East Asian Technology Intelligence
Japan & China tech news — translated, contextualized, and delivered often.
Free. Unsubscribe anytime.
3 Takeaways This Issue
- SK Hynix shares dropped over 15% on the Seoul exchange immediately following its Nasdaq debut as institutional investors pulled back on concerns that the global AI memory chip cycle is nearing its peak.
- TSMC countered cooling macroeconomic indicators by posting a 36% year-over-year Q2 revenue surge to $39.6 billion, driven by sustained demand that has forced the foundry to expand its advanced packaging capacity in Chiayi, Taiwan.
- Japan’s Ministry of Justice is drafting explicit legal protections for voice data to prevent unauthorized AI cloning, a regulatory shift that will force foreign developers to restructure their localized training datasets.
Core Move
TSMC Q2 Revenue Up 36% YoY, Expands Advanced Packaging in Chiayi #
📊 Featured Chart
Source: TSMC
TSMC is building advanced packaging capacity in Chiayi fast. This move shows that the main bottleneck for AI chips is shifting. The problem is no longer leading-edge wafer fabrication. Instead, the challenge is now how to connect different parts on a single chip.
The company grew its revenue by 36% year over year to $39.6 billion in Q2. These numbers are impressive, but they mostly show past demand. The true sign of what is coming is the $9.35 billion expansion in packaging. This investment adds two new plants to the two existing ones. The move ensures TSMC controls the silicon and the assembly technology. This assembly determines real-world AI performance.
Western observers often focus on nanometer process nodes. Yet AI accelerators require multiple chiplets, like logic, HBM, and custom IP. Getting these parts to talk at high bandwidth in one package is just as hard as making the silicon. This expansion shows TSMC is doubling down on that fact. This is Taiwan’s version of industrial policy. It ensures the physical point of assembly stays in their hands as global AI demand surges. They are actively shaping the future supply chain architecture for AI.
Many people assume packaging is a cheap, simple process, but that is wrong. Advanced packaging requires unique intellectual property and specialized tools. It also needs very tight process control, especially for 3D stacking and chiplet integration. This effort is not just about protecting market share. It creates a major barrier for rivals. Competitors cannot easily copy leading AI chip performance, even if they can make similar wafers.
One big risk is that demand for advanced packaging will still outstrip supply. This is especially true for the newest AI accelerators. TSMC is spending a lot of money, but specialized tools have long lead times. Ramping up new factories also takes a long time. Any delay in tool supply or construction could make this bottleneck worse. That would cause problems across the whole AI hardware market.
To see how this plays out, check TSMC’s quarterly capital expenditure reports. Look at how much money goes to packaging versus wafer tools. We should also watch announcements from major AI chip designers like NVIDIA and AMD. See if they use other packaging partners to diversify their supply. Finally, track the progress of Intel’s IFS and Samsung Foundry. Their ability to scale competing services will show if TSMC can keep its lead.
🗾 Japan Radar #
What Japanese media is reporting that Western outlets miss
Japan is tightening AI governance and reorganizing corporate leadership to defensive, risk-mitigating positions rather than pursuing high-growth frontier innovation.
🗾 Enterprise & Cloud
Microsoft unveiled a suite of new AI technologies at its Build 2026 conference, focusing on agent-based business automation. Key announcements include the ‘Microsoft IQ‘ intelligence foundation, a new family of seven proprietary models including the MAI-Thinking-1 inference model, and a full-stack development environment featuring the Surface RTX Spark Dev Box for local LLM development. Microsoft’s strategy, as presented at Build 2026, emphasizes enterprise control and domain knowledge integration within AI agents. While Western narratives often focus on large, general-purpose models, Japanese enterprise customers prioritize secure, auditable systems that integrate with existing IT infrastructure. The introduction of ‘Frontier Tuning’ for compliance and ‘Agent 365’ for agent governance directly addresses these concerns, suggesting Microsoft is positioning its AI offerings to meet the conservative IT requirements prevalent in Japan and other mature markets.
