At the Apsara Conference 2025, Alibaba introduced its PanJiu AI Infra 2.0 with a 128-chip supernode, the most-watched launch of the event. The debut underscores how China's cloud leaders are pushing independent AI infrastructure innovation and moving beyond Nvidia's shadow.

Wiwynn officially began construction of its new headquarters in Neihu District, Taipei, during a ceremony held on October 2. The event was led by chairperson and CSO Emily Hong, with Taipei Mayor Wan-an Chiang and other industry figures in attendance. The company aims to complete the building by 2029 as part of its expansion strategy.

When Intel and Nvidia announced their partnership on September 18, 2025, the US$5 billion equity stake dominated headlines. But beneath the financial fanfare lies a far more consequential shift: Intel is betting its future on a CUDA-first architecture. While mainstream media fixated on the dollar figures, technical forums like SemiWiki and analysts at The CUBE Research began unpacking what this collaboration truly means for the future of computing.

Huatie's CNY3.69 billion (US$518.3 million) computing contract ended without delivery, indicating potential cancellations in China's computing market. Speculative contracts amid rising AI demand and US GPU restrictions have led to supply shortages and financial risks, possibly benefiting Taiwan's server and cloud sectors.

OpenAI CEO Sam Altman arrived in South Korea on October 1 to advance cooperation agreements with leading conglomerates SK Group and Samsung. His visit included meetings with SK chairman Chey Tae-won and Samsung chairman Lee Jae-yong, underscoring the country's growing role in artificial intelligence development.

Google DeepMind has announced the release of two updated AI models, Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, aimed at improving robots' ability to reason, plan multi-step tasks, and use online information to complete assignments. These models also facilitate the sharing of skills across different robotic devices, pushing the boundaries of autonomous robot functionality.

Consumer electronics like notebooks are losing momentum amid a sluggish 2025 market. However, AI-driven server demand is surging, providing sustained growth for ODMs.

Quantum technologies, long confined to academic theory, are now laying the groundwork for a revolution in navigation. While the term "quantum computing" often raises fears about data security, the underlying physics is enabling next-generation navigation systems for defense and civilian use.

Qualcomm's annual Snapdragon Summit officially concluded last week and benchmark scores for the new products are now being released. Overall, Qualcomm has delivered significant upgrades in system-on-chips (SoCs) for both flagship smartphones and flagship PCs compared to the previous generation, and the new chips are expected to perform well in upcoming partner devices.

Shanghai Zhaoxin Semiconductor has unveiled the Kaisheng KH-50000 server CPUs, aiming squarely at AMD's EPYC line. Debuting at Alibaba Cloud's Apsara Conference 2025, the chips were pitched not just as a technical breakthrough but also as a symbolic "National Day gift" to underscore China's advancing domestic compute ambitions.

CoreWeave, a cloud computing services provider, disclosed in a US Securities and Exchange Commission filing on September 25 that it has entered into a new order form with Meta Platforms under their existing Master Services Agreement. The deal commits Meta to pay up to US$14.2 billion through December 2031 for reserved cloud computing capacity, with the option to expand through 2032. The agreement positions CoreWeave as a key infrastructure partner for Meta, providing access to Nvidia's latest GB300 systems to support AI workloads and Meta's growing technology initiatives.

OpenAI is intensifying efforts to expand its artificial intelligence computing infrastructure, relying heavily on Taiwan's semiconductor supply chain. On September 30, OpenAI CEO Sam Altman conducted a discreet visit to Taiwan, holding separate meetings with TSMC and Foxconn. The discussions focused on collaboration in AI chip manufacturing and AI server infrastructure development.

Techman Robot chairman Shi-chi Ho highlighted that rapid advances in artificial intelligence (AI) in recent years, from large language models (LLM) to vision-language models (VLM), as well as the recent integration of physical AI, have transformed the robotics industry from one of fantasy into reality. While robots will eventually enter households, initial applications will focus on practical deployment, with Techman targeting industrial use cases.

