Sam Altman says OpenAI is testing new features by throwing 'a lot of compute' at them (www.businessinsider.com)

OpenAI CEO Sam Altman announced the company will roll out "compute‑intensive offerings" in the coming weeks, framing the push as an experiment to see "what's possible when we throw a lot of compute." Because these features are costly to run, some will be limited to Pro subscribers and others may carry extra fees. Altman emphasized the company's longer‑term goal remains driving down the cost of intelligence and widening access, but for now OpenAI is prioritizing rapid iteration on high‑cost ideas that require large-scale GPU usage. The move highlights the intensifying GPU arms race: OpenAI has publicly targeted more than 1 million GPUs by year‑end, while rivals like xAI and Meta are already operating massive clusters (xAI disclosed ~200k GPUs). Technically, this signals experiments with scale‑heavy capabilities — e.g., larger or more compute‑hungry models, higher‑resolution or longer‑context multimodal inference, or sampling‑intensive features — that are expensive both to train and serve. Short term, expect tiered access and higher prices for premium features; longer term, the lessons from these experiments could drive efficiency gains and new capabilities, but they also reinforce infrastructure and cost barriers that shape who can build and use cutting‑edge AI.