For Western readers: Western enterprises should expect increased competition from Microsoft in providing secure, managed AI agent platforms, and should evaluate their own strategies for integrating proprietary data and governance into AI workflows, rather than relying solely on public models or unmanaged third-party agents. 🗾 Workforce & Culture
GMO Internet Group announced that its Group CEO, Masatoshi Kumagai, has additionally assumed the role of Chief AI Transformation Officer (CAIO) as of July 13. Kumagai will centrally oversee the group’s AI strategy, implementation, governance, talent development, and business transformation. He states that coding is no longer solely a human task and plans to restructure the organization, including engineers, to be ‘AI-nized,’ aiming for a ‘hyper-automation group‘ where all employees use ‘** vibe coding**‘ to solve problems. GMO’s aggressive, CEO-led push for ‘AI-nization’ is notable in Japan, where enterprise AI adoption often proceeds with more caution and consensus-building. Kumagai’s personal involvement and hands-on approach, including practicing ‘vibe coding’ with Claude Code, sets a strong precedent for other Japanese companies, prioritizing rapid execution over gradual integration.
For Western readers: Western business leaders should recognize that when a Japanese CEO takes on a ‘Chief AI Transformation Officer’ role personally, it indicates a serious commitment to deep organizational change, not just a pilot program. If you’re partnering with GMO or similar Japanese firms, expect a direct mandate for AI integration in joint projects and a shift in their workforce skill sets. 🗾 Policy & Regulation
Japan’s Ministry of Justice study group is considering explicit inclusion of a ‘right not to have one’s voice used without permission’ in a forthcoming report, aiming to address concerns over AI-driven voice synthesis and cloning. This move seeks to establish legal protections against unauthorized voice replication, impacting the development and use of AI technologies. The discussion follows increasing public and industry concerns about the ethical and legal implications of generative AI. Japanese government bodies, particularly METI and the Ministry of Justice, are acutely aware of the ‘** data sovereignty**‘ implications of foreign AI models using Japanese personal data, including voice. This move isn’t just about individual rights; it’s also about asserting a degree of control over a critical resource—distinctive Japanese voice data—that AI models require, often framed as protecting cultural identity while serving industrial policy aims.
For Western readers: Western developers and service providers planning to deploy AI models trained on Japanese voice data or offering voice-synthesis services in Japan must anticipate a more stringent legal landscape and potential licensing requirements, or risk legal challenges for unauthorized use. 🗾 Policy & Regulation
Law to Combat SNS Disinformation During Elections Passed in Upper House Plenary Session
Japan’s Parliament has passed new legislation aimed at countering disinformation on social media platforms during election periods. The law mandates social media companies to implement measures against false information, including the prompt removal of problematic posts when identified by authorities. It also requires platforms to enhance transparency regarding politically motivated advertisements and to clearly label AI-generated content used in campaigns. The Japanese government, traditionally less interventionist in online content than some Western counterparts, is moving decisively to establish rules for social media platforms. This is less about censorship and more about defining the responsibilities of platform providers within Japan’s structured electoral system, particularly as AI-generated content becomes more prevalent. The local focus is on maintaining public trust in elections and the integrity of information, a consistent concern for the current administration.
For Western readers: Western social media companies must now implement robust, proactive content moderation systems tailored to Japanese election law, or face potential fines and reputational damage. Semiconductors & Hardware
SK Hynix Shares Drop After Nasdaq Debut Amid Peak Memory Chip Concerns SK Hynix shares fell over 15% on Monday, just days after a successful Nasdaq debut where it initially rose 13%, amidst growing concerns that the memory chip industry’s current uptrend may have peaked. This decline also impacted the wider KOSPI index, reflecting broader market anxieties about the sector’s valuation. While Western financial news might frame this as a typical market correction, the sharp drop in SK Hynix, a key player in HBM (High Bandwidth Memory) vital for AI, suggests a more pointed concern within the East Asian industry about the sustainability of current AI demand forecasts and pricing. It’s not just a correction; it’s a test of whether AI’s promise can truly overcome the memory cycle.
