Databricks ups enterprise push

Databricks and Accenture launched the new Accenture Databricks Business Group on March 17, 2026 announced) to accelerate enterprise adoption of Lakebase, Genie, Agent Bricks and the Lakehouse across industries. The company published Qwen3‑Embedding‑0.6B into public preview on March 17, 2026 launched), a compact 0.6B embedding model that Databricks says supports cross‑lingual retrieval in 100+ languages. Qwen3‑Embedding‑0.6B includes “matryoshka” embeddings that can be truncated from 1,024 down to 32 dimensions for cost/performance tradeoffs and supports a 32k‑token context window, with an instruction‑aware design that Databricks reports typically boosts retrieval by 1–5% details). Databricks made the model available via Foundation Model Serving and positioned it to work directly with Agent Bricks and Vector Search so teams can index and retrieve governed enterprise data without exporting it off‑platform integration). Accenture and Databricks cited existing joint customers — Albertsons, BASF and Kyowa Kirin International — as early users of agent‑ready databases and agentic applications, and said the new business group will be backed by more than 25,000 Databricks‑trained professionals to support deployments at scale customers & scale). Databricks has supported production inference with serverless GPU Model Serving since its GPU/LLM optimizations public preview in 2023, claiming up to 3–5x latency and cost reductions for optimized LLM serving, and its cloud docs list GPU‑enabled compute on AWS and Azure for training and serving workloads GPU & serving). The company is also expanding real estate and hiring: Databricks signed a roughly 150,000‑square‑foot lease at One Sansome in San Francisco (announced March 6, 2025) and told the city it plans to invest over $1 billion in local operations and more than double its San Francisco headcount over the next two years office & investment).