Alphabet Inc.(NASDAQ:GOOGL), the parent company of Google, announced the launch of the Trillium chip, the latest addition to its artificial intelligence data center chip family. This new chip reportedly delivers performance that is nearly five times faster than its previous version. Alphabet CEO Sundar Pichai highlighted the exponential growth in industry demand for machine learning computing, which has increased a million-fold over the past six years.
Google has been at the forefront of AI chip technology for over a decade, developing custom chips that stand as one of the few strong alternatives to the leading processors from Nvidia (NASDAQ:NVDA), which currently holds about 80% of the AI data center chip market. Google’s tensor processing units, along with their integrated software, have enabled the company to capture a substantial portion of the remaining market share.
The Trillium chip, representing the sixth generation of Google’s TPUs, offers 4.7 times the computing performance of its predecessor, the TPU v5e. This chip is particularly optimized for powering technologies that generate text and other media from large models, and it boasts a 67% increase in energy efficiency over the v5e.
Set to be available to Google Cloud customers in late 2024, the Trillium chip also features enhancements in high-bandwidth memory capacity and overall bandwidth, addressing the memory demands that have historically limited AI performance. Google’s engineers have designed these chips to be deployed in scalable pods, each containing 256 chips, allowing for expansion to hundreds of pods as needed.
Featured Image: Megapixl ©Vladsseven
