Anthropic and OpenAI may be spending more than $1000 for every $100 they receive from users. This reflects high operational costs and intensifying competition in AI.
Analysis suggests that Anthropic and OpenAI may be spending more than $1000 for every $100 they receive from users. This highlights the high operational costs and competitive pressures in AI.
AI companies face profitability challenges due to massive computing resources and labor costs. Training and inference costs for large language models are particularly high, leading to a significant gap between revenue and expenditure.
This cost structure raises questions about the sustainability of the AI industry, suggesting a need for price adjustments or cost-efficiency technologies in the long term. It also calls for cautious approaches from investors and the market.
HN users point out that AI inference costs are high, and predict that in the future, dedicated hardware cards with hardcoded model weights will emerge, enabling high-speed inference with a one-time purchase without subscription fees. However, some worry that because companies prefer recurring subscription revenue models, such hardware may not become widespread, and there could be restrictions like vendor lock-in or per-token pricing. Others suggest that current cutting-edge models will remain exclusive to large enterprises, and open-source models need to develop sufficiently for such hardware to become practical.