OpenAI’s latest push to cut costs and boost efficiency is ruthless—GPUs’ costs have been slashed in half, cutting costs directly down the middle. The AI infra arms race is about to change.

CoinWorld News: OpenAI recently found an optimization solution that can cut model inference costs by more than half. According to a person familiar with the matter, OpenAI engineers told some colleagues earlier this month that, leveraging several newly developed optimization technologies, they successfully reduced the number of required Nvidia graphics processing units (GPUs) to just a few hundred, particularly after applying the technology in scenarios where visitors without free or paid accounts use ChatGPT.