Below are the most-read DIGITIMES Asia stories from the week of May 18-24, 2026:
As AI redraws chip industry, TSMC faces its first real rivals
The global semiconductor industry is entering a new phase as the AI boom, geopolitical tensions, and supply-chain security concerns push customers to reduce reliance on TSMC, even though it still dominates advanced chip manufacturing. Rivals such as Samsung Electronics and Intel are gaining interest from major AI chip customers, including AMD, Tesla, Google, and Nvidia, largely because governments and clients want geographically diversified production and lower geopolitical risk.
Samsung is leveraging its memory-chip leadership and aggressive pricing, while Intel is benefiting from strong political backing in the US and growing adoption of its packaging technologies.
Still, most advanced AI chip production remains firmly tied to TSMC because of its superior yields, manufacturing reliability, and advanced packaging capabilities. In response to rising competition and surging AI demand, TSMC has accelerated the largest global expansion program in its history, rapidly building new fabs and advanced-packaging facilities across Taiwan, Arizona, Japan, and Germany while preparing for future sub-2nm and post-1nm technologies. Despite increasing competition, industry executives believe TSMC will remain the central force in advanced semiconductor manufacturing for the foreseeable future.
Agibot claims 100% success rate in factory deployment as humanoid race shifts to real-world validation
At the 5th Mobis Mobility Day in Sunnyvale, Agibot stated that the humanoid robotics industry is entering a commercialization phase where deployment performance matters more than prototypes, highlighting a live-streamed factory deployment in which its humanoid robot reportedly handled more than 3,000 tablet-loading tasks with a 100% success rate over eight hours.
The company credited this achievement to its Genie Sim digital-twin simulation platform, which uses reinforcement learning to train robots virtually and dramatically reduce reprogramming time for new manufacturing tasks. Agibot also showcased its rapid hardware development cycle, progressing from its first prototype in 2023 to advanced humanoid models such as the A2 Ultra and the newly launched A3, which features lighter weight, improved agility, and dual hot-swappable batteries enabling 10 hours of continuous operation.
Beyond hardware, Agibot emphasized its open-source ecosystem strategy through its Link OS platform, Genie Studio tools, and a massive open robotics dataset published on Hugging Face, while also expanding into wheeled humanoids, quadruped robots, dexterous robotic hands, and retail-focused half-size humanoids, culminating in a live demonstration of its Lumi robot on stage.
Broadcom, Applied Materials, GlobalFoundries, Meta, and Synopsys back US$125M UCLA Semiconductor Hub
UCLA Samueli School of Engineering has partnered with Broadcom, Applied Materials, GlobalFoundries, Meta, and Synopsys to launch a US$125 million Semiconductor Hub aimed at accelerating AI-driven semiconductor research and workforce development over the next five years. The initiative will focus on AI-native chip design, advanced materials, packaging, energy efficiency, ultra-broadband communications, and next-generation computing architectures to overcome the limitations of traditional silicon scaling and strengthen US leadership in semiconductor innovation.
UCLA said the hub will unite academia, industry, and policymakers to align research with commercial needs while supporting doctoral research and yearlong internships at partner companies to cultivate engineering talent. Industry leaders described the collaboration as a first-of-its-kind effort spanning the full semiconductor ecosystem, from foundries and packaging to AI infrastructure and design automation, with potential applications in healthcare, transportation, data centers, edge AI, and advanced communications technologies.
India inaugurates Rajasthan's first semiconductor plant as local chip push expands
India inaugurated Rajasthan's first semiconductor plant on May 15 as part of its broader effort to strengthen domestic electronics and chip manufacturing, with Sahasra Semiconductors launching a new ATMP/OSAT packaging facility in Bhiwadi backed by the government's SPECS incentive program. The plant, developed with investment exceeding INR150 crore, is focused on packaging memory chips, Micro SD cards, flash storage products, LED driver ICs, eSIMs, and RFID components, with a current annual capacity of 60 million units expected to expand to as many as 600 million units within a few years.
