By skyforbes 
Semi-analysis ( leading research lab in semiconductors and Ai) have just published a report on the hot topic.
$GOOGL TpU vs $Nvida GpU. They have done the cost comparison and reporting and some crazy things came out of it!
- Google software revamp to make it easier for companies to onboard chips.
-
This becomes a huge threat to Nvida as Cuda system is no longer the only monopoly for Ai chips.
-
Pricing for TpU is extremely cheap! At the 9k standard rack size, Google TpU are 2x cheaper then nvida Gpu.
-
what does this mean for GPU users? Google tpu will crush with margins. They are so cheap that Google cloud is becoming the go to cloud provider for companies as the pricing is just unbeatable.
-
nvida will most likely see margin compression to 60% in the next year or 2 as they need to start to under price to compete with TpU if Google floods the market.
-
OpenAi is in huge trouble. They have billions in commitments and billions in deployed GPU. But the price difference means they will see bad/negative margins for a while. Almost 1/2 trillion in losses for the next 5 years. Compute is expensive & seems like Google has cracked the code with TpU.
-
NeoClouds have been approached by Google and some have started to deploy the TpU at scale for cloud hosting. Initial trials to see how market response is. Other large customer amd contracts that have been rumored for $GOOGL is $META & Xai. Both who are huge nvida customers!
This news will be digested over the weekend but my positions are set on this. I am in a pair trade with long $GOOGL shares + $AVGO calls and short on $NVDA/$AMD.
It does seem like googles full stack approach is working well and we are seeing gemini take on other world models, Google cloud ☁️ dominance over Aws/AZURE and then ToU taking market share from Nvida. Firing in on all fronts.