East Asian Technology Intelligence

Japan & China tech news — translated, contextualized, and delivered often.

Subscribe Free →

Free. Unsubscribe anytime.

3 Takeaways This Issue

• SK Hynix and Nanya Technology are aggressively scaling up capital expenditures to meet HBM and DDR5 demand, betting that the structural requirements of AI infrastructure will permanently break the memory sector’s historic three-year boom-and-bust cycle.
• MiniMax secured a $2.2 billion funding round within 48 hours of a massive employee stock lock-up expiration crash, demonstrating that Chinese state-backed funds and private investors will step in as lenders of last resort to stabilize national AI champions.
• Japan’s Cabinet Office and the Council for Science, Technology and Innovation are launching a coordinated dual-use technology initiative to channel commercial AI and robotics research directly into defense applications, marking a departure from the country’s postwar separation of civilian and military R&D.

Semiconductors & Hardware2 STORIES

East Asian Memory Giants Ramp Up Capital to Feed AI Demand

South Korea’s SK Hynix and Taiwan’s Nanya Technology are aggressively scaling up their financial war chests to combat an unprecedented, AI-driven global memory crunch. SK Hynix secured a historic $26.5 billion through a record U.S. share sale to boost high-bandwidth memory (HBM) production, while Nanya plans to quadruple its capital expenditures by 2027 to expand DRAM manufacturing capacity.

Why it matters: This cash infusion lets SK Hynix double down on HBM production, which is crucial for AI accelerators. The company needs this capital to keep pace with demand from Nvidia and others, and to maintain its competitive edge over Micron and Samsung in the high-margin HBM segment. It’s a clear move to cement their position at the top of a critical supply chain component.

For Western readers: Western AI developers and cloud providers should expect the supply of HBM to loosen somewhat in 18-24 months, but that capacity is still largely going to be tied to existing relationships with major AI chip designers.

Nikkei Asia · Nikkei Asia

🗾 AI & Machine Learning2 STORIES

OpenAI Double Feature: GPT-5.6 Sol and Ultra-Natural GPT-Live Models Debut

OpenAI has expanded its lineup with the general release of its cost-efficient GPT-5.6 suite and the launch of GPT-Live, a conversational model praised for its fluid, human-like voice interaction. While the flagship ‘Sol’ model challenges rivals by matching premium performance at half the cost, GPT-Live’s full-duplex architecture allows it to interrupt, adapt its tempo, and speak without pauses. Together, these releases signal a shift toward highly practical AI that prioritizes economic viability and seamless, real-time human interaction over raw benchmarks.

Why it matters: The immediate comparison to Anthropic’s Fable 5 on cost and speed signals that the race for foundation model dominance is now as much about economic efficiency and deployment friction as it is about raw benchmark scores. For Japanese firms considering large-scale AI integration, the cost reduction argument is compelling, especially for enterprise applications where budget constraints often limit ambitious deployments.

For Western readers: Western enterprises reliant on LLMs for coding or knowledge work should re-evaluate their current model choices, as GPT-5.6’s cost-performance ratio suggests opportunities for significant operational savings without sacrificing capability.

ITmedia AI+ · ITmedia AI+

Policy & Regulation

Japanese government to work with industry and academia on dual-use tech

Prime Minister Sanae Takaichi held a meeting of the Council for Science, Technology and Innovation, emphasizing closer collaboration among government, industry, and academia to advance dual-use technologies. This initiative aims to strengthen Japan’s capabilities in areas that serve both civilian and national security interests, reflecting a broader strategic push. Japan’s framing of ‘dual-use’ often translates to explicit industrial policy aimed at fostering self-sufficiency in areas deemed critical for national security and economic resilience. This isn’t just about military tech; it’s about controlling the underlying manufacturing and research base to prevent reliance on external suppliers, particularly from China, for components and expertise.

For Western readers: Western companies involved in advanced materials, AI, or specialized manufacturing should expect increased Japanese government incentives and mandates for domestic production and collaboration, potentially shifting supply chain options within Japan itself.

The Japan Times

Startups & Funding

MiniMax: Lock-Up Crash, Then $2.2B Fundraise in 48 Hours

Chinese AI startup MiniMax saw a significant portion of its early employee stock options, reportedly 5-10%, sold off immediately after its lock-up period expired, causing a sharp price drop on secondary markets. Despite this, the company quickly secured a $2.2 billion funding round led by Tencent and other investors, highlighting continued investor confidence in leading Chinese AI foundation model companies. The rapid sell-off by early employees followed by a massive fundraise illustrates the internal and external pressures on Chinese AI champions. It suggests that while early employees are eager to cash out, major investors like Tencent see these companies as too critical to let falter, particularly as China pushes for domestic AI leadership.

For Western readers: Western investors tracking China’s AI sector should not assume employee sell-offs indicate a lack of confidence in core business. Instead, treat significant capital injections from state-linked entities or major tech firms as the stronger signal of strategic national interest and long-term viability, regardless of short-term secondary market volatility.

