Google Launches Near Real-Time Voice Translation in Gemini 3.5

Google DeepMind introduced Gemma 4 12B, a multimodal AI model designed to run on consumer laptops with 16GB of RAM. The model uses an encoder-free architecture that processes vision and audio inputs directly into the language model backbone, reducing latency and memory overhead. Performance approaches the larger 26B model while maintaining a smaller footprint, and it is released under an Apache 2.0 license.

Lovable and Google Cloud have signed a multiyear deal that will expand Lovable's usage on Google Cloud infrastructure by 5x, according to a source. The agreement also includes expanded access to Anthropic's Claude AI model. The deal signals growing cloud infrastructure demand from AI-focused companies and deeper integration between Google Cloud and third-party AI platforms.

Alphabet announced plans to raise $80 billion through its first stock sale since 2005, with proceeds directed toward AI infrastructure and compute spending. Berkshire Hathaway has committed to purchasing $10 billion in stock at a discount as part of the offering. The move signals Alphabet's major capital commitment to AI development and represents a significant shift in the company's capital allocation strategy after two decades without a public equity raise.

Sundar Pichai discussed Google's structural reorganization in response to competitive pressure from AI startups, major changes to Search that integrate AI agents and direct answers, and YouTube's shift toward AI-powered summarization and indexing. The moves signal Google's pivot from traditional search results to AI-mediated information delivery, with significant implications for web traffic to publishers and content creators.