Pat Gelsinger Forecasts 10,000-Fold AI Compute Demand
At a Playground press conference, former Intel CEO Pat Gelsinger forecast that global AI computing demand could grow 10,000-fold, while electricity and memory cannot scale at the same rate, requiring at least a 10,000-fold increase in compute efficiency. He said advanced packaging, chiplet-based products, HBM and DRAM have become central issues for AI systems, and rejected the idea of an AI bubble while predicting a 20-year-plus AI investment and expansion cycle.
The remarks focus on the infrastructure constraints and technologies associated with scaling AI computing, including power, memory, advanced packaging and chiplets.
由英特爾前執行長 Pat Gelsinger 作為合夥人的創投公司 Playground 於今（27 日）舉辦記者會中指出，AI 為台灣帶來助益，而台灣也為 AI 帶來助益。AI 工作負載推動台灣經濟加速成長，台灣也促成全球 AI 產業的驚人成長。
Gelsinger 在開場致詞中表示，放眼全球，沒有任何地方像台灣一樣，讓想法如此迅速且大規模地轉化為現實，「我常開玩笑說，早上吃早餐時提出一個想法，到中午台灣就已經做出原型，晚餐前甚至能開始初步量產」。他也認為，台灣是個特別的地方，建立了獨特的產業生態系，這是世界其他地方難以比擬的。