Well I was having so much fun with AI agents that the cost ballooned out of control. I had slowed down using my travel agent and office agent but they were still doing things in the background that were costing me money. This is another lesson learned, shutdown your gateways when you have no jobs running or expected to run to completely cut them off from burning tokens.
I have moved most of the AI models from the high end versions of Claude and ChatGPT over to Chinese versions of Deepseek, Kimi K2.6, MiniMax3 and other lower cost LLMs. No they aren’t as good but they offer more value as I can get more things done at a lower cost basis with those models than I can with U.S. ones.
At some point, I will shell out money to buy AMD’s new Halo Mini PC and start building my own infrastructure for everyday tasks and used cloud providers for specialized tasks.
China largely won the manufacturing war and they are headed to winning the AI war at this point.
AI Chatbot Projected Cost Inflation
The formula is the same as old as time. First, offer your customer a free taste of your good/service, get them addicted, then raise prices through the roof. I’ve written about TV Streaming Inflation here and here and now that I’m getting addicted to AI agents, it made me think what the future may bring to AI chatbot costs.
What better way to figure out future AI costs than to ask AI?
I asked an AI agent to look at the streaming inflation for Netflix and Spotify and figure out what AI chatbot costs may do into the future. I hope AI is right as it only projects a modest increase in prices but I suspect it will be much more unless hardware prices come down and people start their own AI setups at home the same way many now use their own Plex streaming servers to watch content instead of paying every growing Streaming app costs.
This article points out that Microsoft is realizing that AI is now becoming more expensive than paying humans to do the work. There were complaints that Anthropic’s latest LLM was burning through token limits within hours of use!
Large companies may be adverse to using products from China for their enterprise but I have no such qualms. I don’t have anything proprietary or top secret that I will be using AI for in this realm. I’m doing fairly routine and mundane tasks and for the sensitive work such as my trading bots, I will continue to use Claude for that endeavor.
I’ve written this before though that if push comes to shove and I need to cancel subscriptions, it will likely be Netflix or Spotify that I cancel before I cancel my core AI agents.
Share The Wealth
Have you run into AI cost problems yet? If you haven’t, just give it some time….