Compal Electronics' PCB subsidiary, Allied Circuit Co., Ltd. (ACCL), has announced a joint venture with Japan's Meiko to establish a high-layer count (HLC) printed circuit board plant in Vietnam. The new entity, tentatively named AlliedCircuitMeikoVietnam, will set up operations within Meiko's new factory complex in Hoa Binh Province, northern Vietnam.

Qualcomm has launched its new Snapdragon X2 Elite PC chip platform in 2025, aiming to expand its reach in the AI PC market. Mandar Deshpande, senior director of product management at Qualcomm, said during an interview that beyond advancing core CPU, GPU, and NPU technologies, the company places greater emphasis on delivering superior consumer experiences.

DeepSeek unveiled its next-generation large language model, DeepSeek-V3.2-Exp, on September 29, 2025, and released it as open source on Hugging Face. The new model features a sparse attention architecture that reduces computational resource consumption while improving inference efficiency, especially for long-text tasks, advancing China's AI software-hardware collaboration.

Concerns about a possible artificial intelligence (AI) investment bubble have emerged in the US technology sector, with Meta Platforms Inc. CEO Mark Zuckerberg acknowledging the risk while supporting aggressive spending to avoid missing out on AI's transformative potential. Zuckerberg made these remarks during a recent podcast, as reported by Business Insider and Fortune.

Schneider Electric and Nvidia have announced a global strategic partnership to advance the development of infrastructure supporting artificial intelligence workloads. The collaboration was unveiled at the Nvidia GTC conference in Paris.

A serious vulnerability in the Bluetooth Low Energy (BLE) Wi-Fi configuration interface of several Unitree robots could allow attackers to gain root-level control, researchers disclosed on September 20, 2025. The exploit, dubbed UniPwn, affects Unitree's Go2 and B2 quadrupeds, as well as G1 and H1 humanoids. Because the flaw operates over wireless connections, it is "wormable," meaning an infected robot could automatically compromise others within BLE range, creating a self-propagating robot botnet.

Driven by the global AI boom, the Southern Taiwan Science Park (STSP) surpassed the NT$1 trillion revenue mark in the first half of 2025. The total operating revenue reached NT$1.38 trillion (US$45.54 billion) in the first half, representing a significant year-over-year growth of 48.78% compared to the same period in 2024.

As fighter jets patrol the skies and geopolitical tensions rise worldwide, a new arms race is taking place inside Taiwan's leading technology sector. The island's industrial PC makers—known for building ultra-durable computers used in factories and vehicles—are now shifting their focus to supply the digital backbone for the next generation of smart warfare, including AI-powered drones and unmanned naval ships.

Quanta Computer is expanding its manufacturing capacity in the US through an NT$1.336 billion (US$45.6 million) investment to lease a factory in California. The company's subsidiary, Quanta Manufacturing Nashville, LLC, signed a 10-year lease for a manufacturing facility measuring approximately 502,800 square feet.

Microsoft has introduced microfluidics cooling, a chip-level technique that etches microchannels into silicon, allowing coolant to flow directly across the chip surface. The company states that the method could deliver up to three times the efficiency of existing cold plates and eventually render traditional cooling modules obsolete. Yet cooling solution vendors remain unconvinced, stressing that whether the tech can be volume-produced is the real challenge, with an industry insider even believing it may take up to 10 years to happen.

At Qualcomm's 2025 Snapdragon Summit marking its tenth anniversary, CEO Cristiano Amon highlighted in his opening keynote that this is a pivotal moment for Snapdragon to enter a new phase. He emphasized that the future will redefine all mobile devices in the era of artificial intelligence (AI), establishing what he calls an "ecosystem of you."

Foxconn has jumped out to a significant early lead over rivals Quanta Computer and Wistron in the high-stakes race to build Nvidia's next-generation AI servers, according to industry sources. The move positions the company to dominate the production of the world's most advanced computing hardware as cloud providers, tech companies, and entire nations escalate their investments in AI.