Substrate shortages emerge as AI chip market's new bottleneck. Samsung Electro-Mechanics and LG Innotek are expanding AI substrate capacity as long-term customer deals fuel growth but demand still outstrips supply. Published November 7, 2026 - 7:00 a.m. Advanced semiconductor packaging substrates have emerged as a new supply chain bottleneck in the AI chip market, leading to intense competition among global chipmakers to secure critical components. Intel recently disclosed that shortages of packaging substrates have prevented it from fully meeting demand for data center CPUs in its third quarter report filed with the U.S. Securities and Exchange Commission (SEC). The company expects supply constraints for key components, including substrates and memory chips, to persist through next year. After the third quarter, Intel signed two agreements to secure substrate supplies, with estimated financial commitments ranging from $5.3 billion to $6.1 billion. Of that amount, Intel plans to pay approximately $3.3 billion by the end of 2027. The shortage stems largely from the increasing sophistication of AI chips. As CPUs and GPUs used in AI servers become more powerful, the substrates that connect them to other components must also become larger and more complex. Manufacturing these substrates requires precisely stacking multiple layers of circuitry, which makes production technically demanding and capacity expansion time-consuming. The flip-chip ball grid array (FC-BGA) is a prominent example. It is a high-value substrate that transmits electrical signals between semiconductor chips and motherboards. FC-BGAs are widely used in AI servers and data centers. Some customers have even begun helping substrate manufacturers finance capacity expansions while guaranteeing purchases over extended periods. Korean companies are also expanding production capacity to meet growing demand. Samsung Electro-Mechanics announced in September that it would invest a total of 6.78 trillion won ($5 billion) to expand FC-BGA production facilities at its Sejong plant in Korea and its subsidiary in Vietnam. The company plans to invest 4.3 trillion won in Sejong and 2.5 trillion won in Vietnam. The new Sejong facilities are scheduled to begin mass production in September 2028, while the Vietnam facilities are expected to start operations in June 2030. The investment is backed by financial support from global customers and commitments to purchase substrates over the medium to long term. The arrangement allows Samsung Electro-Mechanics to reduce the financial burden of its large-scale investment while securing future sales volumes in advance. Even after announcing its expansion plans, the company has continued to receive requests for additional orders from customers. During its recent third quarter earnings conference call, Samsung Electro-Mechanics said its planned capacity expansion would not be sufficient to meet all customer demand and that it was reviewing the timing and scale of further investments. The company's financial performance has also improved. Revenue from Samsung Electro-Mechanics' package solution division reached 939 billion won in the third quarter, up 58 percent from a year earlier, driven by increased sales of high-value FC-BGAs used in server CPUs, AI accelerators and networking equipment. LG Innotek is also stepping up efforts to expand its presence in the AI substrate market. In September, the company unveiled a large-format FC-BGA substrate for AI accelerators measuring more than 100 millimeters (3.9 inches) on each side and is discussing supply arrangements with potential customers. LG Innotek plans to begin mass production of substrates for AI training and inference applications in 2027 and expand into high-value server substrates by 2028. The company is also in talks with two major customers over long-term supply agreements and potential additional investments. Industry observers say competition in the AI semiconductor market has expanded beyond chip manufacturing to securing stable supplies of essential components. "Demand for AI semiconductors is growing rapidly, but substrate supply has been unable to keep pace," said Kwon Min-gyu, an analyst at Kiwoom Securities. "The shortage is likely to persist for some time, which should have a positive impact on the earnings of substrate manufacturers." BY PARK YOUNG-WOO [[email protected]]

Multinationals tap China innovation to power global growth. * by admin * November 6, 2026 Visitors pose for a photo in front of a billboard themed on the ninth edition of the China International Import Expo (CIIE) at the National Exhibition and Convention Center (Shanghai) in East China's Shanghai, Nov 5, 2026. [Photo/Xinhua] Multinational companies are increasingly leveraging China's innovation ecosystem to strengthen their presence in the country and beyond, with executives at the 2026 China International Import Expo in Shanghai highlighting how locally developed technologies are benefiting global markets. Stephen Ma, executive committee member of Nissan Motor and chairman of the management committee for Nissan China, said China is one of Nissan's three lead markets globally and also serves as a hub for the company's innovation and exports. "Customers overseas really benefit from our experience in China because China is undoubtedly the most advanced automotive market," Ma said. "Not only the new energy vehicle sector is huge, but the sector of intelligent technologies is also huge." He added that for Nissan global, the China market brings "additional ammunition or additional leverage to compete", fusing with many of the company's overseas projects. At the expo, Nissan is showcasing its latest lineup, including the new NX8 SUV in both battery-electric and range-extended versions, alongside two vehicles destined for export markets. Among the expo's "debut products", many are original R&D achievements reflecting a "in China, for the world" approach. The US tech company Intel brought its "Green Computing Pod", which integrates green electricity storage with computing chips inside a container and can be put into production in as little as one week. The product originated from ideas generated during research in China. Alan Wang, chairman of Intel China, said that China-developed research benefiting the world is becoming the new normal, and the company looks forward to showcasing original chip-level products jointly developed with Chinese technology companies at next year's expo. According to him, Intel is a multi-year CIIE participant. This year, it has also brought an enterprise-level all-in-one token machine that can be easily deployed in small and medium-sized enterprises, as well as an Agentic PC experience zone demonstrating continued cooperation with domestic open-source models and the agent ecosystem. Tanks to chinadaily.com.cn Please visit: Its Sponsor

Intel is supplying hardware-backed security for Atsign's new EdgeAI platform, which handles AI and IoT networking. The platform pairs high-throughput data handling with encrypted, device-level protection at the network edge. Separately, Symbolic Mind has reported efficiency gains in AI training and inference using Intel Xeon processors for on-premise enterprise workloads. The deals support the bull case that Intel can win AI workload share with Xeon CPUs and built-in security, rather than only chasing GPU-heavy setups led by Nvidia and AMD. They echo the Oracle Cloud Infrastructure deal. Risks remain. These are early design wins, set against weaker PC guidance, foundry losses, and heavy capital spending on US fabs, 18A and 14A, and advanced packaging.