Google Cloud has launched new AI chips. This move shows growing competition with Nvidia, a major chip maker. Google announced the new generation of its custom AI chips. This is a step to compete with Nvidia.
The company revealed the eighth generation of Tensor Processing Units (TPUs). These are special chips for AI. There are two new chips. The first is the TPU 8t. It is for training AI models. The second is the TPU 8i. It is for inference. Inference means running AI models. It also means handling user requests.
Google says the new chips are much faster. They are faster than older versions. Training AI models is up to three times faster. Efficiency is up to 80% better. This means more performance for the cost. Over a million chips can work together. This offers more processing power. It also uses less energy.
Google does not plan to replace Nvidia chips completely. It will use both technologies. Google Cloud will still offer systems using Nvidia chips. New Nvidia chips are expected later this year.
Other cloud companies are doing the same. Microsoft and Amazon are also making custom chips. These chips meet the growing demand for AI applications.
Experts think this could reduce reliance on Nvidia. This is especially true as companies move to the cloud. But Nvidia is still strong. It makes a lot of money in the chip market.
Analyst Patrick Moorhead predicted this threat in 2016. He said Google chips could challenge Nvidia. However, Nvidia has grown strong. It is now one of the world's most valuable companies.
Interestingly, Google and Nvidia are working together. They are developing network technology. This will improve performance for Nvidia chips in Google Cloud. They are also working on the open-source project "Falcon". Google launched this project in 2023 with the Open Compute Project.
This partnership shows the nature of AI competition. Companies cooperate and compete. They race to lead in advanced computing.