Alphabet’s 8th-Gen TPUs Spark New Revenue Stream at Forward P/E of 28
GOOG•Alphabet's custom tensor processing units (TPUs), developed over a decade, now include specialized eighth-generation variants for training and inference, driving lower AI operating costs. The company has begun selling these chips directly through partners, creating a new hardware revenue stream while trading at a forward P/E of 28, suggesting further upside.
1. Decade-Long TPU Development
Alphabet pioneered custom tensor processing units over ten years ago, and its eighth-generation TPUs now feature distinct variants optimized for both model training and inference, resulting in significantly lower AI infrastructure costs. These developments underscore Alphabet’s deep integration of AI hardware and software, giving it a competitive edge in performance and efficiency.
2. Direct Sales and Valuation Implications
For the first time, Alphabet is marketing TPUs directly through partner channels, opening a new revenue line beyond internal use. With the stock trading at a forward P/E of 28, this hardware push may unlock additional valuation as the market revalues Alphabet’s end-to-end AI capabilities.
