Aug 13, 2025

Last month, I had the chance to lead a group of CIOs, CDOs, and CTOs in China on a Learning Expedition to explore the latest advancements in Artificial Intelligence. Over the course of five intense days, we opened doors to some of the world's leading AI companies— Tencent Baidu, Inc. ByteDance , and SenseTime 商汤科技 Insider Patrice Nordey is Founder & CEO of TRAJECTRY. TechNode Insider is an open platform for subject experts to discuss China tech with TechNode's audience. We also engaged with founders and CEOs of rising AI-powered startups and unicorns such as Versa Inc. Tezign 特赞 Akila PhotoG Chikka.ai and more. Throughout the journey, we experienced remarkable technology firsthand—from smart home AIoT to AI applications in automotive with Xiaomi Technology Beyond the content itself, I am also very proud to have embarked several legacy brands with us— Hermès CHANEL LVMH Kering L'Oréal Cartier , and Louis Vuitton . Together, we explored how frontier technologies could reinvent the luxury industry. Below, are a few takeaways from this remarkable business trip: China is not testing—it's scaling We didn't just meet innovative AI companies; we witnessed China executing a comprehensive “AI infrastructure” strategy. From hardware to application layers, AI is treated as a national priority and developed industrially across sectors—banking, healthcare, supply chain, retail, automotive, manufacturing, software, and hardware. You name it, there is an AI plan for everything! Just as China scaled railways, airlines, telecoms or energy infrastructures, the country is now systematically building an artificial intelligence industry . The contribution of the companies we met— SenseTime (AI as infrastructure), Lenovo (AI for hardware), ByteDance (Volcano Engine platform), among others—is absolutely critical. What many Western companies are only starting to realize is that they're not merely competing against individual Chinese firms, but against an entire coordinated ecosystem. Cities are also racing to lead on AI This is not simply a country-level effort— it's a competition among Chinese cities . Shanghai is investing heavily to become one of China's AI epicenters. We were amazed to see AI integrated into the Shanghai urban planning. Entire districts are dedicated to AI academic research, startup incubation, and corporate investment. For instance, we visited the West Bund AI Valley —a 1.2 million m² district housing over 250 leading AI companies—along with Zhangjiang AI Island , China's first “5G + AI” full-scenario commercial demonstration zone, and Zhangjiang Robot Valley , a 3.9 km² innovation hub dedicated to robotics and AI. Together, they are vivid examples of how the city of Shanghai is actively cultivating world-class AI clusters. DeepSeek is upgrading everything There is clearly a “before” and “after” DeepSeek. As my friend Tianyi Cai, the founder of Versa Inc. told us: “DeepSeek is the best—because thanks to their open-source model, I can leverage the USD 300 million they invested in building it. Then I made it my own, trained it, and personalized it.” Even China's largest tech giants—such as Tencent, which developed its own LLM (Hunyuan)—are using customized versions of DeepSeek's open-source model to meet their own needs. As I wrote previously on Linkedin , the massive adoption of DeepSeek across companies in the country feels like the entire Chinese internet ecosystem received an overnight upgrade. AI-driven retail is the end of channels We met with the IT head of 上海百秋 Buy Quickly , a leading e-commerce service provider managing over 7 million SKUs for premium and luxury Western brands like Nike, Cartier, Lemaire or Bottega Veneta. Their AI roadmap already powers backend operations customer service , and predictive analytics On Romomo , their dedicated live-streaming platform powered with 230 professional live-streamer hosts, AI drives real-time sales management and customer engagemen t. We realized the future of retail isn't about channels anymore—it's about creating stages for seamless AI-orchestrated experiences. AI agents are replacing entire marketing teams Disruptive as AI is, it's already yesterday's story. AI agents are the future! At Tezign 特赞 , we tried atypica.AI , an autonomous marketing research agent that creates go-to-market strategies using AI-generated consumer personas. At Chikka.ai we experienced a conversational agent capable of interviewing real consumers in any language with voice-to-voice AI. and LiblibAI showed us how anyone can produce designer-quality visuals and videos at scale without Adobe tools. demonstrated how AI agents can run marketing workflows that replace entire teams—and how, in this new era, success is no longer just about great content but about outsmarting the social media algorithms of RED or Douyin, much like in the early days of SEO. The AI Agent Economy is just around the corner Agentic development is accelerating so rapidly that it's no longer a question of if, but when. Soon, individuals will equip themselves with personal AI agents Finance agents negotiating rates and trading stocks in real time. Travel agents booking trips and coordinating plans with friends' agents. Shopping agents monitoring preferences and managing wish lists. Wellness agents interpreting diagnostics, scheduling appointments, and recommending routines. Companies are not yet ready to interact with these personal consumer agents acting as proxies. But we can imagine that, the same way a consumer must first interact with a chatbot when contacting his/her bank, companies will have first to interact with AI personal agents before being to reach out directly to their customers. A scary, but very likely scenario. At the