Google expands AI hardware portfolio with two new chips
Google introduces TPU 8tx for secure workloads built with AMD and TPU 8i Max doubling memory capacity for large AI models.
Stock photo for illustration only, not from the actual event
- Google has unveiled two new specialized tensor processing units, TPU 8tx and TPU 8i Max.
- TPU 8tx was designed in partnership with AMD to handle confidential AI training and execution.
- TPU 8i Max combines two inference chips into one package to double memory and processing power.
- The release aligns with industry efforts by tech giants to reduce reliance on Nvidia hardware.
Google is stepping up its custom silicon development efforts to decrease its reliance on external chipmakers for its cloud computing operations. On Monday, the company introduced two new variants of its tensor processing units (TPUs), customized specifically for artificial intelligence workloads, broadening its hardware capabilities to cover confidential inference and enhanced memory capacity for demanding tasks.
The newly announced units, designated as TPU 8tx and TPU 8i Max, are expansions of the company's eighth-generation chip lineup launched in April, further building upon prior improvements in inference capacity and large-scale training support.
Designed in collaboration with AMD and based on the earlier TPU 8t architecture, the TPU 8tx chip is optimized for securely training and executing AI models. Google explained in a blog post that the semiconductor embeds security directly at the hardware layer, ensuring sensitive data and model weights remain protected and isolated even during active processing.
Stock photo for illustration only, not from the actual event
The chip also features a custom inferencing network enabling simultaneous reasoning and training, which allows organizations to distill and train smaller models on the fly without needing to transfer proprietary data across devices.
"builds security directly into the hardware layer, keeping sensitive data and model weights protected and isolated even while in active use"
Google
Meanwhile, the TPU 8i Max integrates two inference-focused TPU 8i chips into a single package, effectively doubling processing power, bandwidth, and memory capacity. This configuration allows AI models to process a much broader spectrum of information during query operations.
This rollout highlights a broader trend among major hyperscalers and AI developers attempting to build proprietary semiconductor ecosystems, mitigating supply chain bottlenecks and optimizing performance for specialized deep learning frameworks.
Source: TechCrunch
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment