Google Unveils New AI Chips and Gemini 2.5 at Cloud Next 2025
- Chaerin Lee
- Apr 11
- 1 min read
April 11 2025
By Chaerin Lee
On April 9 (local time), Google hosted its annual tech conference, Google Cloud Next 2025, at the Mandalay Bay Convention Center in Las Vegas. Under key themes like AI-optimized platforms, open multi-cloud, and interoperability, the company introduced a range of new technologies—including advanced AI models, cloud infrastructure, and next-gen AI chips.
Among the highlights was the 7th-generation Tensor Processing Unit (TPU), named Ironwood. Described as Google’s most powerful chip to date, Ironwood delivers over 10 times the performance of its predecessor. Google emphasized not only its improved computing power, but also its efficiency in energy use and support for high-bandwidth memory (HBM). With this chip, AI training is expected to become faster and more energy-efficient.
Google also introduced a new version of its generative AI system—Gemini 2.5 Flash. The model is now available through Vertex AI and is designed for real-time summarization and document exploration. As a reasoning-based model, Gemini 2.5 Flash can adjust its inference level depending on the complexity of the prompt, offering quicker responses to simpler queries. Businesses can also tailor performance to align with their budget constraints.
Meanwhile, Samsung Electronics announced an expanded partnership with Google Cloud during the event. Samsung’s upcoming home AI robot, Ballie, set to launch in the first half of this year, will be powered by Google’s generative AI. By combining Gemini’s multimodal reasoning with Samsung’s own language model, the robot will be able to provide personalized, intelligent advice to users.