WAIC – World Artificial Intelligence Conference last week, I met startups already building new protocols for multi-agent communication. For sure a great area to focus on! General-purpose robotics is the next auto revolution When it comes to disruptive innovation in China, much of the world's attention is on the automotive sector—long in a catch-up phase, and now suddenly leading in new energy vehicles (NEVs). But another, equally profound shift is underway: the rise of the general-purpose humanoid robot market. Thanks to AI, robotics has received a major boost. Machines can now not only execute tasks but also think, learn, and adapt . Industrial robots are already widely deployed around the world, performing specialized tasks at scale and with remarkable precision. So why focus on human-like robots? For a while, I didn't see the point—until I understood a few fundamentals. For any major industry to emerge, two conditions are essential: standardization and scale . Think about the car industry without the Model T, or the internet before the World Wide Web in 1995—it's hard to imagine widespread adoption without a standard platform. Robotics is no different. To reach mass production, reduce costs, and develop a robust supply chain of component makers and software providers, we need robots of similar shapes and functions capable of performing a wide variety of tasks. But why a humanoid form? Because our entire world—tools, infrastructure, living spaces—has been designed for humans. This makes the human-like shape the most versatile design for deployment across manufacturing plants, hospitals, homes, and even public spaces. Alongside more well known companies in the West like Tesla's Optimus, we met impressive Chinese robotics firms such as Kepler Robotics and Fourier , both preparing to lead this new industrial wave. Many experts predict it will be as transformative—and as disruptive—as the automobile revolution itself. Creativity got a boost With GenAI tools fast-tracking the media world into an era of industrialization, it's natural to wonder: how will genuine, original artwork survive? To explore this, we met a former architect from the Zaha Hadid team, now working as an AI architect and artist. Hearing his story was eye-opening—it showed how these new “tools” can actually expand and deepen creative possibilities . AI enables the blending of art forms that once required entirely different types of artists. Today, a single creator can bring an idea to life as an animated image, a screenplay, a feature film, a poem, or even a physical 3D-printed sculpture—all from the same vision. And contrary to the misconception, mastering AI tools to produce compelling art requires real skills, creative direction, and artistic sensibility. Food for thought: It will take time for brands and agencies to collaborate with AI creators, and for consumers to fully embrace this shift. Current debates center on whether AI-enhanced or AI-produced content should be labeled. But perhaps the better question is—how different is this from the Photoshop-retouched photos in magazines and advertising that no one questions anymore? Luxury can have its AI moment How can an industry founded on craftsmanship, uniqueness, and scarcity embrace AI? That was a central question as we brought luxury executives on this journey. Luxury brands may not be early adopters, but I was surprised by the maturity and curiosity of our participants. While GenAI quality isn't yet fully aligned with luxury standards, it's improving rapidly. We explored how AI solutions can support retail clientelling, selling ceremonies, service delivery, and experience consistency at scale. At ContaAI , we saw AI-powered CRM solutions helping boutique managers to engage clients in personalized conversations. At Kivisense | AI, XR, AR, 3D MarTech , we tested nearly photorealistic virtual try-ons for watches, jewelry, bags, shoes, and apparel—experiences crafted sur-mesure. Hyper-personalized services and the overall future of experience will definitely reach new levels thanks to GenAI and Agentic AI. AI tokens are the new currency As an economist, my final reflection centers on business models, value creation, and distribution. In conversations with various companies about their billing methods and cost structures, one recurring theme kept surfacing: the “AI token” as the fundamental unit of value. In the AI context, a token can be thought of as a single unit of AI output—powered by GPU processing, electricity, and data—making it an ideal currency for an AI-driven economy. For example, at OpenAI, a “token” roughly equals ~4 characters or 0.75 words. Today, we are already extensively using AI tokens to trade, consume, share, and monetize AI capabilities. We can roughly estimates that 60–80 trillion tokens per month are likely in motion across the entire global AI ecosystem. But AI tokens are not just a digital currency, they are enabling new business models, such as AI-as-Infrastructure (e.g., SenseTime selling AI compute power in China from its 20,000 Nvidia GPU facilities assembled before the U.S. export ban) or AI-as-a-Service models (e.g., the billing model used by most AI startups). In this respect, Sam Altman (OpenAI) and Jensen Huang (NVIDIA) share a converging vision of the future Equally intriguing is the rise of AI crypto tokens , from the convergence of blockchain and AI tokens, now collectively valued at over USD 20 billion—a market worth watching closely. And this is only the beginning. We can expect not just new value chains but also market capitalizations increasingly shaped by the volume of AI-tokenized assets held by publicly listed companies. It's astonishing to witness an entire economy in the making.