Anthropic/OpenAI may be spending more than $1000 for every $100 you pay them
ea.rna.nl
June 7, 2026
32 min read
🔥🔥🔥🔥🔥
44/100
Summary
Anthropic and OpenAI may incur costs exceeding $1000 for every $100 charged to users. Coding with Large Language Models (LLMs) is being promoted as a significant application for these technologies.
Key Takeaways
- LLM-coding is currently economically unviable for most applications, with costs potentially exceeding $1000 for extensive usage on a $100 subscription plan.
- Anthropic is working on updates to their models to address cost issues associated with LLM-coding, indicating a potential end to substantial improvements in this area.
- The combination of Claude Code and an experienced programmer can yield impressive results in code creation, but the economic feasibility remains a significant concern.
- Simple conversational uses of LLMs have become very affordable, but serious applications requiring complex reasoning and coding remain costly.
Community Sentiment
Positives
- Anthropic and OpenAI have developed highly efficient token-serving capabilities, allowing them to deliver services at half the cost of open-source providers, which strengthens their competitive edge.
- The trend of smaller, more capable models, such as Qwen3-30B outperforming GPT-4, suggests that ongoing advancements in AI will lead to increased profitability for companies leveraging these technologies.
Concerns
- Claims regarding Anthropic's profitability are misleading, as they often rely on cash flow metrics that do not reflect true profit, raising concerns about the company's financial transparency.
- The assumptions made in the analysis of API costs are flawed, particularly the neglect of caching mechanisms, which could significantly alter the cost dynamics of model inference.
