Below are the most-read DIGITIMES Asia stories from the week of June 22-28, 2026:
South Korea takes physical AI push from policy to practice
South Korea has shifted its Physical AI Alliance into an execution-focused platform, restructuring it to accelerate the development and deployment of domestic physical AI technologies while expanding applications across industries such as manufacturing, logistics, healthcare, and defense.
The move coincides with deepening collaboration between the government, Nvidia and Naver, including plans to secure 260,000 advanced Nvidia chips, build AI factories based on Nvidia's Vera Rubin platform, expand large-scale AI infrastructure through Naver's Gak Sejong data center, and jointly develop a "Seoul World Model" for physical AI using Nvidia's Cosmos models and Naver's spatial data.
Together, these initiatives reflect South Korea's broader strategy to strengthen its AI ecosystem, reduce reliance on foreign technologies, and position itself as a global leader in physical AI.
Europe aggressively pursues non-red supply chains; Thunder Tiger showcases unmanned combat systems in Poland
Taiwanese UAV manufacturer Thunder Tiger is expanding its presence in Europe as demand grows for battlefield-proven, non-China supply chain defense technologies, using the Taiwan Expo in Europe 2026 to showcase its portfolio of attack drones, interceptor drones, long-range UAVs and unmanned surface vessels.
The company said its products, including the Overkill FPV drone with US Blue UAS certification and validation under the US Department of Defense's Drone Dominance Program, have attracted strong interest from European customers seeking rapidly deployable systems.
Thunder Tiger is also strengthening its non-China supply chain through partnerships with Taiwanese suppliers, establishing operations in Ohio, and pursuing local manufacturing, technology collaboration, and investment opportunities with governments and industry partners in Poland, Ukraine, and across the European Union.
TSMC 3nm lead times surpass one year as Samsung faces Intel in foundry push
Samsung Electronics is intensifying efforts to revive its loss-making foundry business, with industry sources describing the next two to three years as critical to achieving its 2028 turnaround target amid surging AI chip demand. The company is focused on stabilizing operations at its Taylor, Texas fab, improving 2nm manufacturing yields, and converting discussions with major customers including Google, AMD, and BYD into production contracts, while leveraging existing partnerships with Tesla, Nvidia, and Groq.
Although TSMC continues to dominate advanced-node manufacturing, its capacity constraints are driving customers to diversify suppliers, creating opportunities for Samsung. However, the company faces growing competition from Intel, making its ability to secure major AI chip orders and demonstrate reliable manufacturing capacity key to regaining momentum in the global foundry market.
Intel deepens ties with Taiwan's chip supply chain as October talks expand
Intel is expanding its collaboration with Taiwan's semiconductor supply chain as rising demand for AI and high-performance computing drives a recovery in its foundry business, with plans to deepen engagement through additional supplier exchanges later this year. The company is broadening partnerships with a growing number of Taiwanese equipment and materials vendors, particularly those already qualified by TSMC, to support its manufacturing expansion in the US and Malaysia, strengthen advanced packaging capabilities, and improve operational efficiency.
Industry sources see new opportunities for Taiwanese suppliers as Intel advances its 14A roadmap and its joint 12nm process platform with UMC toward mass production in 2027, although any move to more advanced nodes such as 3nm is expected to depend on the success of the initial collaboration and Intel's broader foundry turnaround, which CEO Lip-Bu Tan has said could take until 2030-2032.
Intel and AMD unveil ACE in x86 counteroffensive against Arm
Intel and AMD have jointly launched AI Compute Extensions (ACE), a unified x86 instruction set specification aimed at accelerating AI workloads on CPUs and strengthening the competitiveness of the x86 ecosystem as Arm-based processors gain ground in AI PCs and data centers. Developed through the x86 Ecosystem Advisory Group, ACE standardizes matrix acceleration across future Intel and AMD processors, enabling AI frameworks such as PyTorch and TensorFlow to optimize for a common software target while avoiding past compatibility issues.
The initiative reflects a rare collaboration between the two rivals to preserve x86's relevance in the AI era, positioning CPUs as an attractive platform for AI inference and enterprise deployments even as GPUs and Arm-based alternatives continue to expand their presence in AI computing.
AI data center buildout fuels optical interconnect race, but 6-inch InP wafers hit supply wall
Surging investment in AI infrastructure is accelerating demand for high-performance optical interconnects, prompting US optical communications companies Coherent and Lumentum to expand manufacturing capacity and advance 6-inch indium phosphide (InP) wafer production. While 6-inch substrates promise lower long-term production costs, the industry continues to rely on 4-inch wafers because of technical challenges, low yields, limited supplier capacity, and the fragile nature of InP materials.
As demand for AI networking components outpaces supply and export controls tighten, major technology companies such as Nvidia are increasingly securing access to critical substrates through long-term supply agreements, pre-booked capacity, and strategic investments, highlighting the growing importance of resilient optical communications supply chains for AI computing.
Samsung's Texas fab gains momentum with the arrival of ASML engineers
Samsung's Texas foundry is entering a critical new phase as ASML engineers arrive on site to support the installation and optimization of advanced EUV lithography equipment, signaling that the fab is progressing from infrastructure construction toward production readiness. The development supports Samsung's target of beginning commercial 2nm chip production in 2027, with the US$17 billion facility expected to manufacture its AI-optimized SF2P+ process technology and potentially produce Tesla's next-generation AI5 and AI6 autonomous driving chips.
The accelerated ramp marks a significant turnaround from 2024, when equipment deliveries were delayed amid weak customer demand, and underscores Samsung's renewed push to strengthen its foundry business and establish Taylor as a key base for advanced semiconductor manufacturing in the US.
Article edited by Jack Wu
