It’s Time You Take Chinese AI Seriously.

Idea

China’s AI strategy isn’t about catching up to the West – it’s about out-deploying it. While the US chases open-ended research and cloud-based APIs, China’s wiring AI directly into the real economy: across consumer platforms, supply chains, and public services. That real-world focus is already paying off – and it could power the country’s next growth wave, long after exports and property cool down.

Valuations are cheap. Risks are well known. And if even one macro domino falls the right way – a policy shift, a demand recovery, a geopolitics thaw – the upside could be substantial. So I built the China AI Deployment 10 (CHAD-10) investing basket as a targeted way to express that view. These stocks aren’t a moonshot: they’re a leveraged play on one of the most consequential tech shifts in China’s modern economy.

Thesis

• China’s AI strategy is all about real-world deployment. While the US builds frontier models and APIs, China is embedding AI into apps, devices, EVs, and services – fast and at scale.
• China’s tech industry has the tools to do it: world-class talent, state funding, and huge investment in local chips and cloud. Strong government-and-company coordination and interlinked sectors like EVs and robotics make deployment seamless.
• Chinese firms are vertically and horizontally integrated. Vertically, they control the entire AI stack: infrastructure, models, and applications. And horizontally, they operate across multiple industries: EVs, smartphones, and apps. That allows them embed AI more thoroughly, grow faster, and monetize it all across products.
• The CHAD-10 is China’s answer to the Magnificent Seven. It’s a highly investable basket of ten firms powering the country’s AI push.
• Risks like chip bans and economic troubles are real – but they’re largely priced in. With low valuations and investor apathy, this could be a high-upside way to ride China’s structural AI story.

Key risks

• US export bans are limiting China’s access to advanced chips, and the local alternatives aren’t as good.
• Future changes in policy could rattle the industry from time to time. China itself has been known to lay down restrictive tech rules.
• The economic picture isn’t that great: consumer demand, the property market, and overall growth have been lukewarm, at best.
• Strict content controls have limited the global “intelligence” of China’s AI models.
• Plus, the money-making picture is less than even: consumer AI has been growing, but enterprise uptake has been patchy, and margins thin.

China isn’t trying to beat the US at AI. It’s playing a different game altogether – and moving faster. While America pours billions into bigger models and frontier research, China’s been keeping its eye on a broader prize: deployment.

That difference is fundamental. In the US, a handful of giants like OpenAI, Anthropic, and Google DeepMind are competing to build general intelligence and sell it back to the world via APIs and cloud services. In China, a legion of firms are wiring AI directly into the economy – powering logistics, public services, retail, and even factories. The goal is less moonshot, more ground cover.

Instead of chasing one model to rule them all, Chinese firms like Alibaba, Tencent, and Baidu are building entire ecosystems – with control over the data, infrastructure, and distribution. With over a billion users on super apps like WeChat, Alipay, and Meituan, these firms are creating tight AI loops, with faster iteration, simpler integration, and speedier monetization.

At the same time, China is emerging as an AI talent hub. Nearly half the world’s AI researchers are Chinese – backed by university pipelines and state-funded PhD programs. More than half of AI patents are held in China. And while US chip bans still sting, China’s model performance is catching up. Tencent’s latest model just matched Meta’s Llama 3, using just 10% of the compute. And DeepSeek’s new open-source model rivaled OpenAI in math-heavy benchmarks.

This contest isn’t just for the tech giants. Over 400 models have already been built in China, and over 300 of those are already cleared for commercial use. Many of them are open-source, which gives smaller firms and developers a shot at building with real firepower – and that’s a sharp contrast to the West’s walled-off, API-first systems. As a result, a wave of experimentation has been unleashed, sparking AI adoption across industries, not just within the BAT (Baidu, Alibaba, Tencent) trio.

At the same time, the AI infrastructure buildout is continuing at pace. Cloud giants are spending about 400 billion Chinese yuan ($60 billion) a year on domestic chips, regional semiconductor hubs, and national compute platforms for smaller players. That’s much more than their US rivals are plunking down. Even the power grid in China is getting an AI makeover, with new data center zones popping up in hydro- and solar-rich regions. So while the US battles energy bottlenecks, China’s building capacity to spare.

The government’s fingerprints are everywhere – and that’s the point. In China, AI isn’t just a tech trend – it’s state policy. Local governments are deploying models in hospitals, schools, and public services, while the national government is pumping money into state-owned compute infrastructure and planning a venture fund worth an equivalent of $139 billion. With exports and the property market no longer driving growth, AI has become the stimulus plan.

