# Elon Musk takes Grok into Databricks as xAI chases enterprise distribution > Source: > Published: 2026-06-20 11:51:01+00:00 [Elon Musk (@elonmusk)](https://x.com/elonmusk?ref=runtimewire)'s [xAI](https://x.ai/?ref=runtimewire) made Grok models available inside [Databricks Agent Bricks](https://www.databricks.com/blog/agent-bricks-dais-2026?ref=runtimewire) on June 18, extending Musk's AI company into one of the default workbenches for enterprise data teams. The move, surfaced in an [Aligned News](https://di.gg/ai/nnroxkzf?ref=runtimewire) aggregation after posts from Musk and xAI, is not a new model launch or a funding event. It is a distribution deal. Grok becomes a native model choice in Agent Bricks, Databricks' developer platform for building agents against company data, alongside other frontier and open-source models. For Musk, who founded xAI in 2023 after years spent building companies around infrastructure bottlenecks - payments at PayPal, launch capacity at SpaceX, electric-vehicle manufacturing at Tesla and social distribution at X - the Databricks integration fits a familiar pattern. Grok already had consumer distribution through X and direct developer access through xAI's own API. Agent Bricks gives xAI access to a different buyer: enterprises that want model choice without moving proprietary data into a separate chatbot product. Musk kept his public framing short: "Grok now on Databricks," he wrote. [Elon Musk on X](https://x.com/elonmusk/status/2067646181379854418?ref=runtimewire) [xAI (@xai)](https://x.ai/news/grok-databricks?ref=runtimewire) said in its announcement that Grok models are available on Agent Bricks, coinciding with Databricks' 2026 Data + AI Summit. [Ali Ghodsi (@alighodsi)](https://x.com/alighodsi?ref=runtimewire), Databricks' co-founder and CEO, responded with his own concise endorsement: "Excited to have Grok available on our platform!" [Ali Ghodsi on X](https://x.com/alighodsi/status/2067697057222332702?ref=runtimewire) Ghodsi's role matters here because Agent Bricks is a continuation of the company he helped build out of UC Berkeley's data systems work. His Databricks speaker bio notes he became CEO in January 2016 and holds a PhD in distributed computing from KTH/Royal Institute of Technology in Sweden. ### The product change is narrow, but the channel is important Agent Bricks is Databricks' attempt to make agent development look less like scattered framework work and more like governed enterprise software. In its June 16 summit post, [Databricks](https://www.databricks.com/blog/agent-bricks-dais-2026?ref=runtimewire) said Agent Bricks lets developers build agents with different models and harnesses, connect those agents to data, and control security, monitoring and cost inside the Databricks environment. Databricks says more than 100,000 agents have been built on Agent Bricks since the product was introduced last year. The company also says Agent Bricks supports proprietary and open-source models, naming OpenAI, Anthropic, Gemini, Qwen and Kimi in the same section where it announced Grok support. Those are Databricks' own platform metrics and partner claims, not independently audited usage numbers. The practical pitch is simple: if enterprise data already sits in Databricks, the agent should come to that data instead of the other way around. Databricks frames Agent Bricks around three problems - choice, context and control. Model choice lets a team test different providers for quality, latency and cost. Context means connecting agents to structured and unstructured business data. Control means governing models, tools, agents, access rights, traces and budgets through Databricks' platform rather than through a patchwork of provider-specific dashboards. That makes the integration more valuable to xAI than a logo swap. Grok gets placed where enterprise AI decisions are already being made: inside a data platform used by large companies to manage analytics, AI applications and governance. Databricks said in a December 2025 financing announcement that more than 20,000 organizations use its platform, including more than 60% of the Fortune 500, and that it had surpassed $4.8 billion in revenue run rate. Those are Databricks-supplied figures, but they explain why model providers want shelf space inside the platform. ### xAI is moving Grok beyond X This is the second enterprise distribution move for Grok in the same week. RuntimeWire [reported on June 18](/article/xai-grok-4-3-amazon-bedrock-aws-developers) that xAI put Grok 4.3 inside Amazon Bedrock, giving AWS developers another route to the model. The Databricks integration points at the same strategy from a different angle: meet developers and enterprise AI teams inside the software platforms they already use, rather than relying only on Grok.com, X or xAI's direct API. That shift matters because xAI's early public identity was tied tightly to X and Musk's stated goal of building AI to "understand the universe." The company's [about page](https://x.ai/company?ref=runtimewire) still uses that mission language and describes xAI as building frontier reasoning, real-time voice and generative media systems. Its [homepage](https://x.ai/?ref=runtimewire) presents a unified API for text, code, voice, images and video, and says the platform handles more than 1 million API calls per day with under 200 milliseconds median latency. Those figures are company-published metrics. xAI has also been selling the scale of its compute base. In January, xAI said it raised a [20 billion dollar Series E](https://x.ai/news/series-e?ref=runtimewire) from investors including Valor Equity Partners, StepStone Group, Fidelity Management & Research Company, Qatar Investment Authority, MGX and Baron Capital Group, with strategic investors NVIDIA and Cisco Investments. The same post said xAI ended 2025 with more than 1 million H100 GPU equivalents across Colossus I and II. xAI did not disclose a valuation in that announcement. The Databricks deal is what that infrastructure has to become if xAI wants to compete for enterprise workloads: not just a consumer personality or a benchmarked model, but an option inside procurement-approved software. ### Databricks gets another model without ceding the control plane For Ghodsi, Grok is useful because it reinforces Databricks' argument that enterprises should not have to choose one model provider before they have tested the task. Agent Bricks is designed to keep the control plane with Databricks: data access through Unity Catalog, agent tools and traces inside the Lakehouse, and cost controls through Unity AI Gateway. That is the deeper competitive move. Snowflake, cloud providers and model labs are all trying to own the place where enterprise agents are created, evaluated and governed. Databricks is not trying to be only the database underneath those agents. It is trying to be the platform that decides which model gets called, which data the agent can see, how behavior is monitored and when costs are capped. Grok helps that story because it widens the menu. It does not, by itself, prove enterprise demand for Grok inside Databricks. Neither xAI nor Databricks disclosed adoption numbers for Grok in Agent Bricks, named customers using the integration, or pricing for Grok through this route. The public pages also leave a naming wrinkle. xAI's materials use "SpaceXAI" phrasing around the Grok models. Neither page lays out any new corporate structure or commercial terms. The verifiable product change is narrower and more concrete: Grok models are now a native option in Databricks Agent Bricks. That is enough to make the announcement consequential. Enterprise AI is moving away from standalone demos and toward governed agents embedded where company data already lives. Musk gets another path into that market. Ghodsi gets another model on the shelf without letting the model provider own the enterprise workflow.