Lesson 6: Behind the Scenes
Future spend plans
Big AI companies are planning to spend massive amounts of money on the computers, chips, electricity, and data centres needed to run powerful AI.
| Company / group | What they have said they may spend on AI |
|---|---|
| OpenAI | Could be part of a huge AI building plan worth up to $500 billion over several years. |
| Anthropic | Announced a $35 billion plan to build more AI computer power. |
| xAI | Has already spent billions building a giant AI computer system, and may need much more in the future. |
| Amazon | May spend around $200 billion in one year, with a lot going toward AI data centres. |
| Microsoft | May spend more than $100 billion a year on AI data centres and computer equipment. |
| Google / Alphabet | May spend more than $100 billion a year to build AI systems and data centres. |
| Meta | May spend over $100 billion to build AI computers and train future AI models. |
Training a model can take weeks or months. Epoch estimates it took 3 months to train GPT-4. Above are estimated costs for the electricity and hardware to run the model training, not including the teams of engineers, data collection or years of research. (From Stanford AI Index 2025 and Epoch AI).