[redwood]

This is hackernews, not some boilerroom pump n dump forum. Please use more professional language and take your confidence down a notch. Try to learn and add to the discussion.

You seem to believe that the more inference or training value per piece of tech the more demand there will be for that piece of tech full stop when there are multiple forces at play. As a simple example, you can think of this as a supply spike; while you can make the bet that the demand will follow there could be a lag on that demand spike due to the time it takes to find use cases with product/market fit. That could collapse prices over the near term which could in turn decrease revenue. As a reminder the stock value isn't a bet on whether "the gold trader" will sell more gold or not, it's a bet on whether the net future returns of the gold trader will occur in line with expectations, expectations that are sky high and have zero competition built in.