台積電擴建封裝廠，超微電腦出貨新款伺服器
TSMC expands packaging plants and Super Micro ships new servers
Updated at: September 28, 2026 at 03:45 AM
截至2026年年末，人工智慧革命正加速發展，並由台積電(TSMC)與美超微(Supermicro)兩大產業巨擘推動。
As of late 2026, the artificial intelligence revolution is accelerating, fueled by two major industry powerhouses: TSMC and Supermicro.
台積電目前正擴大規模其先進封裝基礎設施，以解決關鍵的供應鏈瓶頸。
TSMC is currently scaling its advanced packaging infrastructure to resolve critical supply chain bottlenecks.
他們的宏偉計畫包括在2028年前將CoWoS產能翻倍至每月26萬片晶圓，並在台灣嘉義科學園區建設新廠，同時向亞利桑那州擴張。
Their ambitious plan involves doubling CoWoS capacity to 260,000 wafers per month by 2028, with new facilities in Taiwan's Chiayi Science Park and expansion into Arizona.
這項封裝技術的進展對於容納次世代人工智慧晶片至關重要。
This progress in packaging technology is essential for housing next-generation AI chips.
與此同時，美超微依然處於硬體整合的最前線，推出了尖端伺服器平台，以支援輝達(NVIDIA)、超微(AMD)及英特爾(Intel)的最新處理器。
Meanwhile, Supermicro remains at the forefront of hardware integration, launching cutting-edge server platforms to support the latest processors from NVIDIA, AMD, and Intel.
透過採用先進的液冷系統和模組化的「積木式解決方案」(Building Block Solutions)，美超微確保了這些高效能晶片能快速部署至全球資料中心。
By adopting advanced liquid-cooling systems and modular 'Building Block Solutions,' Supermicro ensures these high-performance chips can be deployed rapidly into global data centers.
台積電在半導體封裝上的精湛技術與美超微的高效伺服器組裝，共同創造出一條強大的供應鏈，實現了現代人工智慧模型所需的高密度運算。
Together, TSMC’s mastery of semiconductor packaging and Supermicro’s efficient server assembly create a powerful pipeline that enables the high-density computing necessary for modern AI models.
這種綜效正有效地縮短前置時間，並克服了先前限制大規模人工智慧部署的硬體瓶頸。
This synergy is effectively shortening lead times and overcoming the hardware constraints that have previously limited large-scale AI deployment.
