Google Launches Two New AI Chips to Challenge Nvidia

Google has stepped up its competition with Nvidia by launching two new AI chips. These chips are part of Google’s ongoing strategy to enhance its capabilities in artificial intelligence.

Google’s Latest AI Chips

Amin Vahdat, senior vice president and chief technologist for AI and infrastructure at Google, announced the launch of the eighth generation of Tensor Processing Units (TPUs). This announcement came in a blog post on Wednesday. The new chips consist of two specialized architectures: the TPU 8t for AI training and the TPU 8i for AI inference.

TPU 8t: Designed for Training

• The TPU 8t targets compute-intensive training workloads.
• It significantly reduces the development time for advanced AI models, cutting it from months to weeks.
• Key benefits include:
• Three times the compute performance compared to its predecessor.
• Ten times faster storage access.
• Double the data transfer rate.

TPU 8i: Enhanced for Inference

• The TPU 8i focuses on AI inference, which occurs after model training.
• It improves performance by combining high-bandwidth memory with a tripling of static random-access memory.
• This design minimizes latency in data transfer between chips, enhancing collaboration among AI agents.
• The addition of Google’s custom Arm-based Axiom CPU further boosts efficiency.

Performance Improvements

Google claims the TPU 8i offers an 80% improvement in performance per dollar compared to earlier models. This translates to delivering twice the processing volume at unchanged costs.

Impact on Nvidia’s Dominance

Despite these advancements, Nvidia remains a heavyweight in the data center GPU market, holding an approximate 92% share, according to IoT Analytics. Google has historically relied on Nvidia GPUs, making it one of their largest customers. The introduction of TPUs is part of Google’s strategy to reduce dependence on Nvidia while offering cost-effective AI processing solutions for cloud customers.

Conclusion

The launch of TPU 8t and TPU 8i represents Google’s commitment to enhancing its presence in AI technology. While these chips provide unique advantages, Nvidia still retains leadership in computational power for AI workloads. Both companies are expected to play vital roles in shaping the future of artificial intelligence.