>Transformative maybe, but valuation shouldn't be about transformative, it should be about profitable.

My assumption is that LLMs would be ubiquitous right now, just GPT4, not even GPT5, if inference cost is low enough.

Therefore, I expect the market for LLMs to expand exponentially in the future once inference cost is solved.