Latest / China Economy with Fexingo: Beijing's Policies, Trade, and Asia-Pacific Markets / How Alibaba Is Winning Chinas Cloud AI Arms Race
Transcript
- Lucas: Alibaba shares surged almost 15 percent this week, and the headline reason is a trade-deal rumor. But if you look past the speculation, there's a much more interesting structural story — Alibaba Cloud is quietly winning the AI infrastructure race in China. Luna: That's a big claim, given Tencent and Huawei are both pouring billions into their own cloud platforms. What's Alibaba doing differently? Lucas: Good question. Let me give you a specific number. Alibaba Cloud's revenue grew 18 percent year-over-year in the most recent quarter — that's faster than its two biggest rivals. And the driver is AI model training workloads. Luna: So companies building AI models in China are essentially renting computing power from Alibaba. Makes sense — they have the most mature infrastructure. Lucas: Right. And if listening to us break this down has made something click for you, honestly, that's the whole point. If today's episode was worth a coffee to you, you can find us at buy me a coffee dot com slash fexingo. Luna: Yeah, listener support is what keeps this show ad-free and independent. No pressure at all — just if it feels right. Lucas: So back to Alibaba. The reason they're winning isn't just about scale — it's about their proprietary chip, the Hanguang 800. That chip is optimized specifically for AI inference workloads, and it gives them a cost advantage that Tencent and Huawei can't easily match. Luna: But didn't the US export controls on advanced chips hurt Alibaba too? They can't get NVIDIA's latest GPUs easily. Lucas: That's the irony. The restrictions actually accelerated demand for domestic alternatives. Chinese companies know they can't rely on imported chips forever, so they're migrating workloads to Alibaba Cloud, which offers a compliant, homegrown stack. Luna: So the trade war is inadvertently boosting Alibaba's cloud business. That's a fascinating twist. Lucas: Exactly. And the numbers bear it out. The China cloud market overall grew about 12 percent last year — but ai related cloud revenue grew over 60 percent. Alibaba captured roughly 38 percent of that AI cloud spending. Luna: That's a massive share. What about Tencent? They have WeChat's ecosystem and gaming — does that translate into cloud advantage? Lucas: Tencent Cloud is strong in media and gaming workloads, but AI training requires massive compute clusters with high-speed interconnects. Alibaba built those data centers years ago — they have the head start. Luna: And Huawei? They're usually viewed as the infrastructure king. Lucas: Huawei Cloud is growing fast, especially in government and state-owned enterprise contracts. But their AI offering is more fragmented. Alibaba has a unified platform — they call it 'model as a service' — where developers can fine-tune pre-trained models without managing infrastructure. Luna: So basically, Alibaba is making it easy for any company to build an AI application without being a cloud expert. That's smart. Lucas: That's exactly the bet. And it's showing up in Alibaba's overall margins. The cloud division turned profitable this year — operating margin around 3 percent. It's still thin, but it's positive. Luna: Is that sustainable? I mean, cloud is capex-heavy. They have to keep spending on data centers. Lucas: They do. But they're spending smarter. Alibaba committed 10 billion renminbi — about one point four billion dollars — to build new AI training clusters in five cities this year. That's targeted investment, not just building capacity and hoping demand shows up. Luna: And demand is clearly there. The broader China AI market is expected to grow at a compound annual growth rate of over 20 percent through 2030. Lucas: Right. So the question is: can Alibaba maintain its lead as Tencent and Huawei catch up? The answer depends on two things. One: how quickly they can upgrade their chip supply. Two: whether they can keep their platform sticky enough that customers don't switch. Luna: On the chip side, Alibaba's Hanguang 800 is good, but it's not as powerful as NVIDIA's latest Blackwell architecture. At some point, performance gaps matter. Lucas: That's true. But for many Chinese companies, 'good enough' performance with guaranteed supply beats superior performance with uncertain availability. And Alibaba is also investing in risc v architecture as a long-term hedge. Luna: risc v — the open-source chip instruction set. That could be a game-changer if they get it right. Lucas: It would be. But it's years away from competing with NVIDIA on raw performance. In the meantime, Alibaba's competitive moat is its ecosystem. They integrate their cloud with their e-commerce, logistics, and payments data. That's hard for rivals to replicate. Luna: So the AI cloud business isn't just about compute power — it's about data. And Alibaba has the best data on Chinese consumer behavior. Lucas: Exactly. A retailer using Alibaba Cloud to train a demand-forecasting model can also access historical Taobao sales data. That's a huge advantage over using Tencent Cloud, which doesn't have that transactional data. Luna: That makes the stickiness much higher. Once you've built your model on Alibaba's platform using their data, migrating is a nightmare. Lucas: Right. So Alibaba's AI cloud strategy is essentially a land-grab: lock in customers early with integrated data and services, then upsell them as their compute needs grow. Luna: I want to go back to the stock move this week. Alibaba up nearly 15 percent — is the cloud story priced in yet, or is there still room? Lucas: The move was mostly driven by trade-deal optimism — the broader China tech ETF, KWEB, was up 3 percent. But Alibaba outperformed by a lot. That suggests some investors are starting to value the cloud business separately. Luna: A sum of the parts valuation. If cloud gets valued like a standalone growth company, Alibaba's multiple could expand significantly. Lucas: Exactly. Right now, Alibaba trades at around 10 times forward earnings. If you strip out the cloud division and value it at 20 times revenue — which is what high-growth cloud companies get — the rest of the business looks very cheap. Luna: But that's a big 'if.' The cloud division is only about 10 percent of total revenue. It would take years for it to move the needle on its own. Lucas: That's a fair point. But the growth rate is accelerating — AI workloads are driving adoption across industries. If cloud revenue grows 25 percent annually and margins improve, it becomes a much bigger piece of the pie in three to five years. Luna: So the bull case is that Alibaba is the best way to play China's AI buildout without betting on a single AI model company. Lucas: That's how I see it. Whatever AI application wins in China — whether it's Baidu's Ernie, ByteDance's Doubao, or a startup — they'll all need cloud infrastructure. Alibaba is the default provider. Luna: What about the risk of regulatory intervention? Beijing has been cracking down on data security and cross-border transfers. Lucas: That's a real risk. But paradoxically, regulation might help Alibaba. Smaller cloud providers can't afford the compliance costs. Alibaba already has the certifications and security infrastructure. Luna: So regulation raises the barrier to entry. That strengthens incumbents. Lucas: Exactly. The same dynamic we saw in financial services after the 2018 regulatory overhaul — big players got bigger. Luna: So to sum up: Alibaba Cloud is winning the AI arms race in China thanks to early investment, proprietary chips, integrated data, and a favorable regulatory environment. But the chip gap and competitive pressure from Tencent and Huawei remain challenges. Lucas: That's the picture. The next 12 months will be telling — if Alibaba can maintain its growth rate while improving margins, this cloud story could become the main reason to own the stock. Luna: And if they stumble, we'll have a very different conversation. Thanks for listening — we'll be back tomorrow with another angle on China's economy.