GCP updates and free Gemini suite

Google spent the first half of April pushing two tracks at once: new Google Cloud building blocks for companies and new Gemini tools that make solo prototyping cheaper and faster. (cloud.google.com) (blog.google) On the cloud side, Google Cloud’s April 2 roundup and follow-on posts added BigQuery Graph for relationship analysis, reintroduced Looker Studio as Data Studio on April 10, and launched QueryData in preview the same day for natural-language database queries. (cloud.google.com 1) (cloud.google.com 2) (cloud.google.com 3) (cloud.google.com 4) Google also published a unified graph setup on April 14 that pairs Spanner Graph for live operational data with BigQuery Graph for large-scale analysis, aimed at teams that do not want separate graph databases and data pipelines. (cloud.google.com) A graph database maps connections the way a route map shows stations and lines, instead of storing everything as isolated rows. Google’s pitch is that companies can trace links among people, products, accounts, or devices inside BigQuery and Spanner without moving the data into a separate system first. (cloud.google.com 1) (cloud.google.com 2) Natural-language-to-SQL tools try to turn plain English into database commands, which is useful only if the generated query is accurate enough to trust. Google said QueryData launched in preview on April 10 for AlloyDB, Cloud SQL, and Spanner, and described it as delivering “near-100% accuracy” for data agents. (cloud.google.com) The interface changes point the same way. Google said Data Studio now serves as one place for reports, BigQuery conversational agents, and data apps built in Colab notebooks, while BigQuery Studio’s Gemini assistant has been upgraded into a context-aware chat tool that can work across open query tabs. (cloud.google.com 1) (cloud.google.com 2) For media generation, Google Cloud put Lyria 3 into public preview on Vertex AI on April 7, with a 30-second model and a Pro version that can build songs up to three minutes long. Google’s Veo 3.1 Lite post on March 31 said the video model supports text-to-video and image-to-video, 720p and 1080p output, and 4-, 6-, or 8-second clips through the Gemini API and Google AI Studio. (cloud.google.com) (blog.google) The lower-cost developer story comes from Google’s open and semi-open releases. Google unveiled Gemma 4 on April 2 under an Apache 2.0 license in four sizes—Effective 2B, Effective 4B, 26B Mixture of Experts, and 31B Dense—and said the family is designed to run from phones and laptop graphics chips up to cloud accelerators. (blog.google) Google Cloud followed with a post saying Gemma 4 can run through Google Kubernetes Engine, Google Compute Engine, Vertex AI, Agent Development Kit, and Cloud Run, including serverless graphics processing units for the 31B instruction-tuned model. Google also said the 26B Mixture of Experts version would become fully managed and serverless in Model Garden “over the coming days.” (cloud.google.com) The split release strategy is straightforward: proprietary models such as Lyria 3 and Veo 3.1 Lite stay tied to Google APIs, while Gemma 4 gives developers weights they can run and fine-tune themselves. That gives Google a way to sell managed infrastructure to large customers and still court smaller teams that want more control over cost, deployment, and data handling. (cloud.google.com) (blog.google)