---
title: China Published a 15-Year AI Roadmap. The West Can’t Plan Past Next Quarter
description: China’s 15th Five-Year Plan targets 90% AI integration by 2030. With energy dominance, supply chain control, and 14 plans delivered, Beijing can execute. Can anyone else?
author: Darie Nani (Editor-in-Chief)
date: 2026-03-06T13:30:00.000Z
updated: 2026-03-08T01:38:58.888Z
canonical: https://www.sovereignmagazine.com/article/china-published-a-15-year-ai-roadmap-the-west-can-t-plan-past-next-quarter
image: https://cdn.nanimediahouse.com/pexels-majestic-view-of-shanghai-s-illuminated-skyline-featuring-ic-169647.jpg
categories: Politics, Artificial Intelligence
content_type: Analysis
region: China
publication: Sovereign Magazine
about:
  - type: Place
    name: China
---

China’s 15th Five-Year Plan landed on Thursday inside the Great Hall of the People — 141 pages of economic blueprint covering 2026 through 2030. AI appears 52 times. In the previous plan, released in 2021, it appeared 11.

The document introduces a sweeping “AI+ action plan” targeting 70 per cent AI penetration across China’s economy by 2027, rising to 90 per cent by 2030 and ubiquitous deployment by 2035. It commits to scalable quantum computers, a space-earth quantum communication network, humanoid robotics, 6G, brain-machine interfaces, and nuclear fusion. The National Venture Guidance Fund has allocated RMB 121.8 billion ($17.5 billion) across three regional quantum hubs: Beijing-Tianjin-Hebei for computing, the Yangtze River Delta for communications, and the Greater Bay Area for commercial products.

Most Western coverage has treated the plan as either a threat to worry about or a wish list to be sceptical of. Both miss the point. China has published 14 five-year plans since the start of Reform and Opening-up in 1978. GDP has grown through every one. The country eliminated absolute poverty on schedule by 2020. When this government commits to building something, the track record says it gets built.

The more interesting question is not whether China can execute its plan. It is whether any of the countries anxious about it could execute one of their own.

## The energy foundation nobody else has

An AI economy runs on electricity. China’s total installed power generation capacity reached 3.89 terawatts by the end of 2025, up 16.1 per cent year on year. Clean energy capacity hit 52 per cent in February 2026, meaning fossil fuels are now the minority source for the first time. Wind and solar additions in 2025 alone reached 430 GW — roughly eight times what the United States added over the same period.

Solar capacity is on track to surpass coal this year. China manufactures more than 80 per cent of the world’s solar panels and [invested $625 billion in clean energy in 2024](https://www.sovereignmagazine.com/article/china-s-climate-gambit-as-beijing-sets-ambitious-2035-targets-to-lead-global-energy-transitio), nearly a third of the global total. Its projected solar manufacturing capacity for 2030 — 1,255 GW — is 65 per cent higher than what the International Energy Agency’s Net Zero Roadmap forecasts the entire world will need.

In the United States, [AI data centres are fighting for grid connections](https://www.sovereignmagazine.com/article/too-little-too-late-america-lost-the-energy-transition-to-china-trump-s-400m-won-t-change-tha). In China, the grid is expanding fast enough to absorb them.

## The supply chain that runs in one direction

China controls roughly 70 per cent of rare earth mining, 90 to 95 per cent of processing and refinement, and [93 per cent of rare earth magnet manufacturing](https://www.sovereignmagazine.com/article/the-new-tech-hierarchy-s-impact-on-global-manufacturing-as-china-is-winning-the-ai-race). It produces three-quarters of the world’s lithium-ion batteries, 70 per cent of cathode capacity, and 98 per cent of lithium iron phosphate (LFP) active materials. CATL and BYD hold 55 per cent of global market share in EV battery installations.

These are not abstract trade statistics. An F-35 fighter jet requires 435 kilograms of rare earths. A next-generation US destroyer needs 4.5 tons. A nuclear submarine needs 1.5 tons. Over 80,000 parts in Pentagon weapons systems rely on critical minerals subject to Chinese export controls. Every magnet in the 1.2 million combat drones Ukraine produced in 2024 was manufactured in China.

When Beijing restricted exports of seven rare earth elements in April 2025 — a direct response to Trump’s tariff escalation — a US drone-parts manufacturer supplying the military reported order delays of up to two months while searching for non-Chinese sources. Samarium, used in high-temperature magnets for jet-fighter engines, was briefly offered at 60 times its standard price.

No other major economy controls its own supply chain to this degree. The dependency runs in one direction, and the country on the receiving end is not China.

