The Zhitong Finance App learned that according to media reports quoting people familiar with the matter, Google's parent company Alphabet (GOOGL.US, GOOG.US) is developing a next-generation AI server chip codenamed “Frozen v2,” which aims to further improve the operation efficiency of artificial intelligence models. The news boosted Alphabet's stock price by more than 3.5%.

According to the report, “Frozen v2” will be directly integrated into the Google Gemini large model architecture, thereby improving the overall operating efficiency of the model. Employees participating in the project expect that the chip can process 6 to 10 times the number of tokens per unit of power consumption of Google's latest generation tensor processor (TPU). The relevant efficiency evaluation is mainly based on the token throughput that can be provided per unit of power.

With the rapid development of generative AI, global tech giants are generally facing the problem of tight AI computing power, and Google is also continuing to increase self-developed AI chips to reduce dependence on external computing power resources and improve model performance. According to reports, the new chip is expected to be deployed as soon as 2028.

In April of this year, Google unveiled the eighth-generation TPU products at the Cloud Computing Conference held in Las Vegas, including the TPU 8t for AI model training and the TPU 8i for inference tasks, further improving its AI infrastructure layout.

Disclaimer:Webull uses external vendor Google Translation Service for news translations where we endeavour to ensure these are correct, however, we recommend that you please double-check this information accordingly. Webull is not responsible for translation errors or issues.