I periodically review Chris Anderson's Free – a book about the move from physical to digital goods. It's more than fifteen years old but still has lessons for today's genAI environment.

One line in the book is particularly informative. Loosely paraphrased: *when the cost of some raw material is decreasing (pointing to zero) then you should build a business that assumes it's already zero. *

In other words: you want to design a business model for what's coming, not what it is now.

Anderson wrote that line about semiconductors and other goods required to build computers. These days, I see LLM usage through that lens.

Over the past couple of months we've seen companies plowing through tokens, using genAI as though the price is already zero.

The problem?

▶️ LLM usage prices are likely to rise in the future, not fall. ◀️

GitHub recently made news for shifting from a flat monthly fee to a metered, per-token structure. Some developers' monthly bills jumped 50x.

There are also (credible) industry rumors that the LLM providers are heavily subsidizing token prices, which would indicate that eventually everyone's bills will rise.

That uncertainty around token prices translates into uncertainty around your genAI plans. Is it wise to keep burning through tokens, with the expectation that prices will eventually move in your favor? Or are you better off rewriting your plans to protect yourself from future price increases?

▶️ The answer is: yes to both. ◀️

This is a good time to review your company's genAI efforts:

✅ Figure out where you're burning tokens on projects that aren't going anywhere. (Because even if prices fall, those projects are still a waste).

✅ Understand where you're better off not using genAI at all. (Maybe code or ML models would do a better job.)

✅ Decide which projects are worth keeping. (Then map out different price scenarios to understand how the future may turn out.)

(Do you need expert eyes on your genAI projects? An AI assessment can reveal lurking risks (like cost overruns) and missed opportunities (new use cases). Contact me to get started.)