East Asian Technology Intelligence
Japan & China tech news — translated, contextualized, and delivered often.
Free. Unsubscribe anytime.
3 Takeaways This Issue
- Japan’s METI-backed FRONTia Project secures direct access to Nvidia’s Blackwell architecture through a massive state-funded procurement, positioning industrial heavyweights like Mitsubishi and Toyota to dominate the emerging physical AI and robotics supply chain before Western competitors can scale.
- Alibaba Cloud’s debut of Tongyi Qianwen-powered AI agent earbuds at the World Artificial Intelligence Conference marks a distinct shift in Chinese hardware strategy, prioritizing low-cost, all-day consumer wearables that bypass the GPU-heavy cloud requirements bottlenecking Western enterprise software models.
- Google’s rebranding of NotebookLM to Gemini Notebook signals a rapid centralization of its experimental AI tools into a unified platform to defend its core enterprise productivity suite against specialized search and research startups.
Core Move
NVIDIA CEO Huang: ‘Japan AI is a Must’; METI Launches New Domestic Physical AI Project, Minister Akazawa Wears ‘Leather Jacket’ #
Japan’s FRONTia Project is backed by METI. The project is led by Noetra and AIST. This plan shows the nation is shifting toward physical AI uses. Japan wants to use its factory strengths instead of chasing general AI. This is a classic defensive move by METI. The project uses a large group of 44 companies to build local power. This group aims to solve national problems directly. NVIDIA CEO Jensen Huang praised the move. He said that Japan AI is a must. His words show how much the global tech supply chain needs Japan’s factories.
The goal is to build real-world native AI by 2030. Japan wants to use this tech for aging populations and disaster response. The country is not trying to beat OpenAI or Google on LLM benchmarks. Instead, Japan wants to build a national champion for industrial change. This strategy matches its past success in robots and factory automation. The project plans to build a multimodal model for physical AI. This model will connect with Japan’s deep culture of master manufacturing, known as monozukuri.
Local news from ITmedia AI+ calls the plan a necessary step. The country must protect its industrial power and use its special skills. Minister Akazawa wore a leather jacket to match Jensen Huang’s famous style. This move showed a link with top tech firms. Still, he kept the focus on a national effort. This path differs from Western plans. Western groups often focus too much on abstract models. They often miss the vital link to physical tools and supply chains.
The main threat is the common trap of big groups. Having 44 companies can dilute focus and slow down work. Fusing AI with factories is a good goal. However, leading so many firms under a state banner can lead to weak solutions. It can stop bold new ideas. Japan must ensure this group creates real, specific progress. The project cannot just produce basic theories.
This project shows where Japan wants to compete in AI. We should look for specific project results from Noetra. Early tests in factories or infrastructure will be key. We must also watch the 44 partner firms. We need to see if they send engineers and use the new models. Success will not come from basic model progress. It will come from real AI tools that fix actual factory bottlenecks.
🗾 Japan Radar #
What Japanese media is reporting that Western outlets miss
Japan is surrendering consumer LLMs to Silicon Valley to focus its state-backed AI strategy entirely on industrial, physical systems.
🗾 Policy & Regulation · Cross-Regional Analysis · AI & Machine Learning3 STORIES
Japan Launches Sovereign ‘Physical AI’ Push Powered by Nvidia and Industrial Giants
Backed by heavy government funding and a massive procurement of Nvidia’s next-generation chips, a consortium of 44 Japanese giants—including SoftBank, Sony, and Honda—has launched the METI-supported ‘FRONTia Project’ to build a domestic multimodal foundation model by 2030. This strategic initiative, endorsed by Nvidia CEO Jensen Huang during his high-profile visit to Tokyo, aims to leverage Japan’s traditional manufacturing prowess and ‘monozukuri’ culture to pioneer ‘real-world native AI’ tailored for robotics, hardware, and national societal challenges.
Why it matters: The ‘FRONTia Project’ is a classic METI-backed consortium, signaling Japan’s defensive play in AI. The focus on ‘physical AI’ and ‘real-world native AI’ by 2030 through a 44-company coalition indicates an attempt to leverage Japan’s traditional manufacturing strengths and ‘monozukuri’ culture into a new tech domain, rather than aiming for leadership in general-purpose AI models. The emphasis on unique national challenges like an aging society and disaster response, framed as opportunities, highlights a practical, application-driven approach to AI development, which contrasts with the more abstract model-centric focus often seen in Western AI narratives.
