East Asian Technology Intelligence
Japan & China tech news — translated, contextualized, and delivered weekly.
Weekly. Free. Unsubscribe anytime.
3 Takeaways This Week
- Taiwan’s decision to allow up to 60% foreign ownership of satellite operators is a direct bid to lure SpaceX’s Starlink to the island to secure resilient, war-hardened communications infrastructure independent of undersea cables.
- Moonshot AI’s successful launch of its Kimi K3 model proves that select Chinese startups can still achieve frontier-class performance despite U.S. export controls on Nvidia’s H100s by aggressively optimizing domestic hardware.
- The U.S. imposition of new 10% to 12.5% tariffs on Japan and China forces Japanese electronics conglomerates to rapidly shift their semiconductor supply chains out of mainland manufacturing hubs to retain access to the American market.
Core Move
The Shock of 24x AI Token Consumption: The Cost Misconception to Avoid for Production Deployment #
Global token use is projected to grow 24-fold by 2030. This growth shows that Japanese firms are moving past basic AI pilots. They now face the real costs of full-scale production. Western media still focuses on benchmark tests and new models. In contrast, Japanese engineers now focus on task completion cost. This means the total cost to run a business process, not just the price of an API call.
This change shows the practical nature of Japanese industry. Firms there treat new tech as an optimization and quality task. For foreign observers, this shift is like the Total Cost of Ownership focus of the 1990s. Now, firms apply that idea to tokens and hybrid cloud systems. Japanese enterprise IT relies heavily on system integrators and legacy mainframes. These firms cannot simply use public cloud APIs without risking high costs and data leaks.
Domestic media does not see this token growth as a triumph. Instead, they see it as a looming governance crisis. Firms now want to build hybrid systems. These systems send easy tasks to cheap, local models. They save costly top-tier models only for hard tasks that need deep reasoning.
Yet, this practical path carries a major risk. Deep risk aversion may lead to slow decision-making during system design. Firms might fall behind on rollout speed while trying to build the perfect cost framework. Historically, Japanese firms have struggled with the fast, iterative rollouts that AI needs. They prefer to wait for standard best practices, but those do not exist yet in the fast-moving LLM space. If firms spend two years building guardrails before launching features, they will save on token costs but lose market share.
To see if Japanese industry succeeds, look at system integrators like Fujitsu and NTT Data. Watch how they set up their AI contracts over the next year. Look for fixed, outcome-based pricing rather than variable token billing. Also, track how fast Japan’s factories adopt hybrid platforms like Cloudera or Red Hat OpenShift. This adoption rate will show if firms are moving workloads off public APIs to control their own token pipelines.
🗾 Japan Radar #
What Japanese media is reporting that Western outlets miss
US tariffs and satellite shifts force Japan to move past model hype and secure resilient, cost-effective industrial AI deployment.
🗾 Enterprise & Cloud
The Shock of 24x AI Token Consumption: The Cost Misconception to Avoid for Production Deployment
This article from Cloudera advises Japanese businesses to shift their focus from mere token cost to a comprehensive “task completion cost” when deploying AI in production environments. It highlights that global token consumption is projected to increase 24-fold between 2026 and 2030, driven by both enterprise and consumer adoption, making accurate economic assessment and robust governance critical for sustainable AI operations. Japanese businesses, often slower to adopt new technologies at scale compared to their Western counterparts, are now confronting the operational realities of AI. This piece reflects a growing maturity in how Japanese enterprises are thinking about AI: moving past initial experimentation to focus on cost sustainability and governance in real-world applications, which often involves integrating public and private AI environments for different workloads. This is a practical, engineering-focused view, not just a strategic one.
For Western readers: Western business leaders should recognize that the discussion in Japan has moved from “if” to “how” with AI, specifically concerning its cost management and operational governance, implying that the competitive landscape will increasingly be defined by execution efficiency, not just innovation. 🗾 AI & Machine Learning
📊 Featured Chart
Output token prices
Google has launched three new ‘Flash’ series Gemini models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. These models aim for efficiency, low latency, and reliability in building AI agents, with 3.6 Flash and 3.5 Flash-Lite featuring reduced output token usage and lower pricing. Google also announced the start of development for its next-generation ‘Gemini 4′ model. Google’s move to refine its Flash models with improved efficiency and lower costs directly addresses the practical concerns of East Asian enterprises, which often prioritize operational expenditure and predictable performance for large-scale deployments. The specialized cybersecurity model, 3.5 Flash Cyber, is particularly relevant given regional governments’ increasing focus on digital resilience and the ‘dual-use’ nature of advanced AI, which governments here view with cautious interest. Japanese outlets like ITmedia tend to emphasize the direct business applications and technical specifications over the geopolitical positioning often found in Western reporting.