## Tariffs moved the ports, not the products

The tariffs were supposed to reduce American dependence on Chinese manufacturing. On the surface, they worked: China’s share of US imports fell from above 20 per cent before Trump’s first trade war in 2018 to 7 per cent by 2025.

But the overall US trade deficit barely moved. It dropped by $2.1 billion in 2025, driven almost entirely by services. The goods deficit shifted to Mexico and Vietnam. Over $8 billion in Chinese exports were rerouted through Vietnam in the first three quarters of 2025 alone, passing through without substantial transformation — an open secret the administration has acknowledged but struggled to close. Eighty per cent of Vietnamese shipments came from companies with 100 per cent Chinese ownership.

China, meanwhile, stopped buying American exports entirely in April 2025. US goods shipments to the world’s third-largest importing economy fell 26 per cent year on year, back to levels last seen during the 2008 financial crisis. As a share of total US output, sales to China dropped to nearly half their 2017 levels.

The net result: China posted a record trillion-dollar trade surplus. The tariffs amount to the largest US tax increase as a percentage of GDP since 1993 — roughly $1,500 per American household per year. Eighty per cent of the customs duties are borne by US businesses or consumers, not Chinese exporters.

## The chip rejection

In December 2025, the Trump administration cleared Nvidia to sell its H200 AI processors to Chinese companies on a case-by-case basis. One day later, [Chinese customs officers were told not to let them into the country](https://www.sovereignmagazine.com/article/china-bans-nvidia-chips-after-decade-of-us-tech-restrictions). Beijing issued guidance directing domestic companies to purchase H200 chips only when necessary and to match any foreign purchases with equivalent domestic chip orders.

Nvidia has since stopped holding H200 production capacity for China entirely, redirecting TSMC fabrication to its next-generation Vera Rubin architecture. Roughly 250,000 H200 units had already been produced, but the Chinese market for them effectively closed before they shipped.

DeepSeek’s trajectory illustrates the point. Its R1 reasoning model, released in January 2025, [matched OpenAI’s GPT-4 performance](https://www.sovereignmagazine.com/article/china-s-ai-rise-innovation-overcomes-chipmaking-and-investment-gaps) for approximately $5.6 million using older Nvidia A800 chips already subject to export restrictions. A year later, DeepSeek V4 — a trillion-parameter multimodal system optimised for Chinese chips, with a one-million-token context window — is expected to launch in the first week of March 2026. The gap between constrained hardware and competitive output is closing faster than Washington’s export controls can widen it.

Huawei’s Ascend processors are closing the gap. Xi Jinping instructed the Politburo to “concentrate resources to overcome challenges in core technologies such as high-end chips” and ensure China’s AI system operates independently of foreign suppliers. The H200 episode is instructive: China is not waiting for American permission to build its AI stack. It is designing the replacement.

## What the competitors are working with

The United States has no national industrial strategy for AI beyond a Department of Defense paper focused on military applications. The Congressional Budget Office projects a $1.9 trillion deficit for fiscal 2026, with debt held by the public at 101 per cent of GDP, rising to 120 per cent by 2036. The government has been in partial shutdown since 14 February over DHS appropriations.

CSIS, a Washington think tank not known for alarmism, concluded that the US government “lacks the analytical capacity, institutional structures, and personnel with expertise to formulate and execute sophisticated industrial strategy.” Unlike China, where officials study industrial policy as a discipline and spend careers in specific sectors, American policymakers typically come from law or generalist economics. American business leaders, shaped by decades of financial capitalism, focus on quarterly returns and shareholder value rather than systemic industrial strength.

The CHIPS Act, the closest thing the US has to a semiconductor strategy, has faced implementation delays. Lynas Rare Earths’ Texas processing facility, developed with Pentagon support, has been stuck in permitting disputes over wastewater disposal, casting doubt on its 2026 production timeline. New procurement rules banning Chinese-sourced rare earths from the US defence supply chain do not take effect until January 2027 — and the alternative sources those rules require do not yet exist at scale.

The European Union has spent its bandwidth on the AI Act, a regulatory framework it cannot implement on time. The European Commission missed its own deadline for Article 6 guidance on high-risk AI systems in February 2026. Standards bodies tasked with developing compliance frameworks have been delayed until the end of the year. The proposed “Digital Omnibus” package pushes back high-risk requirements by up to 16 months. Twenty-seven member states, fragmented enforcement, and no fabrication capacity. Europe regulates AI. It does not build it.

