$Alphabet(GOOGL.US)
Context: Google‘s biggest competitive threat to Nvidia may not come from a startup trying to build a cheaper graphics processor. It could come from a company that has spent years developing its own AI chips, already has some of the world’s most demanding AI customers and is now trying to put those chips directly inside customers’ data centers. The Alphabet subsidiary is accelerating its efforts to sell Tensor Processing Units (TPUs) beyond its own cloud infrastructure, potentially turning an internal technology advantage into a much bigger external business. Google had approached customers including Meta about installing its TPUs in their own data centers rather than simply renting computing capacity through Google Cloud. Meta was reportedly discussing spending billions of dollars on Google’s chips for its data centers in 2027, while also considering renting TPUs through Google Cloud. The talks were significant because Meta had historically relied heavily on Nvidia’s processors to train its AI models.
My Trade: Doubled down and added more Google shares. I believe in Google more than Nvidia right now.
Takeaway: Google is also developing software called TPU Command Center to make its chips easier to operate, addressing one of Nvidia’s biggest competitive advantages: its CUDA software ecosystem. The ambition is considerable. Some Google Cloud executives have reportedly discussed capturing as much as 10% of Nvidia’s annual revenue through the TPU business. @Captain's Treasure

