Macro Playlist

The cost of using AI models has moved down by about 15% over the last 2 weeks.
 
After a run-up earlier this year, token prices are rolling over. That’s not a bug, it’s the system working.
 
Corporate users are becoming more disciplined and cost conscious about how they deploy AI. At the same time, suppliers are racing to improve efficiency and lower costs.
 
Not every task needs a frontier model. For example, if you’re asking for simple information or using AI as a “life coach,” cheaper and lighter models do the job just fine. Most users will not care in the end which model solved their problem, so why pay a premium to have access to the cutting edge? 
 
That’s where the real story is:

- Smarter allocation of compute
- Better model routing
- Falling marginal costs

Lower prices should support broader adoption. Broader adoption should lead to faster productivity gains. And the clearest path to that future goes through affordability.