India has set out to attract $200 billion in AI infrastructure investment by 2028 — Reliance and Adani are committing $110 billion and $100 billion respectively — but faces structural constraints: unreliable power, water shortages for data centre cooling, skills gaps, and weak institutional linkages between infrastructure build-out and actual utilisation. The ambition is real. The grid is not ready.

## The bar is not 100 per cent

MERICS, a European China policy institute, has questioned whether the 70 per cent AI penetration target by 2027 is a genuine deliverable or a signal to provincial cadres. Cash-strapped local governments may lack the budget and expertise to implement every mandate. China’s property sector is in its fifth year of crisis. The population declined for a fourth straight year in 2025, shrinking by about 3 million. The 2026 GDP target of 4.5 to 5 per cent is the lowest in 35 years.

None of that changes the maths. Suppose Beijing delivers half — 35 per cent AI penetration instead of 70, $8 billion of the $17.5 billion quantum spend deployed, 150 humanoid robot models reaching commercial production instead of 300. That still dwarfs anything Washington, Brussels, or London has attempted. The UK cut HS2 in half, ran billions over budget, and still has not built it — if half of it actually opened, it would be called a national triumph.

Property, infrastructure, and manufacturing investment all declined simultaneously for the first time last year. The plan addresses each gap with programmes already in motion: 300-plus humanoid robot models for labour shortages, DeepSeek V4 on domestic chips for the AI stack, and a trillion-dollar trade surplus funding it all. China’s quantum sector hit $1.61 billion in 2025, growing at 30 per cent annually. Clean energy capacity already exceeds fossil fuels. No Western government has published a comparable programme, let alone started delivering one.

## Further Context

**Q: Does China have more clean energy than the US?**
China’s total installed power generation capacity reached 3.89 terawatts by the end of 2025, with clean energy sources crossing 52 per cent of total capacity in February 2026. Wind and solar additions in 2025 alone reached 430 GW — roughly eight times what the United States added. China invested $625 billion in clean energy in 2024, nearly a third of the global total, and manufactures more than 80 per cent of the world’s solar panels.

**Q: How does US military power depend on China’s rare earths?**
Over 80,000 parts in Pentagon weapons systems rely on critical minerals subject to Chinese export controls. An F-35 requires 435 kilograms of rare earths, a next-generation destroyer needs 4.5 tons, and a nuclear submarine needs 1.5 tons. China controls 90 to 95 per cent of rare earth processing and 93 per cent of magnet manufacturing. New procurement rules banning Chinese-sourced rare earths from the US defence supply chain do not take effect until January 2027, and alternative sources do not yet exist at scale.

**Q: Is China outpacing the US in technology?**
China has closed the gap or taken the lead in several advanced industry sectors. It dominates solar manufacturing, battery production, rare earth processing, and drone manufacturing. In AI, DeepSeek built GPT-4-competitive models on restricted hardware and is preparing to launch its trillion-parameter V4 model optimised for domestic chips. China released more than 300 humanoid robot models in 2025 — over half the global total — and has allocated $17.5 billion to quantum computing across three regional hubs. CSIS has noted that the US has lost its lead in four advanced industry sectors, with China closing the gap in the remaining five.

**Q: Why is China dominating lithium-ion battery production?**
China produces three-quarters of all lithium-ion batteries, 70 per cent of cathode capacity, and 98 per cent of lithium iron phosphate active materials. CATL and BYD hold 55 per cent of global EV battery installations. Chinese companies, backed by government support, have invested heavily in overseas lithium mining assets — often paying above market value — while building integrated domestic supply chains from raw material processing through to finished cells. Europe and the United States have begun efforts to reduce this dependency, but scaling alternative supply chains takes years.

**Q: Was China’s five-year plan successful?**
China has published 14 five-year plans since the start of Reform and Opening-up in 1978. GDP has grown through every plan period. The country eliminated absolute poverty on schedule by 2020 and has consistently met or exceeded its stated growth targets. The success of the plans lies in China’s ability to translate strategic goals into operational mechanisms across decades — from industrialisation and reform to sustainability and innovation. The 15th plan, covering 2026 to 2030, focuses on AI integration, quantum computing, and technological self-reliance.

**Q: What chips is China using for AI?**
Huawei’s Ascend processors are the primary domestic alternative to Nvidia hardware. DeepSeek trained its R1 reasoning model on restricted Nvidia A800 chips in 2025 and is launching its trillion-parameter V4 model in early March 2026, optimised to run on Chinese-made chips. Beijing has directed domestic companies to prioritise homegrown AI chips over imports, and Nvidia has stopped holding H200 production capacity for the Chinese market entirely.