Chinese optical transceiver suppliers are trying to calm investors after reports that the US Federal Communications Commission (FCC) may restrict new China-made data-center components, threatening a supply chain central to global AI infrastructure.
Drones, robots and self-driving vehicles are becoming the new proving ground for artificial intelligence. Machines are taking on more decision-making themselves, rather than relying on the cloud. The reason is speed. When a system has to react in milliseconds, waiting on a round trip to external servers is not an option, whether it's avoiding a collision or detecting an unauthorized drone. Elan Microelectronics chairman I-hau Yeh says this shift, moving AI processing from edge computing directly onto the device itself, is becoming essential for unmanned systems. Few applications demand more speed and accuracy than counter-unmanned aircraft systems, or C-UAS: technology built to detect and respond to unauthorized drones.
On August 5, Taiwan's National Science and Technology Council said it had completed the central government's 2027 budget plan, with NT$182.3 billion (US$5.7 billion) earmarked for technology spending, up about 9.5% from 2026. The biggest increase goes to sovereign AI computing power and infrastructure as Taipei pushes ahead with its Smart Nation 2.0 initiative.
The Trump administration is preparing to extend US technology restrictions deeper into the artificial intelligence supply chain, with the Federal Communications Commission considering a ban on new Chinese-made optical transceiver models used in data centers.
All-photonic networks (APN) are emerging as critical infrastructure for 6G, AI computing and smart cities. Nvidia's plan to expand its AI footprint in the US underscores why dark fiber is essential for distributed AI computing centers, according to Chau-Lyan Chang, director general of Taiwan's National Center for High-performance Computing (NCHC).
Mitsubishi Electric is establishing a cooling-equipment production company in Ohio, joining Daikin Industries in bringing manufacturing closer to US data center customers as delivery times, tariffs and local service become more important competitive factors.
Naver is considering an asset-light structure for a US$10 billion data center expansion under which a Brookfield-led investment vehicle would own GPUs and other infrastructure. At the same time, a Naver subsidiary would sell computing capacity and related services to customers.
Anthropic's US$10 billion computing agreement with seven-month-old startup Volta Infra Holdings Ltd. is the latest sign that the Claude developer is treating compute procurement less like a vendor relationship and more like a hedging strategy.
AMD expects the market for server central processing units (CPUs) to reach approximately US$220 billion by 2030, up from an estimate of more than US$120 billion three months earlier, as the company sees agentic artificial intelligence (AI) creating a new class of CPU-intensive workloads.
AMD CEO Lisa Su indicated during the company's second-quarter 2026 earnings call that an estimate of roughly US$30 billion in 2027 Instinct revenue may be too conservative, telling analysts that the data center AI figure was "probably too low."
AI demand is giving Taiwan's contract manufacturers a rare opening, and Wistron expects the opportunity to broaden in the second half of the year as new customers and products arrive. For global tech buyers, the shift signals tighter supply, firmer prices, and stronger leverage for a small group of server makers.
Roborock Technology, which started out in Xiaomi's ecosystem, has staged a comeback to become the global leader in robot vacuums. After launching its own brand in 2017, the company has spent nine years shedding the label of a Xiaomi contract manufacturer, and now holds a 27% global market share, far ahead of Xiaomi's 8.3%.
