While Silicon Valley pours billions into the next frontier model, Beijing is playing a different game entirely, embedding✦ artificial intelligence into every car, battery, drone, and factory in the country. The West may be winning the AI arms race it defined. But China might be winning the one that actually matters.
There is a conversation happening in Western policy circles that goes something like this: Who will build the most powerful AI? The answer, depending on which think tank you ask, is either OpenAI, Google DeepMind, or some Chinese lab nobody has heard of yet (assuming the export controls fail).
It is the wrong question.
The more important question is not who builds the smartest model, but who puts AI to work fastest across the largest number of industries. And on that metric, China is not just competitive. It is pulling away.
The distinction matters more than most Western analysts appreciate. The United States and Europe have built an AI ecosystem✦ centred on general-purpose foundation models: large language models, multimodal✦ systems, and reasoning engines designed to do everything for everyone. China has taken a fundamentally different path. Rather than competing primarily on model capability benchmarks, Beijing has prioritised what it calls "AI+," the systematic integration of artificial intelligence into specific vertical✦ industries, from electric vehicles to battery manufacturing, from drone logistics to telecommunications infrastructure.
The result is not one Chinese AI champion. It is dozens of world-leading companies that happen to be extraordinary at AI because their industries demanded it.
BYD: The AI Car Company Hiding in Plain Sight
The case study that best illustrates China's vertical AI thesis is not Baidu or Alibaba. It is BYD.
Most Western observers still categorise BYD as an electric vehicle manufacturer, Tesla's Chinese rival. That framing misses what BYD has actually become: an AI-integrated vehicle platform company that happens to sell cars.
In early 2025, BYD rolled out its "God's Eye" advanced driver-assistance system (ADAS) as standard equipment across its entire lineup, from the $10,000 Seagull to the premium Yangwang U8. This was not a luxury add-on. It was a strategic declaration: every BYD vehicle would become an AI platform.
The architecture operates on three tiers. The top tier, DiPilot 600, deploys a triple-lidar configuration with NVIDIA Orin X processors delivering over 500 TOPS of compute✦ power, enabling nationwide autonomous navigation without pre-mapped routes. The mid-tier system handles urban navigation with a single lidar. Even the entry-level tier, installed on cars costing less than $15,000, provides lane-keeping, adaptive cruise, emergency braking, and autonomous parking.
What makes this strategically significant is scale. By late 2025, BYD had over 2.5 million God's Eye-equipped vehicles on the road generating 150 to 160 million kilometres of driving data every single day. That volume of real-world training data, flowing back to a single company's AI systems, represents a vertical integration advantage that no Western automaker can match. Tesla has its fleet data advantage too, but BYD's approach of making advanced autonomy standard (not optional) across budget to luxury vehicles means its data pipeline is broader and denser.
BYD has committed $14.3 billion to AI agent and world model development, supported by a team of over 5,000 engineers dedicated to autonomous driving. The company has integrated DeepSeek's AI reasoning capabilities into its Wanji architecture for real-time decision-making, achieving 98.7% efficiency through knowledge distillation on edge devices. By March 2026, God's Eye 5.0 was bringing high-end autonomous capabilities to vehicles priced at just $14,000.
The AI extends beyond driving. BYD's manufacturing AI has reduced battery defects by 40% and improved average battery lifespan by 20%. The company's 90,000 engineers and $20 billion R&D budget are not split between "AI research" and "car engineering." They are one integrated operation where AI is the connective tissue across design, manufacturing, quality control, and the driving experience itself.
This is what vertical AI looks like. Not a chatbot. A car.
The Vertical AI Ecosystem: It Is Not Just BYD
BYD is the most visible example, but China's vertical AI strategy runs across its entire industrial base. Consider what is happening simultaneously:
CATL, the world's largest battery manufacturer with 38% global market share, has turned battery design from manual trial-and-error into an AI-driven✦ process. Its intelligent design platform, trained on more than 100,000 battery design cases and 600 terabytes of test data, generates optimised cell designs in minutes with 95% prediction accuracy. The company's manufacturing lines monitor over 6,800 quality control points using real-time AI image recognition, producing one battery cell per second. CATL won the World Economic Forum's MINDS Award for AI-driven innovation in January 2026, and it is now deploying humanoid robots for end-of-line inspection that autonomously detect wiring anomalies in real time. From 2026, CATL plans to run large sections of its factories with AI systems and humanoid robots, eliminating human workers from high-voltage inspection environments entirely.
