China Maps AI Dominance in $70B Five-Year Plan
China's National People's Congress opens this week with a 15th Five-Year Plan that puts $70 billion in semiconductor subsidies and AI-plus manufacturing at the center of its tech race with the West.

China is about to formalize the largest state-directed AI investment program in history. When the National People's Congress opens its annual session on Wednesday, lawmakers will table the 15th Five-Year Plan for 2026 to 2030 - a sweeping industrial blueprint that puts artificial intelligence, semiconductors, and robotics at the absolute center of economic policy. The price tag for chips alone: $70 billion in direct subsidies to domestic manufacturers like Huawei and Cambricon.
"The shock is over. Logically, it becomes a question of what China will produce next." - Alfredo Montufar-Helu, Ankura Consulting
This isn't speculative anymore. Beijing spent the last eighteen months proving that its AI labs can compete on performance while spending a fraction of what their American counterparts burn. Now it wants to turn those wins into industrial-scale deployment across manufacturing, logistics, energy, and governance - backed by government money at a scale that no Western democracy can match.
TL;DR
- China's NPC opens Wednesday with the 15th Five-Year Plan (2026-2030) centering AI and semiconductors as national security priorities
- $70 billion in semiconductor incentives earmarked for Huawei, Cambricon, and other domestic chipmakers
- Bank of China has pledged $137 billion over five years to strengthen the AI supply chain
- "AI-Plus" targets: 70% penetration of AI agents across the economy by 2027, 90% by 2030, full deployment by 2035
- Xi-Trump summit on March 31 to April 2 will put technology and trade at the top of the agenda
The Money
China's $70 billion in semiconductor incentives is the largest single tranche Beijing has committed to closing the chip gap with the West.
$70 Billion for Chips
The headline number in the Five-Year Plan's tech chapter is $70 billion in semiconductor incentives - the largest tranche Beijing has ever committed to closing the gap with TSMC and the Western fabrication ecosystem. The money flows primarily to Huawei's HiSilicon division and AI chip designer Cambricon, but the subsidy pool aims to pull in the entire domestic supply chain, from lithography equipment makers to advanced packaging firms.
This is Beijing's answer to the CHIPS Act. The difference is scale and speed. The U.S. allocated $52.7 billion across the entire semiconductor supply chain in 2022 and is still working through disbursements. China's $70 billion is additive to a decade of previous subsidies, and the state apparatus can deploy capital without the permitting delays and political wrangling that have slowed American fab construction.
Huawei has already demonstrated what subsidized innovation looks like in practice. Its Ascend 910C chips power the training runs behind Zhipu's GLM-5, and the Atlas 950 inference accelerator launched globally at MWC last week as a direct challenge to NVIDIA's dominance in the data center.
$137 Billion in Lending
Separately, Bank of China has pledged $137 billion over five years to strengthen the AI supply chain - covering everything from data center construction to talent acquisition to raw materials procurement. This isn't government spending in the traditional sense. It's state-directed lending at concessionary rates, a mechanism Beijing has used before to build out electric vehicle, solar panel, and 5G infrastructure at speeds that caught Western competitors off guard.
Combined with the semiconductor subsidies, the total state-backed financial commitment to AI infrastructure exceeds $200 billion over five years - more than double what the United States has committed through public programs.
Who Is Affected
| Stakeholder | Impact | Timeline |
|---|---|---|
| Chinese chipmakers (Huawei, Cambricon, SMIC) | Direct subsidies, guaranteed demand from SOEs | 2026-2030 |
| Chinese AI labs (DeepSeek, Alibaba, Zhipu, ByteDance) | Cheap domestic compute, state contracts | 2026-2028 |
| NVIDIA / AMD / Intel | Accelerated loss of China market share | Immediate |
| Western AI startups | Price pressure from subsidized Chinese competitors | 2027-2030 |
| Chinese manufacturers | Mandated AI adoption, productivity gains | 70% by 2027, 90% by 2030 |
| Chinese SMEs | Risk of falling behind as SOEs get priority access | 2026-2028 |
Companies
The plan's "AI-Plus" project sets explicit adoption targets: 70% penetration of AI agents across the Chinese economy by 2027, rising to 90% by 2030 and full deployment by 2035. By 2027, Beijing wants three to five general-purpose large language models rolled out in manufacturing, 100 high-quality industrial datasets, and 500 key application scenarios in production.