For Western readers: Western investors betting heavily on continuous, exponential growth in memory demand driven by AI should re-evaluate their assumptions about the longevity of the current chip cycle, particularly regarding prices for HBM and DRAM.
🇨🇳 China Watch #
China’s technology moves, framed for Western readers
Beijing is accelerating its technological self-reliance, leveraging domestic operating systems, unified AI standards, and advanced manufacturing to bypass Western encirclement.
Enterprise & Cloud
OpenHarmony’s Six-Year Transformation: Expanding Ecosystem for Thousands of Industries in China Huawei’s OpenHarmony, China’s open-source operating system, has significantly expanded its presence across various industries, now connecting over 370 million devices and boasting over 80 industry distributions. The platform, developed since 2020, aims to reduce China’s reliance on foreign operating systems, particularly in critical infrastructure and smart devices. The transformation emphasizes its application in sectors like finance, government, transportation, and industrial manufacturing, signifying a national push for digital self-reliance. This isn’t just about consumer phones anymore; OpenHarmony’s expansion into industrial and government applications means China is systematically replacing foreign software at the foundational layer. The sheer breadth of adoption, from energy grids to financial systems, indicates a concerted effort to establish a national digital infrastructure immune to external influence. Western reports often fixate on benchmarks and consumer market share, missing the strategic depth of this industrial pivot.
For Western readers: Western enterprise software and IoT platform providers should assume China’s critical infrastructure and state-owned enterprises will progressively mandate OpenHarmony compatibility, effectively closing off those market segments to foreign OS solutions over the next 3-5 years. AI & Machine Learning
Has Tencent AI Finally Turned the Corner? Inside Hunyuan Hy3, the Model That Got Tencent Back in the Race Tencent has launched Hunyuan Hy3, its latest foundation model, touting significant performance improvements that place it competitively with leading domestic and international AI models. The company emphasized Hunyuan Hy3’s advancements across general intelligence, coding, and multimodal capabilities, aiming to regain ground in China’s intensely competitive AI landscape after previous models lagged. Tencent’s previous AI efforts often felt more like an obligation to participate than a genuine push for leadership, especially compared to Baidu or Alibaba. Hunyuan Hy3 appears to be a more serious attempt to build a foundational model that can actually be deployed across their enterprise and consumer applications, rather than just a benchmark play. The company needs this to avoid losing more ground in the domestic AI race.
For Western readers: Western businesses assessing the competitive landscape for AI tools in China should now consider Tencent’s Hunyuan Hy3 a viable contender, especially for integrations within Tencent’s own ecosystem like WeChat or Tencent Cloud. Semiconductors & Hardware
Google Reportedly First to Adopt TSMC’s 2nm Chip for Tensor G6, Ahead of Apple Google’s Tensor G6 chip is slated to be the first product to use TSMC’s 2nm process, powering the Pixel 11 series set for an August 2026 launch. This marks a shift from Apple’s historical position as the debut customer for TSMC’s leading-edge nodes, with Apple’s iPhone 18 Pro expected to follow in September 2026 with the same 2nm technology. For TSMC, securing a major customer like Google for the 2nm debut demonstrates continued strong demand for its most advanced nodes, beyond its dominant customer, Apple. This also shows Google’s increasing commitment to custom silicon, pushing for an early lead in mobile processor performance by investing heavily in cutting-edge fabrication.