DJI, which dominates the global commercial drone market, has shifted its enterprise strategy from "flying cameras" to "aerial AI platforms." The company's 2026 Enterprise Drone Onboard AI Challenge invites developers to deploy AI algorithms directly on drones for real-time anomaly detection, infrastructure inspection, and agricultural monitoring. The Manifold 3 edge computing module processes sensor data, flight control, and route planning onboard, with no cloud dependency required. DJI is not competing with OpenAI. It is building the world's best AI for flying machines.
Huawei, locked out of advanced Western chips by export controls, has responded by building its own AI computing ecosystem. The Ascend 910C chip delivers 800 teraflops of FP16 performance with 128 GB of memory, and the Atlas 900 A3 SuperPod clusters up to 384 of these chips for 300 petaflops of AI compute. Huawei shipped 700,000 Ascend AI chips in 2025, with 600,000 units of the 910C planned for 2026. The Shenzhen AI cluster built on Ascend hardware is now China's largest homegrown AI compute facility. Huawei's Pangu large models are not designed for consumer chat. They are optimised for industrial applications like predictive maintenance, network optimisation, and manufacturing quality control.
The capabilities these companies are deploying — autonomous vehicles navigating without maps, AI systems designing batteries, drones making real-time decisions without human input — read like science fiction, but they are shipping at industrial scale today. For a look at just how many other "impossible" AI scenarios have already crossed into reality, see our companion piece on when sci-fi stopped being fiction.
Yes, China Builds Foundation Models Too. And That Is Part of the Strategy.
A common Western misconception frames the picture as a binary: the West builds foundation models, China deploys applied AI. The reality is more nuanced. China also builds frontier-class general-purpose models, and it does so with a speed and cost efficiency that has repeatedly blindsided Silicon Valley.
DeepSeek is the headline example. A relatively small Hangzhou lab that reportedly built its R1 reasoning model for a fraction of the cost of comparable Western systems, DeepSeek sent shockwaves through global markets when it demonstrated performance rivalling OpenAI's best on key benchmarks. The model is open-weight✦, free to use, and already integrated into industrial applications across China, including BYD's autonomous driving architecture.
Alibaba's Qwen series has become one of the most widely deployed model families in Asia. In six months during 2025, Alibaba expanded AI integration from smart appliances to AI glasses and smart coffee machines across hundreds of product categories. Tencent's Hunyuan models are embedded directly into WeChat and Tencent Cloud, reaching hundreds of millions of users. Baidu's ERNIE powers enterprise workflows through agent-driven systems.
The critical difference is not capability. It is purpose. Western labs treat foundation models as the primary product, monetised through API✦ access and enterprise licensing. Chinese companies treat foundation models as infrastructure, the plumbing that enables vertical AI applications across manufacturing, logistics, consumer devices, and autonomous systems. DeepSeek's open-weight strategy is not charity. It is an accelerant for ecosystem-wide adoption across China's enormous industrial base.
The March 2026 U.S.-China Economic and Security Review Commission report confirmed this pattern: most major Chinese AI labs publish their model source code and weights openly. The report, titled "How China's Open AI Strategy Reinforces Its Industrial Dominance," laid out how this approach creates a feedback loop. Open models accelerate adoption. Adoption generates proprietary industrial data. Proprietary data feeds back into better models. The loop compounds.
The Policy Architecture: AI+ as Industrial Strategy
China's vertical AI approach is not accidental. It is the product of deliberate industrial policy stretching back more than a decade.
The AI+ Action Plan, issued by China's State Council in August 2025, codifies the strategy: achieve 70% AI penetration across key economic sectors by 2027, supported by intelligent terminals and AI agents. The plan targets a core AI industry worth over $140 billion by 2035, with related industries reaching $1.4 trillion. Four key initiatives drive execution: multi-tier platform cultivation, data aggregation for intelligence enhancement, large-scale industrial AI applications, and ecosystem development for deployment.