The immediate beneficiaries are the labs that have already proven they can compete. DeepSeek's R1 model matched GPT-4 performance at roughly $5.6 million in training costs - a fraction of what OpenAI spent. DeepSeek V4 is expected to launch this week as a multimodal flagship. Alibaba's Qwen 3.5 family overtook Meta's Llama as the most-downloaded model series on Hugging Face. ByteDance's Seedance 2.0 video model reportedly matched or passed OpenAI's Sora 2 in independent evaluations.
Chinese open-source AI model usage surged from 1.2% of global downloads in late 2024 to nearly 30% by late 2025. That arc is exactly what the Five-Year Plan is designed to accelerate.
Users
For developers and enterprises inside China, the message is clear: AI adoption is not optional. The government is building the infrastructure, subsidizing the compute, and setting adoption mandates. Shenzhen's Futian district has already rolled out 70 DeepSeek-powered "AI public servants" handling document processing, public services, and emergency management across 11 government functions.
The risk is that smaller companies get left behind. Analysts warn that the gap between large state-owned enterprises with privileged access to subsidized AI tools and SMEs paying market rates could widen sharply. In sectors like electric vehicles and solar panels, China's subsidy-driven growth eventually triggered brutal consolidation. The same pattern is likely to repeat in AI.
Competitors
For NVIDIA, this is the scenario that export controls were supposed to prevent. Despite U.S. restrictions on advanced chip sales to China, Huawei's domestically produced alternatives are now powering frontier model training. The $70 billion in new subsidies will only accelerate that substitution. Huawei shipped an estimated 200,000 Ascend 910B chips in 2025 alone, and the newer 910C is already in mass production.
China now has over 150 domestic humanoid robot developers, though regulators have flagged the number as unsustainably high.
The robotics angle adds another dimension. China now has over 150 domestic humanoid robot developers - a number that regulators themselves have flagged as unsustainably high. The Five-Year Plan is expected to focus on embodied AI as a strategic sector, channeling funding toward the most viable companies and letting the rest consolidate. The CCTV Spring Festival gala in January featured humanoid robots performing dance and martial arts routines - a signal from the state that this is no longer a research project.
What Happens Next
Bank of China's $137 billion lending commitment covers everything from data center construction to AI talent acquisition.
Three dates matter.
March 5 - the NPC formally convenes. The government work report and budget plans will confirm how much of the Five-Year Plan's tech spending is front-loaded into 2026 versus back-loaded toward 2030. Front-loading would signal urgency. Back-loading would suggest that the ambitious targets are aspirational rather than operational.
March 31 to April 2 - Xi Jinping and Donald Trump meet for a summit where technology, trade, and supply chains will be at the top of the agenda. China expanded export controls on rare earth minerals over the past year, causing 15 to 20% price increases for materials critical to chip manufacturing. Whether those restrictions get tightened or eased depends largely on what happens at this summit.
Second half of 2026 - the plan calls for MiniMax to release its M3 model and for DeepSeek to ship V4 as part of a broader push to show that Chinese labs can maintain frontier capability without access to NVIDIA's latest silicon. If those launches land, the argument that export controls are containing China's AI ambitions will be difficult to sustain.
The numbers are enormous. The question - the one that matters for investors, policymakers, and anyone competing with Chinese AI companies - is whether state-directed capital at this scale can actually produce the sustained innovation that Beijing is betting on, or whether it produces the same overcapacity and waste that plagued China's previous industrial campaigns. The semiconductor industry is a harder problem than solar panels. But $207 billion in combined backing buys a lot of tries.
Sources:
- China's annual parliament meet to unveil roadmap for tech race with the West - Reuters via Yahoo Finance
- China's Parliament Session to Unveil Bold Tech Strategy - Techi
- China to Accelerate AI, Robotics And Space Ambitions in New Five-Year Plan - Arise News
- China Five-Year Plan Bets Big on Tech Self-Reliance - The Dupree Report
- China seeks semiconductor and AI self-reliance in ambitious new 5-year plan - Tom's Hardware
- China's AI push: A driving force in the upcoming 15th Five-Year Plan - CGTN
- China issues guideline to accelerate AI-Plus integration - Gov.cn
