CONNECT WITH US
Sep 18
High-end GPUs may be slowing China's AI compute expansion
China's wide gap in token costs stems largely from domestic companies' difficulty in accessing Nvidia high-end GPUs, forcing them to squeeze more value from each token to recoup costs. But supply-chain sources say the scale of China and US AI compute investment may actually be roughly comparable, despite US hyperscalers' far larger 2025 capex.

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.

Anderon LLC, an IBM company, has finalized a US$1 billion award under the US CHIPS and Science Act to expand quantum wafer manufacturing in Albany, New York. The move could help speed global progress in quantum computing by strengthening the supply chain for a technology with broad implications for industry, security, and research.

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.

AMAX expands liquid-cooled data center in push beyond hardware
Sep 21, 07:45

Amax is expanding its self-built liquid-cooled data center as the company moves beyond hardware into infrastructure and engineering services, with customer demand giving it more than a year of order visibility.

The US is reportedly pressuring Mexico to tighten rules of origin for AI hardware to limit the use of components produced outside North America, particularly in China. Server supply chain sources said the Mexican government began consultations with Taiwanese original design manufacturers (ODMs) around 10 days ago to explore mutually acceptable solutions, signaling a willingness to negotiate.

Digital transformation is no longer the endpoint for Taiwan's technology industries. As artificial intelligence moves from individual applications into core business operations, companies are entering a new phase in which AI, engineering models, and industrial knowledge are increasingly integrated into product development and operations, Dassault Systèmes said.

HEC Compucase AI server products drive 2026 revenue growth
Sep 21, 07:43

HEC Compucase's AI server-related rollout continues to gain traction as AI racks, cooling modules, and high-end power supplies move into mass production or volume shipments. The company expects fourth-quarter 2026 shipments and operating momentum to strengthen further, with full-year 2026 revenue set to maintain double-digit growth and first-half gross margin rising to 24%, about 3 percentage points higher than a year earlier.

High-speed transmission demand continues to rise on the back of AI server and data center buildouts, while cloud service providers (CSPs) accelerate adoption of in-house application-specific integrated circuit (ASIC) platforms. As a result, demand for high-speed connectors and high-speed transmission cables inside servers is also increasing.

Tyntek is rapidly expanding into optical communications as AI data-center demand drives a surge in high-speed transmission needs, with shipments set to rise significantly from the third quarter and gross margin and profitability expected to improve sharply. Backed by clear orders from a major US customer, the company has already launched an NT$1.2 billion (US$37.7 million) expansion plan approved by its board in August, aiming to triple monthly capacity to 1,500 wafers by the end of 2027.