CONNECT WITH US
Dec 26
Huawei details Ascend AI chip roadmap built around in-house HBM, massive clusters
Huawei is clarifying how it intends to compete in global AI computing despite being cut off from leading-edge foundries and US-origin GPUs. Instead of chasing rivals on single-chip performance, the company is leaning into scale, systems engineering, and vertical integration—a strategy it is now preparing to test outside China, beginning with South Korea.
In 2025, the global electronics supply chain transformed due to rising demand for artificial intelligence (AI) and high-performance computing (HPC), boosting servers, chips, and cooling sectors. Geopolitical tensions, export controls, and multi-location approaches also reshaped the supply chain to improve flexibility and risk management. The DIGITIMES Asia news team has summarized the top 10 key developments from this shift.
Japanese firms like Rohm and AOI Electronics built partnerships with Indian partners. Foxconn is mimicking its strategy in China, setting up a factory city in India.
South Korean PNT has secured its first mass-production order for battery copper foil in China, marking a strategic expansion beyond its core battery equipment business into materials and strengthening its foothold in the Chinese market.
The development paths for artificial intelligence (AI) integration in smartphones are diverging as ByteDance extends its collaborations with multiple brands, while technology giants Apple Inc. and Alphabet maintain a standardized application programming interface (API) strategy. Following the December 2025 launch of the Doubao phone M153 in partnership with ZTE Corporation's Nubia, recent reports indicate that ByteDance is broadening its AI smartphone alliances to include Vivo, Lenovo Group, and Transsion.
Market attention in the modern AI computing power competitive landscape has mainly focused on semiconductor chip upgrades. But according to Chih-Ming Lin, chairman of Tatun Electric, the real challenge lies elsewhere. The company, with more than 70 years of industry experience, believes AI will not turn into a bubble—provided power transmission issues are effectively resolved.
Taiwanese pneumatic components maker Chelic said it is redirecting its growth strategy toward semiconductors, liquid cooling, and robotics, as new sensing and energy management products move into customer testing with revenue contributions expected to begin in early 2026.
Google is advancing its TorchTPU initiative to optimize PyTorch performance on its proprietary TPU chips, as reported by Reuters. This effort, undertaken in collaboration with Meta Platforms Inc., aims to reduce developers' switching costs and disrupt Nvidia Corporation's leading position in AI infrastructure.
Global management consulting firm McKinsey & Company has released its "Global Banking Annual Review" and "The State of AI: Global Survey 2025," based on a survey of 2,000 companies across finance, technology, retail, and other sectors.
As generative AI advances, applications are shifting from simple content creation to autonomous action. AI agents have become central, focusing on understanding goals, making independent decisions, and completing tasks with minimal human input. This marks AI's evolution from a tool to an intelligent system that works autonomously.
Apple Inc. has unveiled an AI imaging technology called DarkDiff, designed to improve photo clarity in low-light conditions by integrating generative diffusion models into the camera's image signal processor (ISP). While effective in reducing blur and enhancing detail, the technology's high computational demands currently limit its deployment on consumer devices.
OpenAI launched its AI browser, ChatGPT Atlas, allowing AI agents to access web pages and assist users with tasks such as editing emails. However, this innovation comes with heightened cybersecurity threats from prompt injection attacks, which OpenAI concedes are challenging to fully eradicate.