The cost of launching each new level of AI decreases by approximately tenfold annually, potentially leading to a sharp drop in prices for certain goods. This was stated by OpenAI CEO Sam Altman in his blog.
“In some ways, AI may prove economically similar to the transistor—a major scientific breakthrough that scales well and permeates virtually all sectors of the economy,” he noted.
According to Altman, the price per token in the GPT-4o model (2024) has dropped approximately 150 times compared to GPT-4 (2023).
The developer also added that “the socio-economic value of linear AI growth is inherently super-exponential,” meaning “there is no reason to halt the exponential growth of investments in the near future.”
The implementation of AI agents is now beginning, noted the CEO of OpenAI. Currently, there are only thousands, but soon the number will exceed 1 million, and artificial intelligence will permeate all areas of life—from politics to economics:
“Prices for many goods will eventually drop sharply (currently, the cost of [training] AI and energy limits much), while prices for luxury items and some initially limited resources, such as land, may rise even more,” Altman emphasized.
Back in late January, the top manager described the much-discussed AI model DeepSeek-R1 as “impressive,” considering its capabilities and the resources spent on training.