From Lakehouse to Intelligence Platform: Databricks Declares a New Era at DAIS 2025

The conclusion of the Databricks Data+AI Summit 2025, hot on the heels of the Snowflake Summit, makes it impossible not to contrast the two industry titans. At first glance, the headline announcements from each event are strikingly similar, underscoring a shared trajectory. Both companies are aggressively pushing beyond their respective origins into a unified future of data and AI, advanced data engineering, and generative AI dominance.

However, a deeper inspection of their announcements reveals two fundamentally different routes and philosophies for achieving this vision. While the destinations look the same, the journeys they are charting for their customers are distinct.

Before digging into the technical roadmaps, it’s crucial to understand the financial momentum. The competitive landscape is not just about features, but about market velocity and financial health. While Snowflake is a public market heavyweight, a look at their trajectories tells a compelling story.

• Snowflake: For its current fiscal year (FY26), Snowflake projects approximately $4.3 billion in product revenue, forecasting a healthy 25–26% year-over-year growth. Its non-GAAP product gross margin is consistently strong at around 75%.
• Databricks: As a private company, Databricks is demonstrating explosive growth. It announced it is on track to surpass a $3.7 billion annualized revenue run rate by July 2025, growing at a staggering 50–56% year-over-year. Critically, its subscription gross margins are reported to be above 80%, giving it a significant unit economics advantage over its rival.

At these sustained growth rates, Databricks is not just closing the revenue gap; it is on a clear path to overtake Snowflake.

Market surveys indicate a clear pattern of encroachment. According to analysis based on Enterprise Technology Research (ETR) data, the overlap between the two platforms is significant and growing. In June 2024, approximately 40% of Snowflake customers also utilized Databricks. By June 2025, that number had reportedly climbed to 52%, suggesting that Databricks is successfully landing and expanding within Snowflake’s core customer base.

A primary driver for this growth is Databricks’ established and deepening leadership in AI and Machine Learning. The latest Gartner® Magic Quadrant™ for Data Science and Machine Learning (DSML) Platforms, published in June 2025, places Databricks at the pinnacle of the leaders’ quadrant. It was positioned highest for “Ability to Execute” and furthest for “Completeness of Vision,” outpacing competition from Google, Amazon, and Microsoft.

Market Shift: The Rebundling of Data and AI

Call it unification or rebundling, it is undeniable that large platform vendors are setting the agenda in the data and AI market. In a striking reversal of the past decade’s trend toward specialized tools, major players are rapidly absorbing new functionality, causing their platforms to increasingly resemble one another. As core capabilities like data governance, quality, and observability are integrated, the basis for differentiation is shifting away from simple feature comparisons. Instead, it’s moving toward more intangible, yet critical, criteria: a buyer’s confidence in the vendor’s long-term vision, the perceived strength of its leadership, and its proven ability to execute in the high-stakes arena of enterprise AI.

This consolidation is fueling a surge in M&A activity, driven by a clear customer mandate to reduce vendor sprawl, control costs, and simplify an overwhelmingly complex technology stack. This market-wide movement is playing out across several distinct categories of vendors, all vying to become the definitive platform for data and AI.

1. Cloud Hyperscalers: The Home-Field Advantage

The public cloud giants are leveraging their foundational infrastructure role to offer deeply integrated, all-in-one platforms, making them the default choice for enterprises already in their ecosystem. These include Microsoft Fabric, Google Cloud’s BigQuery, BigLake, and Vertex AI, and AWS SageMaker Lakehouse.

2. Data & AI Stalwarts: The Multi-Cloud Champions

Hot on the heels of the hyperscalers are the data-native leaders who distinguish themselves by being multi-cloud. Snowflake and Databricks have aggressively expanded from their respective strongholds in warehousing and data science to become all-encompassing platforms, now embedding everything from streaming and transactional capabilities to generative AI developer frameworks.

3. Enterprise Software Giants: The Hybrid Titans

Established enterprise vendors are leveraging their deep incumbency and trust within large organizations, particularly by offering robust hybrid and on-premises capabilities that the cloud-native players often lack. Oracle, IBM, Cloudera, and Qlik are augmenting their vast software portfolios with modern data and AI capabilities, positioning themselves as the pragmatic, secure choice for a wide range of industries.

4. Application-First Platforms: Acquiring for Intelligence

A distinct category of vendors is approaching the platform race from the application layer, acquiring technology to embed AI and governance directly into core business workflows. Salesforce (CRM), ServiceNow (ITSM), and SAP (ERP) are actively acquiring companies to build platforms that don’t just store data, but create intelligent data products and power autonomous AI agents.

