Rising Costs of Training AI Models Forecasted

Developing advanced AI models is becoming increasingly expensive, with costs expected to reach tens to hundreds of billions of dollars in the next few years, far surpassing the current training expenses of around $100 million to a billion dollars. Anthropic CEO Dario Amodei predicts that by 2025-2027, AI models could outperform humans in various tasks. The significant cost of training AI models is driven by hardware expenses, such as Nvidia's chips costing over $30,000 each, with concerns arising about the sustainability of the massive energy consumption associated with these models. Despite substantial investments by tech giants in AI, there remains a gap in revenue generation, with estimates suggesting AI firms need to earn around $600 billion annually to cover costs, raising questions about profitability and the need for quick returns to prevent an economic crisis. China aims to reduce the development of foundational large-scale AI models to five by 2028 to streamline competition and reduce energy waste.