The cost of AI’s ‘compute’ is coming into focus, and it’s a lot

The AI buildout is highly expensive due to the significant costs of the hardware, data centers, and cooling systems required.
Initially, tech companies heavily subsidized the cost of AI systems to quickly get them to customers, resulting in less transparency about the actual costs.
There is now a shift towards a more transparent pricing model where companies are charged per data token used, exposing them to the true cost of AI systems.
Financial institutions and corporations, including banks, are facing a dilemma when they discover the actual high costs of using AI systems, leading some to search for cost-effective alternatives.
To reduce costs, companies are exploring various strategies such as building their own AI systems, using older models, or leveraging open-source solutions.
Increased transparency and the new pricing model put pressure on AI companies to turn a profit and manage costs more effectively, which conflicts with their previous strategy of heavy subsidization.
The total spending on AI buildout is estimated at $2 trillion or more, funded increasingly by banks and private credit, which heightens the pressure on AI firms to be profitable.
The higher costs and financial risks associated with AI could impact both investors and creditors.