The 15th Five-Year Plan (2026-2030), unveiled in March 2026, extends these ambitions with specific targets for AI and cybersecurity integration. Where previous plans emphasised building AI capabilities, this one emphasises deploying them in manufacturing, healthcare, renewable energy, agriculture, and smart city infrastructure. Guangdong province's intelligent factories already use 5G-connected HD cameras running AI quality inspection at production line speeds.
China's regulatory framework✦, encompassing the Data Security Law, Generative AI✦ Measures, and a standards framework targeting 1,000+ firms by 2026, is designed not to slow AI development but to channel it toward industrial deployment and economic output. The approach contrasts sharply with Western regulatory efforts, which tend to focus on safety constraints for general-purpose models.
The Western Approach: Brilliant Models, Slower Integration
The contrast with the Western AI ecosystem is stark.
The United States has built the world's most capable foundation models. OpenAI, Anthropic, Google DeepMind, and Meta have pushed the frontier of what AI can reason about, write, code, and understand. These are genuine achievements. But the strategic question is whether model capability translates into industrial transformation at the pace China is achieving.
The Western AI economy is structured around a horizontal✦ model: build a general-purpose platform and let industries figure out how to use it. OpenAI sells API access. Anthropic licenses Claude to enterprises. Google integrates Gemini into its cloud. The assumption is that the best foundation model✦ wins, and the applications will follow.
China's bet is different. It assumes that the best AI in any given industry will be built by companies that understand that industry deeply, and that AI capability should be measured not by benchmark✦ scores but by defect rates, kilometres driven, batteries optimised, and drones navigated.
| Dimension | China's Vertical Approach | The West's Horizontal Approach |
|---|---|---|
| Primary focus | Embed AI into specific industries at scale✦ | Build the most capable general-purpose models |
| Lead actors | Industrial companies (BYD, CATL, DJI, Huawei) | AI labs (OpenAI, Anthropic, Google DeepMind) |
| Success metric | Deployment penetration, industrial output gains | Benchmark scores, model capability, API revenue |
| Data strategy | Sector-specific data flywheels (driving data, battery data, manufacturing data) | Broad internet-scale training data |
| Policy framework | AI+ Action Plan: 70% sector penetration by 2027 | Export controls, safety regulation, compute restrictions |
| Model philosophy | Open-source for rapid ecosystem adoption | Proprietary models as competitive moats |
| Hardware approach | Build domestic alternatives (Huawei Ascend) | Maintain chip export restrictions (NVIDIA controls) |
| Foundation models | Competitive (DeepSeek, Qwen, ERNIE) and treated as infrastructure | Frontier-class and treated as the primary product |
| Integration speed | Standard across product lines (BYD God's Eye on all vehicles) | Optional enterprise add-ons, gradual rollout |
| Risk | Quality and safety at scale, data governance | Model capability plateaus, slow industrial adoption |
What the West Is Missing
The strategic blind spot in Western AI discourse is the assumption that general-purpose model supremacy equals AI leadership. By this logic, whichever country or company builds GPT-7 or its equivalent wins the AI race.
But this framing ignores how value is actually created in the real economy. AI does not generate GDP by passing reasoning benchmarks. It generates GDP by making cars safer, batteries cheaper, factories faster, logistics smarter, and energy systems more efficient. China understands this and has structured its entire national AI strategy around it.
The numbers tell the story. China now controls 69% of the global EV battery market, with CATL and BYD together commanding the majority. BYD is on track to produce 5 million vehicles in 2025. DJI holds over 70% of the global consumer drone market. Huawei's Ascend chips now claim 41% of China's AI server market. These are not AI companies by Silicon Valley's definition. They are AI companies by the definition that matters: organisations using artificial intelligence to dominate their industries.
The Western response (tightening export controls on chips and trying to build AI safety✦ frameworks) may constrain China's access to cutting-edge✦ compute but does little to address the vertical integration advantage. Even if China cannot access NVIDIA's latest GPUs, it does not need them to run quality inspection AI in a battery factory or obstacle avoidance AI on a $14,000 car. The AI models powering these applications are specialised, efficient, and increasingly trained on proprietary industrial data that no Western company possesses.