For Western readers: Western industrial robotics, automation, and AI solution providers should understand that Japan’s long-term strategy for critical applications will prioritize domestic AI development and control, making direct market entry challenging and partnership with Noetra or its affiliates essential. 🗾 AI & Machine Learning
Google Renames ‘NotebookLM’ to ‘Gemini Notebook,’ Strengthening Integration into Gemini Ecosystem
Google has rebranded its AI-powered research tool, NotebookLM, as Gemini Notebook to integrate it more deeply into the broader Gemini ecosystem, including the Gemini app and future Google Search AI mode. The tool, which allows users to generate summaries and Q&A based on uploaded documents, now also assigns secure cloud computers to notebooks for complex data analysis, a feature rolling out to Pro users. Google states the core mission of accelerated learning remains unchanged. From a Japanese enterprise perspective, this renaming and deeper integration clarifies Google’s AI strategy, making it easier for companies to adopt and leverage Gemini-branded tools for research and data analysis. The emphasis on source-based AI is valued in environments where data provenance and accuracy are critical, distinguishing it from general-purpose chatbots. Japanese organizations prioritize structured, verifiable AI applications over flashy, unconstrained ones.
For Western readers: Western businesses using Google Workspace or considering AI research tools should expect a more cohesive and powerful Google AI offering, making deeper integration of Gemini Notebook into enterprise workflows an immediate opportunity.
🇨🇳 China Watch #
China’s technology moves, framed for Western readers
China is shifting AI deployment from abstract software to consumer-ready physical hardware, embedded agents, and state-directed networking infrastructure.
AI & Machine Learning
Alibaba Cloud debuted AI agent earbuds powered by its Tongyi Qianwen large language model at the World Artificial Intelligence Conference (WAIC) in Shanghai. These earbuds offer real-time translation, meeting transcription, and health monitoring, positioning Alibaba directly in the increasingly crowded wearable AI market. The device aims to integrate AI directly into daily life through a discreet, always-on form factor. Chinese tech giants like Alibaba are not just developing large language models; they are rapidly integrating them into consumer hardware to capture market share and user data. This move is less about a single product’s technical breakthrough and more about Alibaba’s aggressive strategy to control the AI application layer, ensuring its Tongyi Qianwen ecosystem extends directly to end-users rather than relying on third-party hardware.
For Western readers: Western hardware manufacturers should expect intensified competition in the AI wearable space from Chinese players leveraging vertically integrated AI stacks, potentially leading to lower-cost and more feature-rich alternatives in Asia before they reach Western markets. Policy & Regulation
30 Entrepreneurs Write to WAIC: What China Expects From AI Development at the 2026 World AI Conference Thirty prominent Chinese technology entrepreneurs and executives have collectively written to the World AI Conference (WAIC) organizers, outlining their expectations and priorities for China’s AI development, ahead of the 2026 conference in Shanghai. Their proposals emphasize the need for robust domestic AI infrastructure, ethical guidelines, and practical applications that drive industrial upgrading. This initiative underscores a concerted effort within China’s private sector to align with national strategic goals for AI self-sufficiency and leadership. The fact that leading Chinese entrepreneurs are making public, coordinated statements about AI strategy, rather than simply responding to policy, indicates a shift towards a more proactive, industry-led dialogue with the state. This points to a maturing domestic AI ecosystem increasingly focused on practical industrial applications over theoretical advancements, a clear move to integrate AI deeper into China’s manufacturing base rather than just its consumer internet.
For Western readers: Western businesses operating in or with China should anticipate that future AI collaborations or market access will increasingly hinge on alignment with China’s domestic AI infrastructure and ethical frameworks, making interoperability and data sovereignty key considerations. Policy & Regulation
China Targets Next-Gen Internet Infrastructure by 2030, Advanced by 2035 China’s Ministry of Industry and Information Technology (MIIT) and three other agencies have issued guidelines to upgrade the country’s internet infrastructure, aiming for “systematic breakthroughs” by 2030 and a more advanced national internet by 2035. The plan includes research into agent-to-agent networks, satellite internet, digital identity, IPv6, and the integration of AI, blockchain, and distributed identifiers. The specific mention of “agent-to-agent networks” and “digital identity infrastructure” points to China’s continued push for a highly controlled and traceable internet environment. This isn’t just about faster speeds; it’s about embedding state-level governance into the very fabric of future network protocols, extending beyond simple content filtering to direct control over digital interactions and identities. It’s a fundamental architectural shift that will impact how data is collected and managed inside China.