The tech’s rollout is already showing up in the numbers. AI is being embedded into smart driving systems (projected to reach 25% of new cars by the end of the year), ecommerce, supply chains, and even a Chinese government app. Gains in productivity, retention, and profit margins are already showing up – with GenAI adoption driving growth of 8% to 25%, depending on the industry. And according to Morgan Stanley, the broader AI opportunity – all-in with infrastructure and services – could top $1.4 trillion by the end of the decade.

There’s another overriding reason why China might actually be better placed than the US to adopt AI at scale: the deep interlinks of the country’s industries. Unlike in the West, tech firms in China tend not to focus on one niche – they operate across a web of connected sectors like EVs, batteries, robotics, smartphones, and AI. That creates a compounding loop: progress in one area speeds the others. Take Xiaomi, which parlayed its smartphone supply chain into smart appliances and now EVs. Or BYD, which makes both cars and batteries. That close integration means Chinese firms can move faster, cut costs, and turn AI into real-world products more easily – with much of the tech built and scaled at home.

Chinese companies operate across interlinked industries, stacking advantages upon advantages. Source: Kyle Chan/High Capacity.

Of course, it’s not all sunshine and rainbows. US chip export bans still restrict China’s access to the most advanced graphics processing units (GPUs), limiting an AI model’s training capability and predictive performance. And, yeah, China’s own chips are making headway, but they’re not quite there yet.

The country’s content rules require models to align with “core socialist values” – which can hamper creative or open-ended tasks like search and ideation. Plus, most Chinese models are trained on domestic-language data, making them less useful abroad. And even though consumer platforms are seeing strong AI returns, enterprise money-making is still case-by-case.

For investors, it’s not about picking a side – it’s about understanding what each side does best. The US still leads in frontier innovation, but China is showing it can scale and deploy at speed. Both strategies create value, and both carry risks. Holding exposure to the two ecosystems isn’t just diversification – it’s a smart way to play the full AI stack while hedging against political, technical, or money-making uncertainty. In a game that’s still being shaped by policy and infrastructure, playing both sides can be a strategic edge.

Like its US counterpart, China’s AI webbing is constructed of three layers: infrastructure, models and platforms, and applications. And there are big players – and some promising niche ones – across each.

Layer 1: Infrastructure

AI needs a physical backbone – chips, data centers, and cloud systems – and in China, that layer is under a ton of pressure. US sanctions have cut off access to Nvidia’s top-tier AI chips, forcing Chinese firms to go local. Huawei has stepped up with some powerful new systems. Cambricon and Hygon have been rolling out replacements for Nvidia’s chips. And other domestic firms have been churning out memory and chipmaking tools. But the gaps are still there – China lacks the most advanced manufacturing gear, and its own chip software isn’t quite ready for primetime.

That hasn’t stopped the buildout, though. Alibaba, Tencent, and others are pouring money into data centers, and operators like Huawei, GDS, and VNET are doing much of the heavy lifting. Even Lenovo – once just a PC brand – has become a rising force in enterprise AI hardware and edge computing.

And that’s just the tip of the iceberg – dozens of smaller, fast-moving startups like Enflame and Black Sesame are racing to build next-gen AI chips and infrastructure, with strong government backing and growing market share.

Layer 2: AI models and platforms

In China’s stack, the platform layer is where the real scale kicks in. Sitting between the chips and the end-user apps, it’s built to achieve three things: training foundation models, hosting them in the cloud, and giving developers the tools to build AI into their own systems.

What sets Chinese platforms apart is their level of control. Alibaba, Tencent, and Baidu don’t just build models – they run them on proprietary cloud infrastructure and plug them into massive digital ecosystems. That gives them the trifecta: scale, speed, and data. Unlike US counterparts that mostly make money through APIs, Chinese platforms can immediately deploy models across messaging, payments, commerce, and industrial workflows – collecting data, tightening feedback loops, and pushing upgrades faster.

Alibaba is doing this through its Qwen models, trained and deployed via Alibaba Cloud. The platform serves internal products like Taobao and Alipay, as well as external developers building retail and enterprise tools. Think of it as AWS hosting and training its own foundation models, then embedding them straight into Amazon and Venmo-style apps – all tailored for the Chinese market.

Tencent is taking a more consumer-first route. Its Hunyuan models run on Tencent Cloud and are being woven into WeChat’s massive ecosystem – everything from ads to content to customer service. While external developers can tap in via APIs, the real magic is in Tencent’s ability to deploy AI instantly across its own apps.

Baidu’s approach, meanwhile, is more industrial. Its Ernie models are embedded in Baidu Cloud and aimed at enterprise and government sectors – think smart cities, manufacturing, and transportation. Rather than chasing flashy consumer features, Baidu is focused on building adaptable, high-performance tools for complex real-world use cases.

