[wvaske]

There are AI models that are too big to effectively run on smaller hardware. The performance benefit of a monolithic system like this vs distributed GPU based systems can be orders of magnitude.