Indian officials highlighted that more than 60% of production is already being exported to markets including the US and Europe, underscoring the facility's global orientation from the start. Alongside the semiconductor project, the new ELCINA Electronics Manufacturing Cluster has attracted over INR1,200 crore in planned investments from 20 companies spanning semiconductor packaging, EV parts, and industrial electronics.
Government leaders framed the developments as evidence of India's rapidly growing electronics sector, supported by initiatives such as the India Semiconductor Mission and production-linked incentives aimed at positioning the country as a global manufacturing hub.
DIGITIMES Insight: Intel and AMD diverge as TSMC prepares price hikes
According to DIGITIMES analyst Luke Lin, the global server CPU market is increasingly diverging between Intel and AMD, with Intel's recent data center revenue growth driven mainly by price hikes of 10% to 20% rather than shipment expansion, while AMD is seeing far stronger momentum through rising server CPU shipments and rapid revenue growth fueled by support from TSMC. Lin said AMD's server CPU revenue grew more than 50% in the first quarter of 2026 and could rise 70% in the second quarter, prompting the company to sharply raise its long-term growth targets.
He argued that Intel continues to struggle with manufacturing bottlenecks, yield challenges, and execution issues despite improved investor sentiment tied to US government backing and political support. Meanwhile, TSMC is reportedly preparing another round of price increases for 2027, particularly for advanced nodes at 4nm and below, as demand for AI-related chip capacity remains extremely strong.
Major customers, including Google and Nvidia, are said to be willing to accept higher prices and even provide large advance payments to secure production capacity, signaling continued tight supply conditions that could reshape data center economics and cloud infrastructure costs globally.
TSMC's COUPE push puts Samsung's silicon photonics ambitions under pressure
Samsung has launched silicon photonics foundry services and begun pilot production as it pushes into the rapidly growing AI optical interconnect market, with plans to mass-produce optical communication modules in the second half of 2026 and eventually offer full co-packaged optics (CPO) foundry services by 2029. The company is investing heavily in photonic integrated circuit (PIC) manufacturing, advanced bonding technologies, and integrated solutions combining memory, foundry, packaging, and silicon photonics to compete with TSMC.
However, industry observers say Samsung still trails significantly behind TSMC, which has already integrated electronic and photonic chips through its COUPE optical engine platform and is preparing mass production of advanced 200Gbps silicon photonics technology in 2026. Samsung also faces competition from established players such as Intel, STMicroelectronics, and GlobalFoundries in the PIC and optical module market.
Analysts suggest Samsung's current strategy may increasingly focus on supporting its own internal AI and logic chip ecosystem rather than directly challenging TSMC's broader foundry leadership, while South Korea as a whole still lags Taiwan in optical packaging infrastructure and manufacturing ecosystem depth despite expectations that optical processors could become a multibillion-dollar market after 2027.
Samsung's quiet Taiwan play: Pairing memory with foundry to chip away at TSMC
Lee Jae-yong reportedly visited Taiwan in May to meet with Rick Tsai as Samsung intensifies efforts to expand its foundry business and reduce customer dependence on TSMC in the AI era. According to supply chain sources, Samsung is attempting to leverage its strength in high-bandwidth memory (HBM) and DRAM by bundling memory supply with foundry services, a strategy already used to secure business from Tesla and regain some orders from AMD. Samsung is also reportedly willing to offer aggressive pricing models, including charging customers only for functional chips, to attract companies such as MediaTek.
Industry observers say the company is under growing pressure to strengthen its position across memory, foundry, advanced packaging, and AI server supply chains as Taiwan's ecosystem becomes increasingly dominant. While Samsung has made some commercial gains and is pursuing new AI and edge computing opportunities, analysts caution that displacing TSMC remains extremely difficult due to Samsung's lower yields, weaker manufacturing stability, and smaller advanced packaging scale, meaning any shift in major AI chip production is likely to happen gradually rather than quickly.
Article edited by Jack Wu