For Western readers: Western hardware companies relying on TSMC for leading-edge process technology should recognize that their access might not be automatically prioritized over other major players, especially if the volume or strategic importance of other customers grows. Policy & Regulation
Chinese Internet Firms Sign AI Agent Data Protection Pact The China Internet Association, with major players like Baidu, Tencent, Alibaba, and Volcengine, has released a self-regulatory pact for personal information protection by AI agents in China. This move aims to standardize data handling as AI agent services become more prevalent across Chinese internet platforms. A separate pact for mini-program ecosystems was also signed by key platform operators. Beijing views AI agents as critical infrastructure, not just consumer features. This pact, while framed as self-regulation, functions as a state-endorsed baseline for how these agents must operate. It’s less about genuine corporate autonomy and more about establishing a common, state-aligned operational standard across the sector, ensuring data flows are predictable and ultimately controllable.
For Western readers: Western firms engaging with China’s digital economy should expect that any AI agent services they develop or deploy in the Chinese market will need to conform to these specific, domestically-defined data protection standards, which prioritize state oversight more than Western equivalents. Policy & Regulation
EU should think twice before weaponising its market against China
Chinese analysts argue that the EU’s assumption of formidable leverage over China, based on its 450 million population and second-largest economy, is flawed. They contend that market size alone does not automatically translate into geopolitical coercive power. This perspective warns the EU against miscalculating its economic dependency on China. Beijing has been increasingly vocal about the perceived hypocrisy of Western nations calling for ‘de-risking’ while simultaneously employing their markets as strategic assets. From their view, this Western approach risks global economic fragmentation, which could destabilize supply chains and ultimately harm all participants, including the EU. Chinese commentary frames this not as a necessary defense, but as a dangerous overestimation of market leverage and underestimation of mutual dependency.
For Western readers: Western policymakers should recognize that China views the EU’s market ‘weaponization’ as a tactical misjudgment that underestimates Beijing’s resilience and willingness to implement countermeasures, rather than as an unassailable source of leverage.
🔺 The Triangle #
Where US, Japan, and China technology interests intersect
East Asian tech giants are bypassing Western bottlenecks by shifting talent, supply chains, and distribution deeper into neutral territory.
Semiconductors & Hardware
Tesla A15 Chip to be Fabbed at Samsung and TSMC on 2nm Process 📊 Featured Chart
A15 chip design specs
Tesla’s A15 chip will be produced using 2nm processes at TSMC in Taiwan and Samsung’s Taylor, Texas fab, with potential for TSMC’s Arizona site and Tesla’s Texas Terafab to join later. The A15 chip aims for 2,000 to 2,500 TOPS and supports 144GB unified memory. This split fabrication strategy diversifies risk and leverages advanced manufacturing capabilities across key East Asian and US players. Tesla splitting its 2nm fabrication between TSMC and Samsung indicates how tight advanced process node capacity has become. It’s not about which foundry is ‘better,’ but about securing volume and diversifying risk, a common strategy for any major chip customer today. The simultaneous use of both Taiwanese and US fabs for TSMC production also reflects geopolitical realities and supply chain resilience efforts, even if the primary driver remains commercial.
For Western readers: Western tech companies designing leading-edge chips should assume multi-vendor fabrication strategies will become the norm for securing capacity and managing geopolitical risk, rather than relying on a single foundry relationship, especially for advanced nodes. Semiconductors & Hardware
Q2 PC Shipments Down 4.9% Amid Memory Shortage; East Asian Vendors Consolidate Supply Global PC shipments declined 4.9% year-over-year in Q2 2026, marking the first drop in nine quarters, primarily due to a persistent memory chip shortage. Major East Asian PC vendors like Lenovo are leveraging their scale across diverse product lines to secure scarce memory supplies, accelerating industry consolidation. While Western reports frame this as a general market contraction, the underlying dynamics disproportionately benefit large Asian players. For instance, companies like Lenovo are using their expansive market reach, from smartphones to servers, to hoard limited memory. This isn’t just about weathering a downturn; it’s about systematically gaining market share by ensuring component access, squeezing out smaller, often Western-facing, competitors who lack that diversified scale.