5. Emerging Platform Contenders

Finally, a set of high-growth, specialized leaders is strategically positioned to expand into broader platforms for data and AI. These include Confluent (data streaming), Palantir (bespoke government and industrial roots), and Celonis (process mining), building a comprehensive platform for AI-driven business automation and intelligence.

This powerful convergence toward unified platforms is putting immense pressure on smaller, independent vendors. After the recent Snowflake and Databricks summits, for example, many point-solution vendors saw their core functionality absorbed and rebranded as mere “features” in the larger platforms. This existential threat is forcing a strategic reckoning, where the primary goal for many innovative startups is shifting from a long-term IPO to a more immediate acquisition by one of the dominant platforms.

Ultimately, no vendor wants to cede control over the critical data assets that underpin their customers’ operations. This intense desire to own the entire data lifecycle drives the relentless pursuit of comprehensive platforms, aiming to minimize customer reliance on external tools and maintain their central position in the data ecosystem.

Databricks Data Intelligence Platform

The Databricks Data Intelligence Platform is a comprehensive, AI-driven platform built on an open lakehouse architecture, designed to unify and simplify data, analytics, and AI workloads across an organization. Many of the announcements from Data + AI Summit (DAIS) 2025 are focused on expanding and enhancing its core functionalities to strengthen this data intelligence paradigm.

Figure 1 shows its components.

Figure 1: Databricks Data Intelligence Platform Key Components

The DAIS 2025 announcements focus on further deepening the intelligence aspects through advanced capabilities, automation, and seamless integration across these core components. Figure 2 classifies the announcements across different areas.

Figure 2: Databricks Data + AI Summit 2025 Key Announcements

We will start by exploring new developments in Databricks’ unified, intelligent, and open platform for all data and AI needs.

Want a video walkthrough? Catch my discussion on the ‘It Depends’ podcast, Episode 91: “Databricks Summit 2025: Data + AI Market Shifts and Key Announcements– June ‘25”. Watch it here: http://www.youtube.com/watch?v=wWqCdIZZTtE

Platform

These announcements are strategically focused on expanding and enhancing core functionalities, with a strong emphasis on improving overall platform usage and adoption across diverse user personas and enterprise needs.

Databricks Free Edition

Databricks is significantly lowering the barrier to entry for its Data Intelligence Platform with a $100 million investment to provide a Free Edition, allowing individuals to explore the full spectrum of data and AI use cases on the platform indefinitely.

This new Free Edition is a game-changer, offering access to nearly the entire suite of Databricks capabilities without requiring a credit card. This stands in stark contrast to the common industry practice of “free tiers” from hyperscalers, where users often discover unexpected charges creeping up, sometimes months after discontinuing use.

While incredibly generous, it’s important to understand the limitations of the Free Edition. It uses the Serverless edition of Databricks and can be deleted if inactive for a prolonged period. It also doesn’t come with SLAs or support. The goal is to learn the platform and not to run production workloads.

Databricks One

One of the most critical elements of any successful platform is its user experience. Recognizing this, Databricks is launching a completely reimagined entry point to its Data Intelligence Platform called Databricks One.

The core objective of Databricks One is to provide business users, those without deep technical expertise, with simple, intuitive, and secure access to the powerful data and AI capabilities of the platform. This new experience significantly broadens the reach of data intelligence beyond traditional technical users (data engineers, data scientists) to empower business teams across the entire enterprise.

Business users can interact with data and AI tools without writing a single line of code. Users can access AI/BI dashboards, ask natural language questions through AI/BI Genie, powered by deep research and quickly find and use relevant dashboards or custom-built Databricks apps (more of this later). It is built on the foundation of Unity Catalog and integrates with identity providers like Entra AD or Okta.

Performance Enhancements

Databricks has been optimizing its SaaS and serverless offerings with several new enhancements. For example, the time to set up Databricks has come down from weeks to under a minute. Some key developments include:

• DBSQL Serverless has achieved a 5x performance gain for interactive analytics and BI workloads and dashboards.
• Predictive Query Execution feature, along with Photon Vectorized Shuffle, adds an additional 25% boost to query performance.
• AI Functions in SQL, which embed generative AI workflows directly into SQL queries, now deliver up to 3x faster performance.
• On-demand, serverless access to NVIDIA A10g GPUs (with H100s coming soon).
• Apache Spark 4.0 improvements for both batch and streaming execution, along with performance optimizations for PySpark workloads.
• Project Lightspeed is a low-latency execution mode for Spark Structured Streaming that delivers p99 latencies under 300 ms for both stateless and stateful queries. This is a direct answer to the demand for real-time data processing with extremely low latency.

