China's IC Design Revenue Nears ¥1 Trillion, but Computing Power and CUDA Gaps Remain Significant
0xBroomberg
China's IC design revenue hit ¥826.1 billion in 2025, topping ¥1 trillion with IDM included — years ahead of the 2030 target. Yet computing chips account for just 7.7% of revenue, and a "triple dependency" on Nvidia exposes the sector's deepest structural weakness heading into the AI era.
Where does the trillion-yuan figure come from?
China's IC design sales reached ¥826.1 billion in 2025, up 24.8% year-on-year. Adding IDM (companies that both design and fabricate) in-house design revenue pushes the total past ¥1 trillion for the first time.
It also marks the first time the sector crossed $100 billion. The twenty-year compound annual growth rate is close to 20%, with no single year of decline on record.
This means → the 2030 scale target has been met years early. Speed is not the problem — structure is.
Nearly 4,000 firms — who is actually making money?
China has roughly 3,900 IC design companies. About 87% employ fewer than 100 people — the vast majority are small teams.
Around 800 firms earn more than ¥100 million a year. The top 50 account for roughly one-third of industry revenue; all firms above the ¥100-million line together contribute 84.7%.
In plain terms = the industry has moved from a start-up boom to a consolidation phase, but thousands of small firms still crowd the tail.
Where is the product mix weakest?
Telecom and consumer-electronics chips make up 66.5% of China's IC design revenue — a traditional strength.
Computing chips — CPUs, GPUs and other high-end processors — account for just 7.7%, far below the global average of roughly 25%.
This means → China is competitive in application-layer chips but lags badly in defining next-generation computing platforms — exactly where AI-era competition is decided.
How is AI reshaping chip demand?
Wei Shaojun (魏少军), chair of the IC design branch of the China Semiconductor Industry Association, argues AI is not a single new market but a reallocation of semiconductor value: demand is spreading from GPUs and AI accelerators into high-bandwidth memory, DDR, power management, SerDes interfaces — high-speed data links — and communications chips.
This reflects a long-term structural driver for semiconductor demand, not a short-cycle boom.
Yet the trend strikes precisely where China is weakest — advanced CPUs, the GPU ecosystem, and AI training chips — while export controls further restrict access to advanced processors and software toolchains.
How can the "triple dependency" be broken?
Wei describes China's reliance on Nvidia as threefold: models, architecture, and the software ecosystem — especially CUDA, Nvidia's GPU programming platform. The bottleneck is not just hardware access; even writing AI software depends on Nvidia's tools.
He sees inference as the more realistic breakthrough: as AI workloads shift from cloud training toward edge and endpoint devices, Chinese firms can target edge AI SoCs (system-on-chip), endpoint inference chips, and domain-specific accelerators rather than competing head-on with Nvidia's training GPUs.
AI is also changing chip design itself. Some Yangtze River Delta firms already use AI-assisted design flows, lifting efficiency by 5%–10%. Wei calls "EDA plus AI" a critical emerging capability.
Talent, capital, supply chain — where does it stall next?
Talent: junior and mid-level engineers are plentiful, but senior experts in architecture, system integration, and advanced process technology are in acute shortage. High turnover among key staff compounds the problem.
Capital discipline: the sector leans heavily on state capital, administrative decisions, and immature market mechanisms. Some firms survive on funding rounds and valuation agreements. Wei is blunt — semiconductors are an endurance race, not a short-term arbitrage play.
Supply-chain fragmentation: global technology restrictions now extend beyond chips to EDA software, materials, and high-end computing equipment. Multinationals may not want to leave China, but geopolitics is forcing supply-chain restructuring, deepening the ecosystem split.
Content is for reference only, not financial advice.