For Western readers: Western companies operating in China should anticipate a progressively more insulated and regulated digital operating environment, particularly around data exchange and user identity verification, making cross-border data flows more complex and potentially more restrictive. AI & Machine Learning
miHoYo Launches AI Companion App BSide: Olivia Lin on Steam Early Access Chinese gaming giant miHoYo has launched BSide: Olivia Lin, an AI companion app, on Steam Early Access, shifting from traditional games to AI-powered digital companionship. The app features a virtual character, Olivia Lin, offering piano performances, music creation based on MIDI files, and an AI letter-based chat system. User feedback has been mixed, with some expecting a game and others questioning if the app primarily serves as an AI data collection platform. miHoYo’s move from game development to an AI utility demonstrates how China’s tech giants are re-purposing existing intellectual property and user engagement expertise to develop new AI products. Rather than emphasizing breakthrough AI models, their focus is often on consumer-facing applications that can quickly monetize existing fanbases and content pipelines.
For Western readers: Western companies should recognize that Chinese AI innovation is not solely focused on foundation models; consumer applications like this, leveraging existing entertainment IP and a large user base, are a significant and fast-moving segment. Robotics & Automation
Xiaomi Open-Sources Robotics-U0: A 38B-Parameter Embodied Generative Model That Unifies Four Robot Tasks Chinese electronics giant Xiaomi has open-sourced Robotics-U0, a 38-billion-parameter embodied generative AI model designed to perform four distinct robot tasks. This move marks Xiaomi’s increased commitment to integrating AI into robotics and making its research publicly available. Xiaomi’s focus here is on practical applications for embodied AI in robots, rather than just abstract benchmarks. Open-sourcing a model of this size for unified robot tasks accelerates development within the Chinese robotics ecosystem and provides a foundation for more sophisticated automation solutions.
For Western readers: Western robotics developers should anticipate faster iteration and lower barriers to entry for advanced robotic applications coming from China due to platforms like Robotics-U0, shifting competitive dynamics in practical deployment rather than just academic research.
🔺 The Triangle #
Where US, Japan, and China technology interests intersect
China’s export-driven manufacturing hardware scales to bypass US-led frontier model moats, forcing global supply chains to diversify through Asia.
Semiconductors & Hardware
SEMI Forecasts Global Chip Equipment Sales to Reach Record $229B in 2028 📊 Featured Chart
Source: SEMI Mid-Year Forecast Global semiconductor manufacturing equipment sales are projected to hit a record $229.5 billion by 2028, driven by increasing AI-related demand, according to SEMI. This growth includes significant expansion in wafer fab equipment and back-end processes like advanced packaging, crucial areas for East Asian chipmakers. The focus here is on the continued buildout of tangible manufacturing capacity. While Western media often focuses on AI model advancements, this data shows the real-world industrial commitment behind the AI boom, particularly in areas like HBM and advanced packaging where East Asian firms hold substantial leads. This isn’t just a funding round; it’s physical plant expansion.
For Western readers: Western companies relying on leading-edge chips or advanced packaging services should expect continued tightness in capacity and potential cost increases as East Asian fabs prioritize AI-driven demand, particularly for HBM and advanced logic. Semiconductors & Hardware
TetraMem, SK hynix Highlight Memristor-Based AI Computing SoC Collaboration TetraMem and SK hynix have jointly developed an analog in-memory computing (A-IMC) SoC utilizing memristor technology, aimed at improving AI inference efficiency by reducing data movement. This collaboration, culminating in a published research paper, focuses on energy-efficient depthwise convolution for AI workloads. The initiative combines SK hynix’s memory expertise with TetraMem’s A-IMC platform to advance memory-centric AI computing. While still at the research paper stage, this project indicates SK hynix’s long-term play to differentiate its memory offerings beyond traditional DRAM, targeting the critical energy efficiency demands of AI. For Korean chipmakers, this kind of advanced R&D collaboration helps them stay ahead of Chinese domestic efforts, which are still focused on catching up to current-generation memory technologies.