SAP Partnership

This collaboration embeds a fully managed version of Databricks within SAP Business Data Cloud (BDC). This allows users to provision Databricks inside the SAP portal in 5 seconds. Key features include:

• Zero-Copy Data Sharing: Data flows seamlessly and securely between SAP BDC and Databricks using Delta Sharing, eliminating the need for complex, costly data replication.
• Unified Governance: Databricks’ Unity Catalog extends its powerful governance, access control, and lineage tracking to your SAP data, ensuring consistency and trust across all your datasets.
• Unlock AI on SAP Data: Businesses can now apply Databricks’ leading AI, machine learning, and data engineering capabilities directly to rich SAP data from sources like S/4HANA. Use cases can include real-time forecasting, AI-powered chatbots on internal docs, and optimized supply chains.

This partnership helps blend SAP data with external sources in a single lakehouse, speeding up data onboarding and model building. Insights can even flow back into SAP BDC for operational use. By simplifying data access and management it reduces overall complexity and TCO.

In short, this partnership empowers businesses to turn their most critical SAP data into a strategic asset for AI-driven insights, bridging the gap between operational systems and cutting-edge analytics.

Transactional Support (Lakebase)

Weeks before DAIS 2025, Databricks made a significant move by announcing its $1 billion acquisition of Neon, a cloud-native, serverless PostgreSQL database. Shortly thereafter, Snowflake announced it is acquiring Crunchy Data, another PostgreSQL database provider. These parallel acquisitions signal a clear intent from both cloud data giants to enter the Online Transactional Processing (OLTP) market while also offering Online Analytical Processing (OLAP) capabilities within a single vendor.

Databricks introduced a new OLTP database architecture category at DAIS 2025, called Lakebase: an operational database built on open standards, with the separation of compute and storage, serverless implementation, that is ideal for modern development workflows, including AI agents and deeply integrated into the lakehouse. Its goal is to enable companies to build applications and AI agents more quickly on a single platform. If “lakehouse” took time to adopt, get ready for “Lakebase” to become the new lexicon.

It’s important to note that the Lakebase is not a replacement for Hybrid Transactional/Analytical Processing (HTAP) databases, which inherently integrates both OLTP and OLAP functionalities within the same database engine. Examples include SingleStore, TiDB, Oracle MySQL HeatWave, and SAP HANA. While both Databricks and Snowflake are now pursuing strategies where HTAP may be an outcome, their approaches and the acquired technologies diverge significantly.

Lakebase (Databricks, built on Neon technology):

Neon is a lightweight, developer-focused, serverless PostgreSQL database designed for modern, highly dynamic applications. Its key differentiator is its ability to spin up instances in under a second, while leveraging copy-on-write branching. Over 80% of its instances reportedly were created by AI agents. This “developer-first” and “AI-native” design, coupled with features like database branching (Git-like versioning for data) and scale-to-zero capabilities, makes it ideal for highly agile development, ephemeral environments, and direct integration with AI agents.

Databricks intends to primarily leverage Neon for AI use cases, particularly for powering the transactional layer of AI-native applications and agents. This strategy culminates in a new service called Databricks Lakebase, built on Neon technology and grounded in the tenets of Lakebase. It promises low-latency, high-concurrency transactional capabilities built on open-source Postgres. The Public Preview of Lakebase is already being used by customers for data serving, feature/model serving, and application state use cases.

Snowflake Postgres (Built on Crunchy Data):

In contrast, Crunchy Data is a more traditional, enterprise-grade PostgreSQL database provider. It specializes in offering robust, secure, and highly available PostgreSQL solutions tailored for typical relational transactional needs and mission-critical workloads. Crunchy Data brings expertise in enterprise-hardened security, compliance, and operational standards, appealing to organizations with established PostgreSQL deployments or stringent regulatory requirements.

Snowflake’s acquisition, aiming to create “Snowflake Postgres,” focuses on providing a secure, compliant, fully managed Postgres experience within its AI Data Cloud, enabling developers to build production-ready AI agents and applications with familiar tools and enterprise rigor.

In essence, while both are extending into transactional capabilities with PostgreSQL, Databricks’ acquisition of Neon is a bolder bet on a future driven by AI agents and highly dynamic, ephemeral transactional workloads, while Snowflake’s acquisition of Crunchy Data caters more to traditional enterprise transactional needs with a focus on robust, production-grade reliability and compliance.

