“5年烧掉3万亿美元”，越来越喂不饱的AI

Core Viewpoint - The article discusses the escalating competition among global tech giants in building AI infrastructure, highlighting the massive investments and the urgent demand for computing power in the AI sector, which is likened to a "power war" reminiscent of historical resource struggles [2][30]. Group 1: Investment and Spending - Current AI infrastructure spending by major companies has already surpassed half of the total expenditure on building the information superhighway over 20 years [24]. - NVIDIA announced a $100 billion investment in OpenAI for large-scale computing centers, marking a significant commitment to AI infrastructure [6][7]. - Morgan Stanley predicts that global spending on AI data centers will reach approximately $3 trillion over the next five years, equivalent to France's GDP in 2024 [7]. Group 2: Computing Power Demand - The demand for computing power in AI is experiencing explosive growth, with Microsoft Azure processing over 100 trillion tokens in Q1, a fivefold increase year-on-year [14]. - The training of top-tier AI models, such as ChatGPT-4, requires exponentially more computing power compared to previous versions, with a reported need 446 times greater than GPT-3 [14]. - By 2035, global computing power demand is expected to increase by 100,000 times, according to Huawei's report [15]. Group 3: Supply Constraints - The supply of computing power is constrained by chip availability and energy resources, with NVIDIA holding an 80% market share in AI chips but facing production and geopolitical challenges [15]. - Data centers are significant energy consumers, with projections indicating that by 2035, their electricity demand in the U.S. will double, representing the largest increase in energy demand since the 1960s [15]. - The rising electricity costs have led to a 267% increase in local electricity bills in areas with dense data center activity [15]. Group 4: Collaborative Strategies - Major companies are forming alliances to strengthen their positions in the computing power market, with NVIDIA's investment in OpenAI reinforcing its influence over AI chip supply [18][19]. - OpenAI's recent contracts with Oracle and NVIDIA illustrate a trend of strategic partnerships aimed at securing computing resources and enhancing operational capabilities [25][28]. - The competition has evolved from individual corporate strategies to a collaborative ecosystem where companies work together to establish industry standards and practices [29]. Group 5: China's AI Infrastructure Development - Chinese companies are also ramping up investments in AI infrastructure, with Alibaba planning to invest over 380 billion yuan in cloud and AI hardware over the next three years [32]. - The "East Data West Computing" initiative aims to optimize resource allocation across regions, addressing supply and demand imbalances in computing power [36]. - China's approach emphasizes a coordinated national strategy to enhance computing capabilities, contrasting with the more fragmented strategies of Western companies [40].