That was all based on the assumption that scaling LLMs would lead to AGI. That didn't happen and won't, it can be proven for anyone capable of digging into the details of what models actually are and what they are doing (I recommend Chris Hay's videos).

It's becoming more obvious everyday to people. They'll realize that 1000x the cost for marginal improvements isn't worth it and the market will open up. It'll become more based on tooling and smaller more task-focused models instead of this crazy project to create a data center god to rule over humanity.