Stanford University’s 2025 AI Index reveals that AI training costs have reached unprecedented levels, with Google’s Gemini 1.0 Ultra costing approximately $192 million to train. The comprehensive report, analyzed by Eliza Strickland for IEEE Spectrum, highlights both challenges and progress in the AI field through 12 significant graphs.
Despite rising training expenses, the cost of using AI is actually decreasing, with inference costs dropping dramatically. The report indicates U.S. companies remain leaders in producing notable AI models, although Chinese models are quickly closing the performance gap.
Environmental impact remains an important topic, with some models like Meta’s Llama 3.1 generating nearly 9,000 tonnes of CO2 during training, equivalent to the annual emissions of nearly 500 Americans.
Corporate investment in AI reached unprecedented levels in 2024, with private investment hitting record highs. However, businesses have yet to see transformative returns, with most reporting cost reductions below 10 percent and revenue increases under 5 percent.
The report also examines policy developments, noting that while numerous AI-related bills have been proposed in the U.S. Congress, very few have passed, with regulatory action shifting to state governments.
Public opinion shows surprising optimism, with 60 percent of global survey respondents believing AI will change their work, but only 36 percent expecting it to replace their jobs entirely.