Technology Desk: Amidst the increasing demand for Artificial Intelligence (AI), Google has introduced two new AI chip systems. The company launched TPU 8tx and TPU 8i Max on November 9. Their objective is to make the training of AI models, data processing and work of AI agents faster and safer. This move will further strengthen Google's efforts to challenge Nvidia in the AI chip market.
What is Google's new AI chip?
Google's new chip systems are part of its eighth-generation Tensor Processing Unit (TPU) family. TPUs are processors specifically designed for AI-related tasks. These are used to train AI models and prepare answers from them.
In April this year, Google introduced TPU 8t and TPU 8i. Now with the addition of TPU 8tx and TPU 8i Max, the company has a total of four eighth generation TPU options.
TPU 8tx: Data will remain safe even during processing
TPU 8tx is designed for companies and government institutions that need to protect sensitive data when using AI.
Data usually remains encrypted when stored or sent over the Internet, but has to be decrypted during processing. Google claims that the TPU 8tx provides security at the hardware level. With this, important parts of the data and AI models can remain safe even during processing.
The system works with AMD's fifth generation EPYC processors. According to Google, its latency can be reduced by up to 2 times compared to TPU 8t. This will enable companies to improve and train AI models using their personal data.
TPU 8i Max: Bigger AI models will get more power
TPU 8i Max is designed to process large AI models and long data simultaneously. It combines two TPU 8i chips in a single package.
According to Google, it has double the computing capacity, SRAM, and HBM memory bandwidth compared to the TPU 8i. In simple words, this chip can handle more information at a time and will help in completing large AI tasks faster.
It can be used for complex tasks like AI coding assistant, mathematical analysis and drug discovery.
Why will Nvidia be challenged?
Nvidia has a strong hold in the AI chip market. Its GPUs are used by many of the world's biggest tech companies for training and running AI models. Google wants to reduce this dependency through its TPU.
The company is making its chip system available to Google Cloud customers and other developers. According to Google, it has around 100 customers associated with its TPU business. These include AI companies, trading firms, retail companies and universities.
Competition will increase in AI chip market
Google's new TPU systems show that the next race in AI isn't just about building better models. Companies are also investing in hardware that helps AI run faster, keep data secure, and reduce costs.
However, challenging Nvidia will not be easy. Its GPU and CUDA software are already being used on a large scale. In such a situation, Google's success will depend on how fast companies adopt its TPU.