These strategic moves signal the outbreak of a new arms race, as other platform vendors lacking a credible PostgreSQL offering are now actively evaluating the remaining independent transactional database vendors for potential acquisition.

Data Engineering

The data engineering landscape is rapidly evolving, with a clear trend toward simplified, intelligent, and highly automated data pipelines. Snowflake’s announcement of its Apache NiFi-based Openflow capability in June 2025, designed for data extraction and loading, signals its move into an area where Databricks has already established a strong presence with Lakeflow.

Lakeflow represents Databricks’ solution for data ingestion, transformation, and orchestration. It officially reached General Availability (GA) at DAIS 2025, solidifying its position as a mature offering. It was initially launched, in 2024, with three core components:

• Lakeflow Connect: Provides a rich set of pre-built ingestion and Change Data Capture (CDC) connectors for various sources, including databases (DBMS), SaaS applications (like Salesforce, Workday, SAP), and storage locations (like SharePoint). A significant enhancement is the new Zerobus API, which enables direct writes to the lakehouse with ultra-low latency, typically within 5 seconds. Connect is built on Databricks’ acquisition of Arcion.
• Lakeflow Pipelines: This component leverages the Delta Live Tables (DLT) technology, allowing users to define data transformations using either SQL or Python. It automates common operational tasks, including schema evolution, data quality enforcement, and error handling, simplifying the development and management of data pipelines.
• Lakeflow Jobs: Designed for robust orchestration, Lakeflow Jobs automatically manages and monitors the health of data pipelines, ensuring reliable execution of complex workflows.

At DAIS 2025, Databricks introduced two pivotal enhancements that further refine Lakeflow and align with the conference’s theme of declarative, AI-driven operations:

Lakeflow Declarative Pipelines

Built on the open source Spark Declarative Pipelines framework announced at Data + AI Summit, this component lets users define end-to-end pipelines with just a few lines of SQL or Python.

Embodying the “ask the system for an output, not how to execute” philosophy, Declarative Pipelines empower users to simply describe their desired data transformations and outputs. The underlying engine intelligently handles the intricacies of execution, including:

• Automated Data Lineage: The engine understands dependencies and relationships within the data.
• Optimized Execution Order: It determines the most efficient sequence of operations.
• Built-in Data Quality: Incorporates automated checks and enforcement.
• Proactive Monitoring & Retries: Handles failures gracefully and ensures pipeline resilience.

Declarative Pipelines effectively succeed DLT and provide a broader range of data sources beyond just Delta Lake, while still handling both batch and streaming data pipelines in SQL or Python. A key enhancement is their direct integration and availability within Apache Spark 4.0, leveraging the latest performance and feature improvements.

Lakeflow Designer

Lakeflow Designer is a no-code ETL interface that democratizes pipeline construction. It allows users to visually design data pipelines using a drag-and-drop interface, further augmented by natural language input that understands your data through Unity Catalog integration. This intuitive environment automates the entire workflow, from initial build to production deployment.

Behind the scenes, Lakeflow Designer compiles these visual workflows into standard Lakeflow Declarative Pipelines SQL code. This critical capability means that instead of solely relying on data engineers for pipeline creation, business analysts can now build sophisticated data pipelines. This fosters collaboration, allowing business analysts to initiate and iterate on pipelines while data engineers can easily inspect, refine, and maintain the underlying code, thereby reducing redundancy and maintenance overhead.

Crucially, all Lakeflow components are deeply integrated with Unity Catalog. This ensures consistent governance, maintaining unified logic, lineage tracking, granular access controls, and comprehensive observability across all data and pipeline assets within the Data Intelligence Platform.

Finally, it’s worth noting the vibrant and competitive ecosystem of independent vendors in the data engineering space, offering a diverse range of tools such as Informatica, Prophecy, dbt, SQLMesh, starlake.ai, Qlik/Talend, Talaxie, and Apache Hop, among others. Databricks’ Lakeflow aims to provide a comprehensive, integrated solution while still enabling interoperability within this broader ecosystem.

Data Warehouse (Databricks SQL)

Databricks’ data warehousing offering, Databricks SQL, launched in 2021, has seen remarkable adoption, now serving over 12,000 customers. This growth is underpinned by significant performance enhancements, with query execution now 5x faster than at its inception.