Pandaily

AI & Machine Learning

ByteDance Seedream 5.0 Pro Hands-On: Three Directions, 10 Scenarios Tested

ByteDance’s latest AI model, Seedream 5.0 Pro, demonstrates enhanced multimodal capabilities across text, image, video, and audio generation, positioning it as a competitor to leading Western models. The review highlights its strong performance in complex content creation, with notable improvements in video generation and real-time interaction for Chinese language users. This iteration focuses on practical applications and scenario-based performance, moving beyond raw benchmark scores. The focus on ‘three directions’ (multimodality, real-time, long-context) and ’10 scenarios’ for Seedream 5.0 Pro indicates a strategic shift towards practical, deployable AI applications rather than purely academic benchmarks, aligning with China’s broader industrial policy for AI. While Western coverage often fixates on model size or benchmark scores, Chinese reporting emphasizes how these models can be integrated into real-world products and services, especially for domestic users, which ByteDance is doing with this model.

For Western readers: Western AI companies evaluating competitive threats should pay less attention to headline benchmark numbers and more to how Chinese models like Seedream 5.0 Pro are being optimized for specific real-world applications and multimodal fluency, particularly for non-English content.

Pandaily

Robotics & Automation

DJI Launches EV50 VTOL Fixed-Wing Cargo Drone, Claims Everest Altitude Record

Chinese drone maker DJI has released its first vertical takeoff and landing (VTOL) fixed-wing cargo drone, the EV50, designed for long-range regional transport. The company states the EV50 set a new high-altitude record for VTOL drones, operating at 8,861 meters during a scientific expedition on Mount Everest. The EV50 showcases DJI’s engineering depth beyond consumer drones, moving into complex industrial platforms. The fixed-wing VTOL design is important for longer ranges and heavier payloads, which is where the commercial viability of drone logistics starts to shift from novelty to practical utility in hard-to-reach areas. China’s focus on national infrastructure and remote connectivity means these platforms get serious attention.

For Western readers: Western logistics and aerospace firms should recognize that China, through DJI, is actively pushing the boundaries of drone capability for challenging, high-value applications like scientific research and remote area resupply, rather than just consumer markets. This isn’t just a niche product; it’s a testbed for future capabilities that will find broader applications.

TechNode

Semiconductors & Hardware

DeepSeek Initiates In-House AI Chip Development to Reduce NVIDIA Dependency

Chinese AI startup DeepSeek has commenced an internal project to develop custom AI chips, primarily for inference workloads, aiming to reduce operational costs and lessen reliance on foreign suppliers like NVIDIA. This move is driven by the escalating compute expenses associated with its growing foundation models, such as DeepSeek-V3 and R1. Chinese AI firms like DeepSeek are increasingly taking on hardware development themselves, a costly and complex undertaking usually reserved for dedicated chip companies. This push illustrates the severity of the supply chain constraints from US export controls and Beijing’s strategic urgency to localize critical AI infrastructure, well beyond what the average Western business news story might convey.

For Western readers: Western AI model developers should assume that China’s leading AI firms will continue to prioritize vertical integration into hardware, creating an increasingly bifurcated global AI ecosystem with distinct domestic supply chains.

TechNode

Policy & Regulation

How China’s ‘New Left’ ideology shaped 21st-century China’s economic path

This opinion piece argues that China’s “New Left” has successfully revised the narrative around Deng Xiaoping’s reforms, emphasizing Mao Zedong’s foundational role in establishing a disciplined party and economic structures that enabled ‘gaige kaifang’ (reform and opening-up). It challenges the Western neoliberal view that Maoist command economics and Deng’s capitalist reforms were diametrically opposed, asserting a continuity often overlooked. The ‘New Left’ ideology provides the intellectual justification for Beijing’s continued emphasis on state-directed industrial policy and its current push for technological self-reliance. This differs from Western views that often credit market liberalization alone for China’s economic ascent. The Chinese framing sees state discipline and a unified national project as fundamental preconditions, not impediments, to economic success, implying a continued preference for state guidance in key tech sectors.

For Western readers: Western businesses operating in China or engaging with Chinese tech partners should recognize that the Chinese government’s strategic decisions, particularly in AI, semiconductors, and data, are driven by an internal logic that values state control and national cohesion as much as, if not more than, pure market efficiency.

South China Morning Post — Tech

Cross-Regional Analysis

IMF: South Korea, Taiwan, Thailand, Malaysia Lead AI Hardware Export Surge

The IMF reports that South Korea, Taiwan, Thailand, and Malaysia led global economic growth in Q1 2026, driven by their roles as major net exporters of AI-related hardware. These four economies averaged a 4.4 percentage point growth surprise above forecast, significantly outperforming other countries. Western analysts tend to focus on the performance of generative AI models and the major cloud providers. This IMF data reminds us that the actual production and export of AI-enabling hardware — the physical layer — is where much of the immediate economic value is being created and captured, especially in East Asia. The projections confirm the hardware cycle’s strength.

