OpenAI just cut the price of one of its models by 80%, to roughly $0.20 per million input tokens. The cost of running AI is collapsing. Two years ago, “Let’s add AI to this” could trigger a serious budget discussion. Today, for many applications, AI costs are becoming a rounding error.
When intelligence becomes cheap, the bottleneck moves. It’s no longer “Can we afford the model?” but “Do we know what problem to point it at?” That’s great news for students, researchers, startups, and small teams without the resources of big AI labs.
I’ve spent years optimizing models to run on limited hardware, including systems with just 8 GB of GPU memory, because access matters. Cheap APIs are attacking the same barrier from the other side. AI isn’t just getting smarter—it’s becoming cheaper, more accessible, and easier to experiment with. The next advantage may belong less to those with the biggest budgets and more to those with the best ideas and fastest execution.