생성 AI 서비스의 비즈니스 모델 유형화와 수익성 결정요인: SaaS·API·임베디드 모델 비교 = Typology of Business Models and Determinants of Profitability in Generative AI Services: A Comparison of SaaS, API, and Embedded Models

다국어 초록 (Multilingual Abstract)

This study classifies emerging business models of generative AI services and analyzes their key profitability determinants. Unlike traditional digital services, generative AI involves usage-based costs and risks, including input and output tokens, model calls, inference costs, retrieval-augmented generation, security, copyright, and hallucination risks. This study categorizes generative AI business models into SaaS, API, and embedded models and compares actual services such as ChatGPT, Perplexity, OpenAI API, Anthropic Claude API, and Google Gemini API.
It also proposes a profitability framework, consisting of value creation, cost-to-serve, usage monetization, and risk adjustment, and presents detailed profitability models based on customer lifetime value, token spread, model routing, and price premium. The analysis shows that customer retention and usage control are critical for SaaS models, token margin and call volume for API models, and existing customer base and workflow integration for embedded models. This study provides a theoretical classification and analytical framework for pricing, commercialization, and investment evaluation of generative AI firms.