Of course, that does not mean there will not be significant advances in the future.
A substantial portion of this hardware is also being used for rapid model development rather than simply operating existing models and building a stable revenue stream. These companies are focused primarily on training and improving the models, not merely running them, which is significantly cheaper.
A substantial portion of this hardware is also being used for rapid model development rather than simply operating existing models and building a stable revenue stream. These companies are focused primarily on training and improving the models, not merely running them, which is significantly cheaper.