DJI

Where is AI actually paying off?In this episode of Inside China AI, Gabrielle Chou, Associate Professor at NYU and based in Shanghai, speaks with AI analyst Lucy Tan about five industrial AI use cases drawn from confidential interviews with senior executives at large Chinese industrial groups.The episode looks beyond demos and pilots to ask where AI has produced measurable returns inside real operations: factories, supply chains, leasing platforms, logistics systems, and compliance infrastructure.The key finding: the first ROI often came from the least glamorous work.Predictive maintenance, inventory optimization, quality inspection, route optimization, asset utilization, and compliance systems produced some of the clearest returns. Frontier AI projects were often still in pilot, while operational AI was already showing up in the P&L.We discuss:* Why predictive maintenance delivered some of the fastest payback* How AI ROI compounds when one company owns more of the value chain* Why AI changed the business model in industrial leasing* How compliance AI became a market-access tool* Why China is not a monolith: some firms are early adopters, others are still cautious* Why the real split is not geography, but adoption stage* What CEOs should bring back to the boardroom when evaluating AI ROIThe episode’s central takeaway: the AI that pays first may not be the AI that gets the headline.Inside China AI is hosted by Gabrielle Chou. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit insidechinaai.substack.com/subscribe

Tencent Is Changing What the AI Model Race Is ForHy3 is not just another model launch. It is Tencent’s bet that models become infrastructure — while the real value moves to Yuanbao, WorkBuddy, CodeBuddy, Tencent Cloud, and Weixin.Tencent released Hy3 preview on April 23. But the launch itself is not the real story.In this episode of Inside China AI, Gabrielle Chou speaks with AI analyst Lucy Tan about why Tencent’s AI strategy may be less about winning the model race — and more about changing what the race is for.Hy3 reached #1 globally on OpenRouter’s weekly leaderboard for the week of May 15, with 2.4 trillion tokens served. But Tencent’s deeper strategy may not be model-layer revenue. The more important question is where Hy3 will live: inside the surfaces Tencent already owns.We discuss:Why Hy3 should not be read as just another Chinese AI model launchWhy Tencent is treating the model as infrastructure, not the final productThe significance of “deep co-design” between Hunyuan and Tencent’s productsWhy Yuanbao switching from DeepSeek to Hunyuan mattersHow Tencent differs from DeepSeek, Moonshot’s Kimi, and Z.aiWhy the U.S. model-lab strategy does not fully apply to TencentHow open models can become a funnel for Tencent CloudWhy Weixin may become Tencent’s most important AI surfaceThe model gets the attention. But the surface holds the user.Inside China AI is hosted by Gabrielle Chou. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit insidechinaai.substack.com/subscribe

China AI This Month: Regulation, Capital, Compute, Grey Markets, and MonetizationInside China AI — Monthly Podcast, May 2026The China AI story this month isn't about model releases. The stronger signal is that China's AI market is becoming structured — regulated, funded, protected, priced, and integrated into infrastructure.In this episode, I'm joined by Lucie [last name / role] to unpack the five forces that mattered most this month, and what they mean for executives, investors, consultants, and board members trying to read China's AI economy from the outside.What we cover🏛️ Regulation — China is defining what an AI agent actually is Regulators are moving from governing models to governing agents. What the new definitions mean for product design, liability, and which companies can scale.💰 Capital — State-linked money is flowing to frontier labs The shift from private VC to state-adjacent capital at the frontier. Who's getting funded, who isn't, and what it signals about Beijing's read on which labs matter.⚙️ Compute — Scarcity is forcing architectural efficiency Cut off from the latest Nvidia chips, Chinese labs are competing on efficiency rather than scale. Why this is producing a genuinely different model design philosophy — and why Western analysts keep misreading it.🕳️ Grey markets — Blocked access is producing parallel infrastructure What happens when you ban an API but not the demand for it. The infrastructure layer that has emerged to route Chinese developers around export controls.💳 Monetization — Will Chinese users pay for compute-heavy AI? The consumer AI companies running the live experiment right now. Early signals on willingness to pay, and what it means for unit economics across the sector.Why this mattersFor executives, investors, consultants, and board members, the question is no longer "Which Chinese AI model is best?"The better question is: Where is China's AI market becoming regulated, funded, protected, priced, or integrated into infrastructure?That's the map worth building. Newsletter this week with details that were discussed in the podcast This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit insidechinaai.substack.com/subscribe

At the 2026 Beijing Auto Show, the biggest story was not just electric vehicles. It was AI moving into the car.In this episode of Inside China AI, Gabrielle Chou looks at how Chinese AI is moving from models and apps into hardware — cars, cockpits, appliances, glasses, robots, and other physical products.The central question: when Chinese hardware goes global, does Chinese AI travel with it? Or does it get stripped out and replaced by Google, Apple, Alexa, OpenAI, or another Western platform AI?This episode introduces the idea of the “control layer”: the intelligence layer that interprets what the user wants, decides what happens next, and triggers the action.We discuss:The 2026 Beijing Auto Show and why AI became the new battlegroundWhy Chinese cars are becoming “agents on wheels”Huawei’s role in the AI cockpit and intelligent driving stackWhy Chinese AI may be stripped out of mature categories like phones and smart-home appliancesWhy newer categories like AI glasses and humanoid robots may give Chinese AI more room to travel globallyHow the control layer may define the next phase of AI competitionLast week, we looked at four new Chinese AI models from DeepSeek, Tencent, Alibaba, and Moonshot. This week, we move from models to hardware — and ask who will control the intelligence layer inside the products we use every day.Inside China AI is a podcast by Gabrielle Chou on China’s AI economy and what it means for international business. If you want to read more about this, there is a Inside CHina AI Newsletter on this topic too! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit insidechinaai.substack.com/subscribe

The DJI issue is now a 23-minute audio portrait — same piece, different pace.If you prefer listening, the podcast is in your Substack feed or in the app. And if someone on your team missed the written version, the audio is often the easier forward.This is the first in a new monthly format — The Clever Move — where once a month I take one company that makes the month's argument concrete. This month it's DJI. Next month, someone else. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit insidechinaai.substack.com/subscribe