For Western readers: If you are an AI hardware architect, factor in that East Asian memory manufacturers like SK hynix are actively developing entirely new memory architectures rather than just incremental improvements; these could change system design requirements for future AI accelerators. Semiconductors & Hardware
China’s H1 Export Surge Driven by Packaged Chips, Industrial Robots 📊 Featured Chart
Source: South China Morning Post citing customs data
China’s exports of chips, industrial robots, and automatic data processing machines saw significant growth in the first half of 2026, with chip exports nearing $177.28 billion, a 96% year-on-year increase. The reported chip volume of 179.44 billion units, averaging a dollar per chip, points to a product mix dominated by mature memory, PMICs, MCUs, and re-exports of chips packaged and tested within China. The export figures underscore China’s critical role in the lower-value segments of the chip supply chain, particularly in assembly, testing, and packaging (ATP), which often gets overlooked in Western analysis focused on advanced nodes. It’s a reminder that global supply chains remain intertwined, with China handling a massive volume of chips, regardless of where they were originally fabbed. The growth in industrial robot exports signals Beijing’s continued push to upgrade its manufacturing base and become a more significant player in automation technology.
For Western readers: Western companies sourcing mature chips or industrial automation equipment should factor in China’s growing export capacity and competitive pricing for these specific product categories, recognizing this isn’t about leading-edge fabrication but critical mid-stream supply chain functions. Semiconductors & Hardware
Tesla A15 Chip to be Fabbed by Samsung and TSMC; China Lands Reusable Rocket
Tesla’s A15 chip will be manufactured concurrently by TSMC in Taiwan and Samsung in Texas, with TSMC’s Arizona fab and potentially Tesla’s Texas Terafab as future sites. Separately, China achieved a major aerospace milestone by successfully landing a reusable Long March 10B rocket booster developed by the China Academy of Launch Vehicle Technology. Tesla’s decision to split A15 production between TSMC and Samsung, and potentially move some to its own Terafab, is a shrewd move to diversify foundry risk and exert leverage. It points to where industrial policy meets chip manufacturing, securing capacity and potentially driving down costs. China’s reusable rocket success is a more direct indicator of increasing self-reliance in a critical strategic domain. Don’t expect China to use this new capability to launch Western satellites — this is about Beijing building its own capabilities to project power, not about offering commercial services abroad.
For Western readers: Western automotive and aerospace companies should understand that China’s advancing capabilities in areas like reusable rocketry are primarily for its own strategic and commercial benefit, not a new opportunity for partnership, and that dual-sourcing is becoming the new norm for critical chips. AI & Machine Learning
China’s AI Gap with US Narrows as New Open Model and Nvidia Alternatives Emerge A Chinese startup has launched what is being called the world’s largest open AI model, competing with offerings from OpenAI and Anthropic. This development, coupled with traction gained by Chinese alternatives to Nvidia, has contributed to a decline in AI and semiconductor stock prices globally. Chinese President Xi Jinping is concurrently positioning China as an AI partner for developing nations, emphasizing an open-source strategy. A Chinese startup launching a foundational model that impacts global stock prices is not just about a technical leap; it reflects China’s progress in closing gaps in advanced AI development, not just imitation. The focus on open-source and positioning China as an AI partner for developing nations is a clear play for market and influence beyond domestic shores. This is industrial strategy in plain sight.
For Western readers: Western AI developers and hardware suppliers should reassess the timeline for China’s indigenous AI capabilities, particularly regarding open models and alternative hardware, as domestic solutions are gaining real-world traction and market impact. 🧩 Pattern This Issue
Japan: METI launches FRONTia project to anchor Nvidia chips in physical AIChina: Xiaomi open-sources Robotics-U0 to standardize embodied generative model tasksPolicy: China targets next-gen internet infrastructure for physical AI by 2030
East Asian industrial policy is rapidly shifting from digital LLM benchmarking to physical and robotic AI integration, meaning Western software-only platforms risk losing control over the physical execution layers of the global economy.
AsiaAI.FYI · Written by Dick Weisinger ·