A few niche players are also worth flagging. iFlytek is doubling down on speech and language models, aimed at education and healthcare. Meanwhile, DeepSeek is making waves as China’s open-source AI darling – boasting GPT-4-level reasoning but requiring a fraction of the computing power. It’s early days, but its lightweight, developer-friendly models are already helping democratize China’s AI infrastructure.

Layer 3: AI applications

Layer 3 is where China’s AI story gets real (and potentially lucrative). It’s where the focus shifts from building the tech to actually using it. That means turning everyday apps into faster, smarter, more automated versions of themselves.

On the consumer side, China’s mobile-first ecosystem has a huge head start. Super apps like WeChat, Taobao, Meituan, and Douyin are already entwined in people’s daily routines, so layering in AI – for search, recommendations, or customer service – is seamless. Tencent’s Yuanbao, Alibaba’s Taobao AI, and Meituan’s Long Cat are rolling out AI as core utilities, not just novelties. ByteDance’s Doubao, Bilibili’s generative content tools, and DeepSeek are scaling quickly too. And with agentic AI – autonomous bots that operate across apps – already on the horizon, platforms like WeChat could become launchpads for the next big leap. Meanwhile, Zhihu, Kuaishou, and other platforms are testing their own AI-powered content tools – but the big question is whether all this engagement will translate into actual revenue.

Enterprise AI is heating up, too. Adoption is still early, but it’s moving fast. Companies like Baidu, iFlytek, Beisen, Kingsoft Office, and Fourth Paradigm are embedding GenAI into productivity tools, analytics, and workflow automation. But China’s enterprise buyers are famously frugal and prefer open source, so vendors will need to prove their AI tools save time or cut costs. The winners will be those with proprietary data and deep industry expertise.

Then there’s industrial AI – less flashy, but potentially enormous. Baidu, Huawei, Foxconn Industrial, and others working with state-owned firms to build AI systems for factories, logistics, and smart cities. It’s a tricky space – lots of bespoke deals, slow scaling – but the productivity payoff could be huge.

Meanwhile, edge AI is becoming a key feature in devices. Xiaomi, Roborock, DJI, and Midea are building AI into appliances, and letting them process data locally for faster, smarter responses. It’s opening the door to new use cases in smart homes, energy management, and offline workflows.

Mobility and robotics are moving, too. XPeng and BYD are putting AI into smart cockpits and self-parking systems. And humanoid robotics – led by UBTech – are finally crossing over from lab demo to deployment, thanks to local supply chains and policy support. And the impact – from services to manufacturing – will be massive.

Finally, AI is starting to reshape critical sectors like education and healthcare. Firms like TAL Education and Alibaba Health are using it for diagnostics, tutoring, and virtual consultations. These aren’t mass-market moves yet – but they tick some key boxes for policymakers, and the potential disruption is real.

Part 3: Let’s call it China’s “CHAD-10” basket.

Now, let’s talk about what’s actually investable in China’s AI boom. Some of the players here are privately held, or trade only on the mainland exchanges, which most foreign investors can’t access. And plenty are still long on hype, but short on profit. So I’ve narrowed the pool to just ten names – all listed in Hong Kong or accessible as American depositary receipts (ADRs) – each of which offer clear AI upside, scale, and tradability.

Let’s start – carefully – with infrastructure.

AI needs it, but investing in it is no cakewalk. Key players like Huawei and Cambricon are off-limits, and the listed names – like SMIC or Lenovo – are either not pure plays or come with weak margins or hot geopolitical risk.

Still, I wanted at least one infrastructure play, and GDS Holdings made the cut.

#1: GDS Holdings (9698.HK)The firm runs the data centers that power AI training and deployment. It’s scaling fast, works with China’s top tech firms, and is Hong Kong-listed. Not exactly Nvidia, I know, but a solid way to get AI infrastructure exposure, nonetheless.

Next, let’s move to where the real value lies: models, platforms, and applications.

For the rest of the CHAD-10, I looked for companies doing more than just talking a big AI game. And that meant scanning for these things:

• Ecosystem control. I needed firms that sit at the center of a big, sticky ecosystem – whether it’s consumer super apps, cloud platforms, EVs, or robotics. They’d need built-in distribution, rich data loops, and a real path to monetizing AI at scale.
• AI on the profit and loss. Forget hype: I wanted firms that are already seeing AI gains in productivity, revenue, or margins.
• Strategic moats. Each company would have to be aligned with China’s tech priorities – with control over data, compute, or models. That hands them staying power as the ecosystem scales.

