Core Insights - The article discusses the significant shift in the internet paradigm from a focus on user scale to the cost of energy and computational power, highlighting the implications for AI companies and their business models [1][2][3][4]. Group 1: Transition from Scale to Energy - OpenAI's indefinite suspension of Sora's free testing service signals a critical change in the industry, emphasizing that the cost of computation and energy consumption is now a primary concern [1]. - The traditional internet model, which relied on fixed costs and low marginal costs, is being challenged by AI products like Sora, where each user interaction incurs substantial computational costs [2][3]. - The growth in user base is no longer a profit lever but a cost amplifier, leading to increased losses for AI applications despite high revenue growth [3][4]. Group 2: Cost Structure Reconfiguration - The cost structure for AI companies has shifted from one-time capital expenditures to ongoing operational costs, similar to energy companies, which pressures profit margins [5][6]. - The phenomenon of "negative scale effects" emerges, where increased user engagement leads to higher operational costs, complicating the path to profitability for AI firms [6]. - Competitive advantages are transitioning from algorithmic capabilities to energy and infrastructure efficiency, with companies focusing on reducing the cost per token generated [6][7]. Group 3: Future of Internet Competition - The future of the internet may revolve around energy monopolies rather than user traffic, with companies that can access low-cost energy gaining a competitive edge [7][8]. - Investment trends are shifting towards energy assets and data center infrastructure, with companies like Amazon and Google securing long-term power purchase agreements to ensure energy supply [8][9]. - The geographical landscape of internet infrastructure is changing, with energy-rich regions becoming new hubs for computational power, potentially leading to a decentralization of internet architecture [9][10]. Group 4: Philosophical Implications - The article posits that the evolution of productivity is tied to energy efficiency, suggesting that the future of AI will require a balance between intelligence generation and energy consumption [10][11]. - The concept of "smart restraint" emerges, indicating that the pursuit of unlimited generation may be unrealistic, and a more calculated approach to energy use will be necessary [11].