The 2025 AI Index Report published by Stanford University revealed the dramatic cost changes in the field of artificial intelligence and the end point of global competition. According to the report, the cost of using high-level AI language models has dropped from $20 to $0.07 per million tokens in just 18 months.

The cost of artificial intelligence has dropped 280 times

The sharp decline in the cost of use was driven by the proliferation of smaller, optimised AI models and a drop in hardware costs. The cost of enterprise AI hardware fell by 30 per cent last year. New generation processors were found to be 40 per cent more energy efficient.

While this decrease significantly increases accessibility on the user side, the opposite direction is followed on the model training side. The costs of training next-generation large language models are increasing exponentially. Technology giants such as OpenAI, Google and Meta have allocated record resources to training flagship models.

The report also comprehensively assessed the current state of global artificial intelligence competition. The United States was the country that allocated the most resources to artificial intelligence and developed the most models. China ranked second with 15 models, while Europe was able to find a place in the list with only 3 models.

In terms of performance, the differences are closing. In blind tests at the LMSYS Chatbot Arena, the best US model outperformed the best Chinese model by only 1.7 per cent. Similar results were found in industry-recognised benchmark tests such as MMLU and HumanEval

Stanford’s report revealed that artificial intelligence has become a rapidly maturing industry on a global scale. So what do you think about this issue? You can share your opinions with us in the comments section below.