Are they behind in models, or behind in VC money to burn on subsidized compute offered to the public and early customers?

Genuine question.