eric sheridan

Alphabet, Tesla, and other hyperscalers are raising capex aggressively (Alphabet to $75B, Tesla to $25B) even as free cash flow turns negative, because compute supply constraints make capacity a prerequisite for revenue; Goldman Sachs argues the industry is shifting from token maxxing to token optimizing, where efficiency and deflation drive adoption, and full-stack players like Alphabet with TPUs gain advantage.

The AI economy is shifting from maximizing token output to optimizing token efficiency and cost; every compute shift historically brings unit growth through deflation, and Alphabet's TPU stack plus third-party compute deals position it to benefit as enterprises demand wider arrays of models at lower cost.