The AIinASIA View
Western policymakers and investors need a category update. They are still evaluating China's AI capability through the lens of "can they build GPT-5?" when the more relevant question is "can they build the world's best AI-integrated car, battery, drone, and factory?" The answer to that second question is increasingly: they already have.
This does not mean China's approach is without risk. Scaling AI across millions of vehicles and factory lines raises serious safety and quality challenges. The recent rollbacks of agentic AI in consumer phones show that Beijing's deployment-first approach can outrun its own guardrails✦. And while China's foundation models are advancing rapidly through labs like DeepSeek and Alibaba's Qwen, the frontier models from OpenAI and Anthropic still lead on many capability benchmarks. China is building both, but the emphasis matters.
The strategic takeaway is clear. The AI race is not a single competition with a single finish line. It is multiple races across multiple industries. The West is leading in the race to build the smartest general-purpose AI. China is leading in the race to build the most AI-integrated economy. Both matter. But if you had to bet on which approach generates more economic value over the next decade, the smart money is on the country that is embedding AI into everything its people buy, drive, and use, not the one building ever-larger models and hoping the applications will follow.
The question is not whether China can build a better ChatGPT. The question is whether it needs to.
Frequently Asked Questions
What is China's vertical AI strategy?
China's vertical AI strategy, formalised through the AI+ Action Plan and the 15th Five-Year Plan, prioritises embedding artificial intelligence into specific industries like automotive, battery manufacturing, drone technology, and telecommunications. Rather than focusing primarily on building general-purpose foundation models, China channels AI investment into sector-specific applications that improve industrial output, product quality, and operational efficiency at scale.
How does BYD use AI in its vehicles?
BYD has deployed its God's Eye ADAS system as standard across its entire vehicle lineup, from the $10,000 Seagull to premium models. The system uses a tiered architecture with up to triple-lidar configurations, NVIDIA Orin X chips, and DeepSeek AI integration for real-time driving decisions. With over 2.5 million vehicles generating 150 to 160 million kilometres of data daily, BYD's fleet creates one of the world's largest real-world autonomous driving datasets. Beyond driving, BYD uses AI to reduce battery manufacturing defects by 40%.
How does China's AI approach differ from the United States?
The US AI ecosystem is built around horizontal, general-purpose foundation models where companies like OpenAI, Anthropic, and Google build large models and sell API access. China's approach is vertical: industrial companies embed AI directly into their products and processes. US policy focuses on export controls and safety regulation. China's policy focuses on deployment targets (70% AI penetration in key sectors by 2027) and open-source model distribution to accelerate industrial adoption. Both countries build advanced models, but they prioritise different outcomes.
Which Chinese companies exemplify the vertical AI approach?
Key examples include BYD (AI-integrated vehicles), CATL (AI-driven battery design and manufacturing), DJI (onboard AI for commercial drones), Huawei (Ascend AI chips and industrial cloud), Alibaba (Tongyi Qwen integrated across consumer products), and Tencent (Hunyuan models embedded in WeChat ecosystem). Each company uses AI to dominate a specific industry rather than competing as a general-purpose AI provider.
Does China's vertical AI strategy mean it is falling behind in foundation models?
No. Chinese labs like DeepSeek, Alibaba's Qwen, and Baidu's ERNIE have made significant progress, and DeepSeek's R1 model rivalled OpenAI's output at a fraction of the development cost. China's open-source model strategy has accelerated ecosystem adoption across its industrial base. The strategic difference is that China treats foundation models as infrastructure supporting industrial deployment, while the West treats them as the primary product. Both approaches carry distinct risks: China risks deploying AI faster than safety mechanisms can manage, while the West risks building powerful models that transform industries too slowly.
Sources: World Economic Forum, "Blueprint to Action: China's Path to AI-Powered Industry Transformation" (2025); USCC, "How China's Open AI Strategy Reinforces Its Industrial Dominance" (March 2026); CSET Georgetown, China AI+ Action Plan translation; USSC, "Intelligent Everything: China's Policy to Supercharge AI Adoption"; CATL official announcements; DJI Enterprise developer documentation; CarNewsChina; Carbon Credits.
No comments yet. Be the first to share your thoughts!
Leave a Comment