AI computing is changing how global hardware is designed, built, and deployed, with implications that extend well beyond Taiwan's tech sector. At a September 19 event, Wiwynn chair Emily Hong said the industry now faces three central challenges: power, cooling, and high-speed interconnect, as demand for AI infrastructure rises worldwide.
China's fast-rising AI models are running into a compute supply bottleneck as demand for advanced packaging substrates surges, with ByteDance reportedly preparing to spend CNY100 billion (US$14.9 billion) a year on large-scale chip substrate purchases by 2030. Supply chain sources say the scramble for packaging substrate capacity is intensifying across Huawei HiSilicon, Hygon Information Technology, ByteDance, and major outsourced semiconductor assembly and test (OSAT) players, even as the industry races to expand output.
Acer Medical said Taiwan's and South Korea's long-standing health insurance and medical records could underpin a new wave of sovereign healthcare AI as the company moves beyond image recognition into generative AI and AI agents. The company highlighted uses in remote care, motion analysis, and preventive medicine at a Next AI event in Wonju, South Korea.
Faraday Future used its annual 919 launch event on September 19 to complete what it calls "Robot World 2.0," unveiling nine new embodied AI (EAI) devices and four industry productivity solutions, putting an all-new Futurist humanoid on sale, and pricing a Master Mini from US$9,990. The lineup spans humanoids, quadrupeds, and mobile manipulators in large, medium, and small sizes, which the company claims is the widest range of device forms of any US robotics company.
Acer founder Stan Shih stated that slowing artificial intelligence (AI) growth is not easy, as semiconductor capacity is still insufficient to support the current pace of AI development. He added that Taiwan's tech manufacturers have always been known for their flexibility and speed, and they will only become stronger in the face of fluctuations in technology cycles.
Anthropic is reportedly being asked to prove two opposite things at once. To list at the valuation its bankers have floated, the company must show public-market investors that Claude remains competitive among enterprise buyers. To keep the safety-first identity that separates it from OpenAI, it must show it is willing to move slower than it can. The next few weeks will show which of those claims it is prepared to compromise.
A new Anthropic threat intelligence report offers an unusual glimpse into how advanced AI systems are being used under heavy, real-world workloads. The findings suggest that the decision to buy or host AI servers may depend less on model quality alone, and more on which tasks consume tokens, how long they run, and how much operational state they must preserve.
