As of August 2024, major technology firms are significantly increasing their investments in AI data centers to address the surging demand for computational power.
Microsoft leads the industry with a total expenditure of $46 billion on AI capital and operating costs, maintaining 300 data centers. Amazon follows with approximately 215 facilities.
In September, Microsoft and BlackRock announced a $100 billion initiative through the Global Artificial Intelligence Infrastructure Investment Partnership (GAIIP) aimed at developing AI-focused data centers and their supporting energy infrastructure.
Both Google and Amazon are reportedly spending over twice as much on training AI models compared to the costs associated with running them for end users. The training of major AI models has become increasingly costly due to the requirement for extensive data sets and substantial computational resources.
This trend indicates that as AI model deployments expand, the overall costs for inference may soon surpass initial training expenses, a scenario already observed with OpenAI's ChatGPT.