That lens helped cut through the noise – ruling out early-stage hopefuls, the hard-to-own, and the all-sizzle-no-steak names.

What I had left was a group of big, liquid companies with serious scale, efficient ecosystems, and clear AI upside. Here are the stocks that stole the nine remaining spots and why…

Platform + Ecosystem Leaders

#2: Tencent (0700.HK)Think of Tencent as Meta plus PayPal, built into one super app. The company’s WeChat platform has over a billion users and blends messaging, payments, maps, and loads of mini-apps – all in one place. That gives it one of the richest data sets in the world. Gaming is still Tencent’s bread-and-butter, but it also whips up revenue from ads, fintech services, and cloud. And these days, it’s layering AI across everything: smart replies in chat, targeted ads, personalized content, and enterprise services via Tencent Cloud. With devoted users, strong money-making streams, and rigid control of key interfaces, it’s a slam dunk for the CHAD-10.

#3: Alibaba (9988.HK / BABA.US)If you merged Amazon, AWS, and PayPal all into one empire, you’d get something that looks like Alibaba. The firm’s core ecommerce platforms (Taobao, Tmall) drive massive retail volume, while AliPay and its Ant Group affiliate dominate digital payments. Logistics and cloud (via Cainiao and AliCloud) round out the stack. And now Alibaba is embedding AI across the whole thing, all while selling AI infrastructure via AliCloud. That dual exposure – serving AI users and providers – makes it one of the most leveraged names in China’s AI stack, and a must-have for this basket.

#4: Baidu (9888.HK / BIDU.US)Baidu is still best known as China’s Google – and, yeah, like Google, it’s much more than a search engine. Sure, search and advertising are its core revenue drivers, but Baidu is deeply invested in AI infrastructure, autonomous driving, and smart cloud services. Baidu is also injecting generative AI across its products, like Baidu Search, Baidu Drive, and its smart assistant platform. At the same time, its Apollo autonomous driving unit is among China’s most advanced, already operating robotaxis in multiple cities.

#5: Meituan (3690.HK)If DoorDash, Uber, and Yelp merged, you’d get Meituan. It dominates China’s local services economy – from food delivery to ride-hailing to hotel bookings. And AI powers the whole operation: dynamic pricing, courier routing, personalized offers, and automation across millions of transactions.

Edge AI + Devices + Mobility

#6: Xiaomi (1810.HK)Xiaomi makes smartphones, sure, but it also makes TVs, wearables, robot vacuums – and now electric vehicles. And it’s embedding AI across all of that – and pricing it for mass adoption. Its EV sedan is already a hit, and those devices generate rich user data, which feeds the AI – in turn, fueling demand for more goods and services.

#7: Xpeng (9868.HK / XPEV.US)Xpeng is one of the few EV players in the world building its entire autonomous driving system in-house – from chips to software and user interface. Think Tesla, but more urban-focused and software-heavy. Xpeng is already piloting AI-powered city driving, and its whole roadmap is geared toward smart mobility.

#8: Li Auto (2015.HK / LI.US)Li Auto is a rising star in the smart EV space, with its plug-in hybrid EVs and a line of all-electric models in the works. Its AI stack powers driver assistance, voice interaction, and one of the smartest in-cabin experiences on the market. It’s not just peddling cars – it’s selling a tech-first driving experience. And that’s paying off: with strong sales growth, solid margins, and a differentiated product roadmap, Li Auto is one of the most promising smart mobility plays in China.

#9: UBTech (9880.HK)UBTech builds humanoid robots for education, security, and customer service – think Boston Dynamics meets Alexa… on wheels. With advances in generative AI, these robots are becoming more interactive and more commercially viable. It’s a pure play on embodied AI, and now that it’s listed in Hong Kong, this company is finally investable. Small-cap and early, but with high potential – this is one of the few ways that you can buy into the robotics layer directly.

#10: Horizon Robotics (9660.HK)Horizon designs AI chips for edge computing – things like autonomous vehicles and smart devices. So it’s essentially Nvidia for China’s edge stack (okay, a bit of an overstatement, but you get the idea). Its chips process data on-device, not in the cloud – which means faster response, lower cost, and more privacy. With major auto partnerships and a recent IPO, it’s been helping to push China’s AI ambitions beyond data centers.

So, that’s the lineup. The CHAD-10 basket focuses on the most accessible, AI-leveraged names in China – from dominant ecosystem platforms (Tencent, Alibaba, Meituan, Baidu) to applied AI in consumer devices and mobility (Xiaomi, Xpeng, Li Auto). It even includes two smaller, earlier-stage bets (UBTech and Horizon Robotics). And sure, firms at that stage can be riskier, but these two offer a rare pure-play exposure to robotics and edge AI.

