watch

For investors, the next stage of China's semiconductor self-reliance offers opportunities in three areas:
1. Advanced Packaging and 3D Integration: Companies like Piotech Jianke, backed by Big Fund III, are positioned to benefit from this niche.
2. R&D-Driven Startups: Firms focusing on domain-specific accelerators or privacy-preserving AI tools could gain traction as the ecosystem matures.
3. Corporate Partnerships: Collaborations between domestic players and international firms (e.g., NVIDIA's technologies integrated into broader AI platforms) may unlock hybrid solutions that bridge the gap, according to a TS2 Tech article.

While specific 2025 budget figures remain opaque, the scale of Big Fund III's investments suggests a long-term commitment. The RMB 71 billion fund is expected to prioritize sectors with low self-sufficiency, such as equipment and materials, creating a fertile ground for venture capital and private equity.

The Global Implications

China's AI chip ambitions are not just domestic-they're reshaping the global machine learning ecosystem. As more countries invest in localized compute and AI middleware, we can expect diversified benchmarks, parallel ecosystems, and a shift toward standards-based collaboration, the MERICS report argues. For investors, this means opportunities in cross-border partnerships and shared evaluation protocols, even as geopolitical tensions persist.

In the end, China's path to semiconductor self-reliance is a marathon, not a sprint. But with subsidies, strategic investments, and a focus on 3D integration, the next stage is already unfolding-one chip at a time.