
Counterpoint Research預測伺服器AI ASIC出貨量將於2027年超越GPU,短期供應鏈合作預計維持穩定
輝達執行長黃仁勳12日辭職,震撼全球科技業。Counterpoint Research指出短期AI晶片產品路線及供應鏈合作可望維持穩定,但中長期需觀察新管理團隊策略。預測伺服器AI ASIC出貨量將於2027年超越GPU,AI運算市場將朝GPU與客製化ASIC並行發展。
AI-generated summary
輝達執行長黃仁勳12日辭職,震撼全球科技業。Counterpoint Research提出六大觀察重點,分析短期與中長期對AI晶片產業的影響,包括產品路線、供應鏈合作、客製化AI晶片發展及全球監管政策演變。
輝達(NVIDIA)執行長黃仁勳12日辭職,震撼全球科技業,也牽動AI晶片產業版圖變化。
〔記者林薏茹/台北報導〕輝達(NVIDIA)執行長黃仁勳12日辭職,震撼全球科技業。研調機構Counterpoint Research提出六大觀察重點,認為短期內NVIDIA既有AI晶片產品路線及供應鏈合作可望維持穩定,但中長期仍須觀察新管理團隊的策略方向、供應鏈合作模式,以及客製化AI晶片發展。Counterpoint預測,伺服器AI ASIC出貨量將於2027年超越GPU,隨著大型雲端服務供應商(CSP)持續發展自有晶片,AI運算市場將朝GPU與客製化ASIC並行的方向發展。
Counterpoint指出,NVIDIA從Blackwell到Vera Rubin的產品規劃已逐步建立,既有研發團隊及供應鏈合作關係,可望支撐未來18至24個月的產品推進。目前NVIDIA生態系涵蓋台積電先進製程與CoWoS封裝、SK海力士及美光HBM記憶體,以及鴻海、廣達、緯創等伺服器製造商,因此短期產品交付及供應鏈合作預料不會立即受到管理層異動影響,不過未來產品更新節奏若有調整,仍將牽動供應商產能規劃、研發投入及資本支出。
Counterpoint進一步分析,NVIDIA的競爭優勢已不再侷限於GPU,而是延伸至能源、晶片、基礎設施、AI模型與應用等多個層面,並透過CUDA軟體生態系形成整合優勢。隨著公司進入新的管理階段,新團隊能否延續跨技術平台整合能力、維持主權AI及雲端基礎設施合作夥伴投資計畫,將是影響未來成長的重要因素。此外,NVIDIA過去採取高度扁平化管理架構,未來組織職責重新分配後,核心技術人才留任、管理團隊穩定性及跨部門決策效率,也將成為觀察重點。
另一方面,客製化AI晶片發展也可能加速。隨著Google TPU、AWS Trainium等大型雲端業者自有AI晶片持續發展,AI運算市場正逐步形成GPU與ASIC並行的格局。Counterpoint研究副總監Brady Wang預測,伺服器AI ASIC出貨量將於2027年超越GPU,NVIDIA領導層變動也可能促使部分客戶重新評估供應商組合。不過,實際採購決策仍取決於運算效能、成本、軟體生態系及投資報酬率,此外,電力供應及資料中心建置進度,也將持續影響AI基礎設施投資與部署。
Counterpoint Research共同創辦人兼研究副總裁Neil Shah認為,NVIDIA此次領導層變動為AI產業帶來新的觀察變數,但不代表既有技術發展與市場需求將立即改變。短期仍以產品交付、供應鏈合作及客戶需求為主要驅動因素,中長期則須關注新管理團隊策略、AI晶片供應商競爭、客製化ASIC滲透率,以及全球監管政策演變。隨著AI產業從高速擴張轉向更重視規模化部署、成本效率與投資報酬,企業的技術整合能力及營運執行效率將成為下一階段競爭關鍵。
AI outlook — possibilities, not facts
伺服器AI ASIC出貨量將於2027年超越GPU
Likely · Within years
NVIDIA既有AI晶片產品路線及供應鏈合作在未來18至24個月內可望維持穩定
Likely · Within months

Lex Fridman, an American technology podcaster, posted on the

Professor Xu Zigui's team from National Pingtung University of Science and Technology developed "dual-target edge computing single-lens UAV smart vision following technology" and won the gold medal and the Best Invention Award at the 20th International Invention Exhibition in Warsaw, Poland. This technology uses AI and software innovation to significantly reduce hardware costs and improve drone endurance.
The 11th China Museum and Related Products and Technology Expo opened in Haikou. Guan Qiang, chairman of the China Museum Association, emphasized that AI is reshaping the museum industry, but it must adhere to the principle of "human beings taking the lead and AI supplementing", strictly guarding against historical distortions caused by technologicalism, and ensuring that technological applications serve the protection of cultural relics and the spread of civilization.

The "Disruptive Technology Industrialization Index Report (2026)" was released in Tianjin. The report shows that the industrialization of China's disruptive technologies has entered a period of rapid development, with ecological support and market transformation capabilities significantly improved. However, industrialization is still a common shortcoming, and regional competition is turning to a systematic contest of industrialization efficiency.

On November 13, the Ministry of Transport’s Key Laboratory of Low-altitude Transportation Equipment and Systems started construction at Beihang International Campus in Hangzhou. The laboratory is led by Tianmushan Laboratory and co-constructed by Beihang University and Beijing Jiaotong University. It aims to promote the low-altitude transportation field from a single breakthrough to system integration, and introduces academician workstations to strengthen safety assurance.

The State Administration for Market Regulation has released the national standard "Intelligent Manufacturing and Process Industry Smart Factory Business Management Specifications", which will be implemented on May 1, 2027. This standard aims to solve the problems of inconsistent management framework and insufficient system coordination in the construction of smart factories in the process industry. It covers five major modules including R&D, production and operation.