Zoom out and you’ll probably notice the basket leans heavily on consumer retail, robotics and EVs. And it’s not a fluke: that’s where China’s AI story is already hitting the real economy. It’s also where local players have a real advantage, with control over supply chains, vertical integration, and government backing. And frankly, it’s where monetization looks most promising right now. No surprise that many of China’s biggest tech firms are doubling down on those three product lines.

Part 4: Naturally, there are risks as well as opportunities.

You rarely find opportunities without risks. So here is a look at the risks involved, as well as the potential opportunities.

Okay, so there are risks…

Let’s be clear: there’s still a bit of a gamble here. China’s economy is still fragile. Growth is sluggish, youth unemployment is high, and the property market’s still a drag. Policymakers have been slow to roll out meaningful stimulus, and if that doesn’t change, the recovery could struggle even more. And that’d pose a challenge for consumer-facing companies.

The global picture may not be much rosier: if the US economy slows or interest rates stay too high for too long, emerging market risk – particularly Chinese tech – might not hold much appeal. In fact, any risk-averse turn of events could push investors away.

And the current geopolitical backdrop isn’t doing China’s AI scene any favors. US chip restrictions continue to throw up roadblocks, limiting access to top-tier hardware just as model training ramps up. New sanctions or tighter controls are a constant threat. Plus, Chinese ADRs still face delisting pressure in the US. And back home, the government hasn’t exactly gone soft: regulators are keeping close tabs on data use, platform dominance, and algorithms. And those policy risks could constrain how aggressively firms can turn AI into profit.

Execution is another wild card. Some of the basket’s names (like UBTech, Horizon Robotics, or Xpeng) are early-stage or still burning cash. Most face stiff competition. Their stories rely on scaling quickly and turning hype into hard numbers. And that path isn’t likely to be smooth – or short.

And of course, there’s the big, overriding risk: that the AI boom itself is overhyped – at least right now. Some investors are already questioning whether Chinese firms have simply done a better job selling their story than building the tech. In fact, most of the bounce back since last September has come from multiples expansion – not earnings. For the rally to stick, companies now need to prove they can grow into those valuations.

And, yep, those risks help create the opportunity.

Most of the risks – weak demand, policy crackdowns, chip curbs, and geopolitical storms – are well-known. And that’s why Chinese tech stocks are still down two-thirds from their 2021 highs and remain largely shunned by global investors.

And sure, valuations have bounced since September, and tech has outperformed. Still, most of the CHAD-10 names are still trading at steep discounts to their US peers and global ones alike.

Here’s the rundown: Chinese tech’s median price-to-earnings (P/E) ratio sits at just 17.5x versus 32x for those US rivals, price-to-sales sits at 3x versus 10x, and price-to-book at 4x versus 11x. And yet, expected revenue growth is actually slightly higher – 16% versus 12% next year, and 19% versus 14% the one after. And sure, gross margins tend to be lower (33% versus 50%), but that alone doesn’t explain the valuation gap.

Tap here (or the image below) to open the spreadsheet.

Despite attractive fundamentals and predictions, Chinese stocks trade at a big discount to their US peers – and are way below their previous highs. Sources: Finimize, Koyfin.

In fact, that discount looks less like a bright and shining signal, and more like dull muscle memory. America’s tech names have crushed it over the past five years, so it makes sense that investors might assume that they’ll keep winning. Meanwhile, China’s top AI stocks are still down 67% from their highs – dragged by policy risks, chip bans, and geopolitical heat.

That doesn’t guarantee a turnaround, but it does set up a powerful asymmetry: if the macro picture improves even a little – with stronger consumer demand, new government support, or easing US tensions – these stocks could improve in a hurry.

And with a median beta of just 0.5, Chinese AI stocks don’t move in lockstep with US ones. That makes them a low-correlation play in the crowded global AI trade.

More importantly, this remains a long-term structural opportunity. China’s leaning hard into AI, and these firms are at the heart of that push. And while sentiment isn’t in panic mode, it’s still cautious: economic data is weak, tariff noise is clanging, and investors have put some money on the sidelines. That creates an interesting setup – a long-term growth story trading at a discount, with big fear offering a potential entry point.

To be clear, this isn’t about calling the bottom. It’s about gaining exposure to a long-haul trend while prices are still attractive, leaning in with the help of dollar-cost averaging. If things worsen in the short term, great: you’ll take advantage when the prices dip. If they don’t, that’s fine too: you’re already in for the gains.

—

Originally Posted on June 4, 2025 – It’s Time You Take Chinese AI Seriously.