For Western readers: Western businesses building AI applications in East Asia should expect heightened competition in enterprise AI platforms as Google aggressively targets efficiency and specialized applications, particularly in cybersecurity, necessitating a re-evaluation of current platform choices based on cost-performance and security features. Policy & Regulation
Trump rebuilds tariff wall with new rates on 60 countries, Japan and China impacted 📊 Featured Chart
Tariff rates as of July 24, 2026
The U.S. has imposed new tariffs ranging from 10% to 12.5% on dozens of trading partners, with Japan and China specifically facing a 12.5% levy on their goods. Other Asian nations like India, Malaysia, Bangladesh, and Cambodia are subject to a 10% tariff. This isn’t just a political announcement; it’s a direct operational cost for Japanese and Chinese companies. While Western media focuses on the political motivations, the real impact is in logistics and sourcing. Japanese firms, in particular, have been trying to ‘friend-shore’ production to avoid these exact scenarios, but moving complex supply chains takes years, not months.
For Western readers: Western companies relying on components or finished goods from Japan and China should assume increased import costs and longer lead times due to disrupted supply chains and re-routing efforts by East Asian partners. Begin re-evaluating sourcing strategies now, as these tariffs are likely to persist. Policy & Regulation
Taiwan eases rules on foreign telecom satellites, eyes Musk’s Starlink Taiwan has relaxed foreign ownership restrictions on satellite operators, aiming to enable services like Starlink amidst concerns over potential communication disruptions with China. This move addresses Taiwan’s lack of indigenous low-Earth orbit satellites and seeks to bolster its communication resilience. Taiwan’s easing of satellite rules isn’t just about internet access; it’s a strategic infrastructure play. While Western media might frame this as a market opening, for Taipei, it is explicitly about crisis readiness and bypassing Chinese cyber or kinetic attacks on terrestrial communications. The goal is survivability, not just competition.
For Western readers: If you are assessing supply chain risk for technology operations in Taiwan, factor in that the island is prioritizing redundant satellite communication pathways via foreign operators as a national security measure, which implies a government-backed fast-tracking of deployments. AI & Machine Learning
Chinese AI sensation Moonshot’s gamble on big models pays off Chinese AI startup Moonshot AI is finding success with its Kimi K3 large language model, which was featured prominently at the World Artificial Intelligence Conference in Shanghai. This indicates a growing maturity and user acceptance of domestic Chinese LLMs, contrasting with earlier concerns about their capabilities relative to Western counterparts. Chinese coverage often frames the success of companies like Moonshot AI as a testament to national technological resilience, especially in the face of external pressure. This view emphasizes the ability of Chinese firms to innovate and compete using domestic resources and talent, rather than solely relying on imported technology or frameworks.
For Western readers: Western businesses should recalibrate their assumptions about the pace of Chinese LLM development; Moonshot’s market traction shows that China is developing effective, competitive models for its domestic market even without access to the most advanced US chips.
🔺 The Triangle #
Where US, Japan, and China technology interests intersect
Hardware-level design security and physical AI integration are the new battlegrounds as supply chains hedge against Geopolitical volatility.
Semiconductors & Hardware
Apple and Samsung Defy Global Smartphone Market Decline in 2Q 2026 📊 Featured Chart
Source: Omdia
Global smartphone shipments fell 4% year-on-year in 2Q 2026 due to a memory crisis and rising component costs, but Samsung and Apple grew their market share. Chinese vendors Xiaomi, OPPO, and vivo maintained their positions in the top five, but faced pressure in the sub-$400 mass market segment. The ‘memory crisis’ hitting the mass market is not just a general supply crunch, but a critical cost issue that disproportionately squeezes Chinese manufacturers relying on thinner margins. This forces them to pivot from a volume-driven strategy, which has been their strength, towards optimizing product portfolios for value, a shift that is harder to execute for companies built on scale.
For Western readers: If you are a Western component supplier to Chinese smartphone brands, expect continued pressure on order volumes for lower-end components and a strategic shift from Chinese OEMs towards higher-value, albeit lower-volume, offerings. Semiconductors & Hardware
Cadence Launches AuraStack AI Super Agent for PCB Design, Advanced Packaging Cadence Design Systems has introduced AuraStack AI Super Agent, a new agentic AI platform for PCB and advanced packaging design, building on its existing suite of AI Super Agents. This platform aims to automate and accelerate the system design cycle, from planning through manufacturing, for complex electronic systems. The technology is particularly relevant as hyperscale data centers and other industries deploy more intelligent, high-performance systems. The introduction of agentic AI in PCB and advanced packaging design directly addresses the escalating complexity in high-performance computing, especially for AI infrastructure. For Japanese and Chinese firms building out their AI capabilities, this means new opportunities to streamline system design, potentially reducing development cycles for domestic hardware. However, it also means continued reliance on Western EDA toolchains for cutting-edge design, reinforcing existing supply chain dependencies.
