大模型能干的事很多，智能体赚钱的其实不多

Core Insights - The article discusses the financial struggles of large AI models and their inability to generate profits, with companies like OpenAI projected to lose $14 billion annually until 2029 [1]. - It highlights the reliance on venture capital (VC) funding for sustaining operations in the AI sector, particularly in China where VC support is lacking [5]. - The article contrasts the profitability of certain AI applications, noting that while some companies like Midjourney are profitable, many AI agents face significant challenges in achieving financial viability [8][24]. Financial Performance of AI Companies - OpenAI is expected to be unprofitable until 2029, with annual losses of $14 billion [1]. - Anthropic is projected to reach profitability by 2028 but is currently also incurring losses [2]. - Newly listed MiniMax is losing $500 million annually, similar to another company, Zhipu [3]. - DeepSeek is reportedly profitable, but the overall trend indicates that foundational models are generally unprofitable [4]. AI Agents and Profitability - Glean, an AI agent, has an annual recurring revenue (ARR) of $200 million, but the company is likely to continue seeking additional funding due to the high costs associated with AI operations [6]. - Many successful AI agents are heavily reliant on large-scale financing and acquisitions to sustain their operations [7]. - Companies like Midjourney have demonstrated profitability with a small workforce, generating $500 million annually with only 40 employees, indicating a different operational model [8]. Challenges in AI Implementation - The article identifies two main barriers to profitability for AI agents: precision and cost [9][20]. - Precision refers to the ability of AI to reliably perform tasks in specific business contexts, where high-stakes decisions require near-perfect accuracy [11][15]. - Cost considerations include model costs, which vary between international and domestic models, and the additional expenses associated with personnel and operational overhead [20][21]. The Concept of "Unmanned Companies" - The emergence of "unmanned companies" is presented as a necessary evolution for achieving profitability without relying on VC funding [24]. - These companies would utilize a system of AI agents to drive business operations, with human roles being supportive rather than central [24]. - The transition to unmanned companies is complex and requires a rethinking of production relationships, emphasizing the need for AI to take precedence in operational roles [24][26]. Future Outlook - The article suggests that the pace of AI development may be slower than that of the internet due to ongoing technological instability and the complexities of restructuring production relationships [26]. - Companies are encouraged to consider long-term strategies rather than short-term gains, as the journey towards profitable AI applications is likened to a marathon [26].