AI is getting cheaper, but your computer is getting more expensive: How OpenAI and Anthropic started a price war

In September, the AI market saw a significant shift as OpenAI and Anthropic released new models with significantly lower usage costs than their predecessors. Simultaneously, DeepSeek introduced cached input pricing at $0.003 per million tokens. The article analyzes how this price reduction is driven not only by engineering innovations, such as KV-cache optimization and improved model efficiency, but also by shifts in the financial models for computing power. Despite the falling cost of AI services, hardware prices—including DDR5 RAM and H100 GPU rentals—continue to rise. The piece examines the drivers behind this price war, the impact on the open-weights ecosystem, and the current state of high-performance hardware availability. The article provides a deep dive into the economics of the modern AI industry and how companies are balancing lower user costs with rising infrastructure expenses.
This is a summary. Read the full article at the original source:
HabrRelated stories
SGLang Without Magic: How Core LLM Inference Settings Work
This article from Ecom Tech on Habr provides an in-depth analysis of the configuration settings for SGLang, a popular LLM inference framework. The aut…
Google's experimental Playground platform uses AI to create games for you
Google has introduced an experimental platform called Playground, which leverages generative artificial intelligence to simplify the game creation pro…
Non-obvious AI capabilities: Why one should remain optimistic
The author shares their personal experience with artificial intelligence, noting that many specialists underestimate the true potential of modern neur…