For Western readers: Western hardware designers and AI infrastructure providers should recognize that this advancement in EDA tools will accelerate the design and deployment of advanced electronics, including those developed by competitors in East Asia, potentially shortening time-to-market for complex AI systems globally. Semiconductors & Hardware
QuickLogic, PQSecure Demo Reprogrammable Post-quantum Cryptography for Future-proof SoCs
QuickLogic and PQSecure Technologies have demonstrated a reprogrammable post-quantum cryptographic (PQC) IP core using QuickLogic’s eFPGA Hard IP, developed for the Intel 18A process node. This solution allows for in-field updates to PQC algorithms without requiring costly silicon redesigns, addressing the evolving standards set by the U.S. National Institute of Standards and Technology (NIST). The collaboration aims to provide crypto-agile hardware for future-proof System-on-Chips (SoCs). The shift to post-quantum cryptography is a defensive play for countries like Japan and China, aiming to secure their digital ecosystems against future threats and reduce reliance on foreign cryptographic standards. This kind of reprogrammable IP helps them future-proof their domestic chip designs and manage the transition without immediate, costly hardware overhauls, allowing for greater control over their own security postures.
For Western readers: Western SoC designers supplying East Asian markets should recognize that the ability to update cryptographic IP in the field will become a core requirement for securing new design wins, particularly for sensitive government or industrial applications in Japan and China. Semiconductors & Hardware
Melaka Strengthens Semiconductor Strategy Through Government-Industry Dialogue Melaka, Malaysia hosted a dialogue with government agencies, industry leaders like Infineon, and financial institutions to advance its semiconductor and advanced manufacturing ambitions. The event focused on implementing the Melaka Semiconductor Strategic Plan 2026-2035, which supports Malaysia’s broader National Semiconductor Strategy. The Melaka plan is a localized version of Malaysia’s national push to secure more advanced semiconductor investment. While Malaysia has long been a key hub for outsourced semiconductor assembly and test (OSAT), these dialogues signal a coordinated effort to move up the value chain, which means more domestic R&D, design, and potentially even front-end manufacturing over time. It’s a long play, but it’s a real strategy to capture more capital from global chipmakers looking to diversify beyond traditional hubs.
For Western readers: Western semiconductor firms with significant operations in Malaysia, like Infineon, will find local governments increasingly pushing for greater technology transfer, talent development, and deeper integration of Malaysian firms into their supply chains, rather than just offering cheap labor and land. AI & Machine Learning
Thermal Imaging as Infrastructure for Physical AI Meridian Innovation, with contributors Hock Leow and Stan Markov, highlights thermal imaging’s role in advancing “Physical AI,” which allows machines to understand real-time physical states like temperature and energy flow, not just visual appearance. This technology addresses a critical gap in sensor fusion, enabling AI systems to predict industrial failures, monitor vital signs, and navigate challenging environments by adding a “state channel” to perception. The article emphasizes the efficiency of thermal sensors, citing a 160×120 thermal sensor generating 170,000 pixels per second of 16-bit data. While the article is vendor-agnostic, the emphasis on real-time physical state analysis resonates strongly with East Asian industrial policy. Both Japan’s drive for smart manufacturing and China’s ‘Industrial Internet’ initiatives require precise, real-time sensing for automation, predictive maintenance, and quality control. Local companies are keen to integrate such capabilities to enhance operational efficiency and reduce reliance on human oversight in factories and infrastructure.
For Western readers: Western businesses in industrial automation and smart infrastructure should recognize that East Asian competitors are rapidly integrating advanced sensor fusion, including thermal data, into their AI deployments, potentially leading to more resilient and autonomous systems. 🧩 Pattern This Week
Policy: Trump-era tariff walls pressure global supply chains, hitting Japan and ChinaJapan/Taiwan: Taiwan relaxes satellite rules to secure backup communications against regional threatsMalaysia: Melaka aligns with Western chip firms to strengthen its assembly footprint
Rising trade barriers and regional security risks are forcing Indo-Pacific nations to restructure their supply chains and critical infrastructure, leaving companies without diversified manufacturing bases exposed to severe operational bottlenecks.
AsiaAI.FYI · Written by Dick Weisinger ·