AI Adoption Soars, Google Cloud Relies on Its Own Chips to Deal with GPU Scarcity

Vice President of Technology and Customer Engineering Google Cloud, APAC, Moe Abdula (photo: Dinda Buana/VOI)

JAKARTA - Amid the high global demand for chips and GPUs (Graphics Processing Units) due to the surge in the adoption of artificial intelligence (AI), Google Cloud claims to be able to meet customer needs.

Vice President of Technology and Customer Engineering Google Cloud, APAC, Moe Abdula admitted that the surge in AI computing needs indeed occurred in almost all industries.

However, he stated, Google has so far not experienced supply constraints for infrastructure to support AI services on its cloud platform.

"Every time a new technology emerges, there is indeed an increase in demand. However, so far we have been able to manage this increase in demand in terms of, for example, our AI Accelerator," he said in the Google Cloud Media Briefing on Wednesday, July 15.

Because according to him, Google has a strong supplier network to ensure the availability of GPUs for Google Cloud customers. One example is through its partnership with NVIDIA, which currently dominates the global Al GPU market.

In addition to GPUs, Google is also strengthening its position by developing TPU (Tensor Processing Unit), which is a special chip developed by Google to handle AI and machine learning workloads.

Moe explained that the development of TPU began when Google realized that AI systems needed a different computing architecture. One of the triggers was the need to develop voice-based search features.

"So we did an analysis, for example, what if the search was done using voice. And for that a lot of technology is needed. That's where TPU came in. And we've been using it for quite some time to enable parallel processing," he said.

However, Google does not rely entirely on TPUs. Moe said that the company still uses GPUs in various services and data centers, especially in areas whose infrastructure has not been designed to support TPUs.