Domestic 10k-Card Cluster Unveiled: A domestic manufacturer showcased an ultra-large-scale computing cluster capable of deploying over ten thousand AI accelerator cards, with total computing power exceeding 5 EFLOPS. Some of its technologies are claimed to be leading.
Nvidia Plans to Resume Sales to China: According to reports, Nvidia plans to resume sales of AI chips, such as the dedicated H20 model, to the Chinese market to address supply shortages.
Tech Giants Challenge Software Ecosystem: Google and Meta are collaborating on a project aimed at optimizing the performance of mainstream AI frameworks on their in-house developed chips, seeking to break the monopoly of the existing dominant software ecosystem.
Nvidia Evaluates Production Increase: Due to strong demand for advanced AI chips from the Chinese market, Nvidia is considering increasing production capacity for models like the H200.
Domestic Chip Companies List Densely: Recently, several domestic AI chip design companies have completed listings or are advancing related processes on the Sci-Tech Innovation Board, showing active market performance.
Professional GPU with Large Memory Released: Nvidia launched a new professional-grade GPU with significantly increased video memory capacity to meet the demands of processing larger AI models.
Domestic Industry Focuses on Open Standards: The domestic industry chain emphasizes the establishment of unified, open hardware and software architecture standards to reduce application migration costs and promote ecosystem collaboration.
Huawei Discloses Chip Roadmap: Huawei disclosed its product plan for Ascend AI chips over the next few years, with multiple new products expected to be launched between 2026 and 2028.
Domestic Cloud Providers' Capex May Rebound: As the tight supply of AI chips is expected to ease, analysts predict that capital expenditure from major domestic cloud service providers will resume growth.
Progress in Domestic Non-GPU Approaches: Beyond the traditional GPU path, rapid progress has been made in the R&D of domestic AI chips following other technological paths, such as reconfigurable computing and TPUs. Some companies have announced the mass production of self-developed TPU chips.