For Western readers: Western firms dependent on AI hardware supply chains should expect continued tightness and potentially rising costs, as demand from these key exporting nations points to sustained global appetite for components.

Technode Global

Semiconductors & Hardware

Samsung’s AI Chip Sales Drive 1,800% Profit Jump Amid Global AI Market Concerns

Samsung Electronics reported an 1,800% jump in profits, achieving its third consecutive record quarterly profit, driven by booming AI chip sales. This success comes as global concerns grow about an overinflated AI market bubble and US companies increasingly explore cheaper Chinese AI models. Samsung’s earnings are a real-world measure of AI’s hardware demand, contrasting sharply with the speculative fears in some Western financial analyses about an AI bubble. The underlying trend here is that, regardless of how ‘overvalued’ software companies might seem, the physical demand for silicon is concrete and growing, directly benefiting major East Asian component suppliers. The reported cost-driven interest from US companies in Chinese AI models is a pragmatic, bottom-line decision that could reshape market dynamics faster than geopolitical rhetoric about ‘decoupling’ might suggest.

For Western readers: Western enterprises currently using or developing AI models should evaluate the cost-performance ratio of Chinese models, as their peers are already doing, and assess the strategic implications of adopting non-Western foundational AI technologies for their long-term data and supply chain strategies.

MIT Technology Review

Semiconductors & Hardware

Stackpole Expands Current Sense Resistor Series, Leveraging Asian Manufacturing Base

Stackpole is expanding its CSSU series of high-power metal alloy current sense resistors, adding 0.5 mΩ and 20 mΩ options to serve high-current applications like battery management and automotive electronics. The company, part of the Akahane Stackpole Manufacturing Group, operates facilities in Japan, Taiwan, and China, which are central to its global production and supply chain strategy. The article is straightforward product news. What’s more interesting is Stackpole’s manufacturing footprint, which includes facilities in Japan, Taiwan, and China. This reflects how deeply integrated these East Asian economies are in the global component supply chain, even for companies headquartered elsewhere. It’s not just about final assembly; it’s about specialized component manufacturing, too.

For Western readers: Western businesses building high-current power electronics should recognize that key component availability, even from US-headquartered suppliers, depends on manufacturing stability in Japan, Taiwan, and China, making their supply chains inherently exposed to East Asian industrial and geopolitical dynamics.

Electronics Weekly

Semiconductors & Hardware

US ITC Upholds Import Ban on Innoscience GaN Products in Infineon Patent Case

The U.S. International Trade Commission (ITC) has finalized an import and sales ban on products from Chinese GaN power chip maker Innoscience, upholding an earlier ruling that the company infringed on Infineon’s gallium nitride patents. This follows multiple favorable rulings for Infineon in German courts against Innoscience on similar patent infringement claims. Chinese firms like Innoscience have been aggressively expanding their GaN product lines to capture market share, often at lower price points. This ITC decision and the German court rulings create a significant hurdle for their market access in key Western economies, limiting a growth vector that might have allowed them to catch up on advanced materials technology. While Western coverage might emphasize the legal aspect, the practical effect for East Asian firms is about market access and competition strategy.

For Western readers: If you are a Western company relying on Innoscience as a low-cost GaN component supplier, assume that supply chain is now significantly restricted for US and German markets, and potentially other regions as Infineon continues its enforcement.

EE Times Asia

Semiconductors & Hardware

Wolfspeed Sues Navitas Over WBG Semiconductor Patent Infringement

Wolfspeed has filed a patent infringement lawsuit against Navitas Semiconductor in a U.S. court, alleging Navitas’s gallium nitride (GaN) and silicon carbide (SiC) products infringe multiple wide bandgap (WBG) semiconductor patents. The complaint specifically targets Navitas’s GaNFast, GaNSlim, GaNSafe, GeneSiC MOSFETs, and SiCPAK modules. This action follows a recent U.S. ITC ruling upholding an import ban on Innoscience GaN products in an Infineon patent case, highlighting intensifying IP disputes in the WBG sector relevant to East Asian manufacturing. Patent litigation among Western WBG leaders directly affects Asian suppliers and manufacturers, who are either licensing this technology or developing their own. The outcome of these cases dictates which process technologies and materials will be viable for Japanese and Chinese companies looking to scale their own EV and industrial power solutions, especially given the existing US-China technology friction.

For Western readers: Western businesses in WBG semiconductors should anticipate continued patent disputes as the market matures, potentially driving up licensing costs or limiting sourcing options for GaN and SiC components manufactured or integrated by Asian partners.

EE Times Asia

🧩 Pattern This Issue

• Korea/Taiwan: SK Hynix and Nanya scale capital expenditure to break memory cycles
• China: DeepSeek initiates custom AI silicon design to bypass NVIDIA bottlenecks
• Southeast Asia: IMF identifies Malaysia and Thailand leading global hardware exports

East Asia is aggressively expanding and localizing every tier of the physical AI hardware stack, ensuring that even if Western software leads, the physical infrastructure and supply chain control remain firmly anchored in Asian manufacturing corridors.