ChatGPT was a research project that went megaviral, it wasn't intended to be as big as it was.

Training a massive LLM on the scale of GPT-4 required a lot of lead time (less so nowadays due to various optimizations), so the timeframe makes sense.