豆包们，开始“上链接”

Core Viewpoint - The article discusses the evolving monetization strategies of closed-source AI models, highlighting the shift from traditional methods to new approaches like e-commerce and advertising, while emphasizing the challenges posed by high operational costs and intense competition in the industry [6][7][20]. Monetization Strategies - Closed-source AI models are exploring new monetization avenues, particularly through e-commerce and advertising, as traditional methods like API sales and subscriptions are insufficient to cover rising costs [7][9][10]. - E-commerce is viewed as a more favorable option compared to advertising, as it focuses on transactions rather than mere exposure, making it more acceptable to users [10][11]. - Major domestic AI assistants have integrated e-commerce links, allowing users to make purchases directly from responses, indicating a trend towards embedding shopping functionalities [12][14]. Current Trends in E-commerce - Companies like Doubao and Tencent Yuanbao are embedding shopping links in their responses, with various platforms like Taobao and JD.com being utilized for product recommendations [12][14]. - The integration of e-commerce links is considered a basic level of monetization, but it faces challenges in differentiating from existing e-commerce platforms [15][19]. - Internationally, OpenAI has introduced shopping features in ChatGPT, allowing users to make purchases directly, which could redefine the AI's commercial value [16][17]. Traditional Monetization Methods - The three primary traditional monetization methods remain API sales, subscription services, and customized enterprise solutions, which still constitute the bulk of revenue for AI companies [21][30]. - API sales are a significant revenue source, with companies like OpenAI and Anthropic leading in this area, generating substantial income from API usage [22][24]. - Subscription models, while successful for OpenAI, face challenges in user adoption, particularly in the domestic market where willingness to pay is low [27][28]. Cost Challenges - The high costs associated with training and inference are a major concern for AI companies, with OpenAI reporting significant losses despite high revenues [32][34]. - Training costs have skyrocketed, with estimates indicating a rise from several million dollars in 2020 to over 300 million dollars by 2025, reflecting a 66-fold increase [34]. - The industry is experiencing a "burn rate" scenario, where companies must continuously invest to improve models and retain users, leading to a cycle of high expenditure and low profit margins [37].