Google parent Alphabet has introduced a new product called Trillium, which is part of its artificial intelligence data center chip family. According to Alphabet CEO Sundar Pichai, the demand for machine learning computers has increased by a factor of 1 million in the last six years. Google has been working on AI chips for over a decade and believes it is well-suited for the current market demands.
Alphabet’s custom chips for AI data centers offer a viable alternative to Nvidia’s processors, which currently dominate the market. The company’s tensor processing units (TPUs) have allowed Google to capture a significant share of the market, with Nvidia controlling about 80% and Google’s TPUs making up the remaining 20%. Google does not sell the chips directly but offers access through its cloud computing platform.
The Trillium chip, Google’s sixth-generation offering, boasts 4.7 times better computing performance compared to the previous version, TPU v5e. It is also 67% more energy-efficient than its predecessor. The new chip is designed to power technologies that generate text and other media from large models and will be available to cloud customers in late 2024.
Google’s engineers have made significant performance gains by increasing the high-bandwidth memory capacity and overall bandwidth of the Trillium chip. AI models require large amounts of advanced memory, which has been a bottleneck in improving performance. The chips are designed to be deployed in pods of 256 chips that can be scaled to hundreds of pods for larger applications.
Overall, Google’s Trillium chip represents a significant step forward in the company’s efforts to compete in the AI data center chip market. With increased computing performance and energy efficiency, the Trillium chip offers a compelling alternative to competitors like Nvidia. Google’s focus on AI and machine learning technologies positions it well to capitalize on the growing demand for advanced computing solutions.