For Western readers: If you are a smaller Western PC vendor or rely on direct-to-consumer PC sales, plan for continued component scarcity and higher costs through 2027, as large East Asian players will continue to prioritize their own integrated supply chains. Cross-Regional Analysis
Nobel Chemist Departs US for China AI Lab, Apple Sues OpenAI Over IP Nobel-winning chemist Omar Yaghi is leaving the United States to lead an AI institute in China focused on new materials discovery, amidst China’s efforts to recruit US scientists. This move contrasts with reports of reduced science spending in the White House. Separately, Apple has initiated a lawsuit against OpenAI, alleging trade secret theft and staff poaching related to its consumer hardware development. China’s strategy of recruiting top Western scientists like Yaghi is a direct path to acquiring advanced research capabilities, bypassing slower organic development. While Western coverage often emphasizes the US-China race in large language models, China’s focus on foundational sciences like materials discovery, powered by AI, provides a more concrete and industrial long-term advantage in manufacturing and supply chain control.
For Western readers: Western R&D leaders should assume China will continue to aggressively recruit high-value scientific talent, particularly in fields with dual-use implications for AI and advanced manufacturing, and factor this into national security and industrial policy. US policy decisions on science spending directly affect its ability to retain leading researchers. Cross-Regional Analysis
China’s Zeekr Expands Malaysian Dealer Network to 11 Outlets, Targets 18 by Year-End
China’s premium EV brand Zeekr, part of Geely Holding Group, has expanded its Malaysian dealer network to 11 outlets by adding three new partners, with a target of 18 by the end of 2026. Zeekr entered Malaysia in December 2024 and claims the top-selling premium EV brand position for January-April 2026. This expansion includes new models like the 2026 Zeekr X, 7X, and 009, with the 9X and 009 Grand recently opened for booking. Zeekr isn’t just selling EVs; they’re establishing a full sales and after-sales infrastructure in a key ASEAN market. This isn’t about one-off car sales; it’s about building long-term brand presence and service capability, which is a harder barrier to entry for Western or Japanese competitors. The speed of their build-out, going from market entry to 11 outlets in less than two years, is what stands out.
For Western readers: If you are a Western automaker, assume Chinese EV brands like Zeekr will rapidly replicate this market entry and infrastructure build-out strategy across other ASEAN and emerging markets, and that their success will rely on deep dealer networks, not just vehicle price points. Startups & Funding
Gobi Partners and NTT Partner to Connect Japanese Enterprises with Southeast Asian Startups
Malaysia-based Gobi Partners and Japan’s NTT have formed a strategic collaboration to link Japanese enterprises with Southeast Asian startups, focusing on business validation, proof-of-concept initiatives, and cross-border commercial engagement. This partnership aims to bridge Gobi’s extensive startup network with NTT’s enterprise capabilities to facilitate the adoption of emerging technologies across both regions. Gobi’s move into Japan, following prior relationships with East Japan Railway and JETRO, demonstrates a persistent effort to connect Southeast Asian entrepreneurial dynamism with Japan’s industrial base. This isn’t just about venture capital deals; it’s about NTT leveraging its enterprise scale to scout and integrate new technologies, providing a direct pipeline for SEA startups into a major Japanese corporate ecosystem.
For Western readers: Western tech investors and companies operating in Southeast Asia should recognize that Japanese corporate capital, channeled through entities like NTT and partners such as Gobi, will increasingly compete for promising regional startups, potentially driving up valuations or tying up strategic partnerships. 🧩 Pattern This Issue
Taiwan/Korea: TSMC posts 36% growth while SK Hynix falls on peak memory fearsChina: Google taps TSMC 2nm for Tensor G6 to leapfrog Apple’s hardwareChina: Tencent launches Hunyuan Hy3 model to challenge domestic foundation rivals
While Western tech-stack competition focuses heavily on proprietary foundation models, the actual leverage in the AI transition is concentrating in advanced hardware manufacturing, exposing companies that rely on standard memory and lagging foundry nodes.
[AsiaAI.FYI](https://asiaai.fyi) ·
Written by Dick Weisinger ·