09What it costs, and what using it changes

What training and using a model cost, and why using it does not teach it anything.

Training versus using

Training happens before the model is released. Using it afterwards does not change it: your conversations do not alter its parameters. This answers a question every executive asks: "does it learn from our data?" Not by being used. As the second article explained (chapter 4), what a model seems to "remember" is the context the application sends it, not a change in the model itself. Whether a vendor may later include your data in a future training run is a contractual matter, not a technical one, and worth checking in the terms you sign.

What it costs

The leading vendors do not publish training costs, but independent estimates give the orders of magnitude. Epoch AI, which tracks the field, estimates that the final training run of a frontier model now costs hundreds of millions of dollars, and projects that the largest runs will exceed a billion dollars by 2027[15]. Such a run takes months on thousands to tens of thousands of specialised chips.

Published figures need careful reading. DeepSeek reported about 5.6 million dollars for its V3 model, roughly two months on some 2,000 chips[14], but that figure covers only the final run, not the experiments, failed attempts and fine-tuning that preceded it[15]. It shows both how much efficiency matters and how easily a headline figure can mislead.