At this year’s summit, prepare for another new term: Lakebridge. In a move mirroring Snowflake’s recent announcement of SnowConvert AI (a free data migration service), Databricks unveiled Lakebridge, a free migration tool specifically designed to accelerate the transition from legacy data warehouses to Databricks SQL. However, while Snowflake’s offering leans on an AI agent for code conversion, Databricks’ Lakebridge leverages the sophisticated LLM-powered code conversion technology from its February 2025 acquisition of BladeBridge.

Lakebridge provides a comprehensive, end-to-end migration workflow:

• Source Profiling and Estimation: It begins by analyzing historical query data and statistics from the source system, providing clear cost and effort estimates for the migration.
• Usage Pattern and Dependency Analysis: Lakebridge intelligently analyzes existing usage patterns and identifies critical dependencies within the legacy data warehouse environment.
• Automated Code Conversion: This is a core strength, automating the conversion of complex SQL and ETL code, including stored procedures, into Databricks-compatible SQL or Spark SQL. This powerful conversion is driven by Databricks’ Test-Time Adaptive Optimization (TAO). TAO is a reinforcement learning technique applied to LLMs that improves model performance without requiring labeled data. Behind the scenes, TAO utilizes the Databricks Reward Model (DBRM) to score and rank multiple candidate responses, ensuring the highest quality code translation.
• Data and Code Migration: Leveraging Lakeflow Connect, it seamlessly migrates both the transformed code and the actual data to the Databricks lakehouse.
• Validation and Reconciliation: Post-migration, Lakebridge ensures data integrity by validating and reconciling results, confirming parity in row counts, schemas, and data quality.
• Progress Reporting: Intuitive dashboards and reconciliation reports provide transparent tracking of migration progress and ongoing data integrity.

This powerful automation allows Lakebridge to automate up to 80% of migration tasks, significantly speeding up what is traditionally a complex and time-consuming process. It supports a wide array of source platforms, including major data warehouses like BigQuery, Synapse, Snowflake, Redshift, Teradata, Oracle, Vertica, SQL Server, Db2, and various Informatica products, demonstrating its broad applicability in diverse enterprise environments.

Data Governance (Unity Catalog)

Unity Catalog is the crown jewel of the Databricks Intelligent Platform as V3 whichit is the single pane of glass to discover various assets and manage them. DAIS 2025 witnesses significant new announcements in this space.

Full Apache Iceberg™ support in Databricks

Last year, Databricks acquired Tabular, a company known for its deep contributions to Apache Iceberg and set in motion a better collaboration between the Apache Iceberg and Delta Lake open table formats.

At DAIS 2025, Databricks announced Public Preview for Apache Iceberg™Apache Iceberg support in Databricks, unlocking the full Apache Iceberg™ and Delta Lake ecosystems with Unity Catalog.

This means:

• External Engine Read/Write: External engines can now directly read (GA) from and write (public preview) to performance-optimized, Iceberg managed tables within Unity Catalog via Unity Catalog’s Iceberg REST Catalog API.
• Fine-Grained Governance: This access comes with Unity Catalog’s robust, fine-grained governance capabilities, ensuring consistent security, access control, and lineage tracking for all Iceberg data assets.
• Iceberg Catalog Federation: Now in Public Preview, Unity Catalog also allows you to govern and query Iceberg tables managed in AWS Glue, Hive Metastore, and Snowflake Horizon without copying data.
• Delta Sharing for Iceberg. Delta Sharing, the open protocol for secure data sharing, has expanded its capabilities. Now in Private Preview, allowing you to share Unity Catalog tables and Delta tables with any recipient using Delta Sharing and consume them in any client that supports the Iceberg REST Catalog API.

These developments underscore Databricks’ dedication to an open lakehouse ecosystem, significantly improving interoperability and empowering organizations with greater flexibility and control over their data assets, regardless of the chosen open table format.

In addition to the Iceberg support in Unity Catalog, Databricks is also taking the lead to bring the Iceberg and Delta communities together to make the two open table formats more compatible. The recent release of Apache Iceberg™v3 is the first step in that direction. Key features in Iceberg V3 that enable this include Deletion Vectors (making row-level deletes more efficient), Row Lineage (tracking how individual rows change), and the Variant data type (for handling semi-structured data more flexibly).

Unity Catalog Metrics

Both Snowflake and Databricks are acutely aware of the critical need to make business metrics easily accessible to analytics and AI providers. However, their approaches to achieving this differ significantly. Snowflake recently introduced a powerful metrics capability that defines metrics directly with the schema of the underlying tables. In contrast, Databricks has integrated its new metrics definitions directly into Unity Catalog.

Unity Catalog Metrics stores business metrics as first-class assets. Essential business metrics and KPIs can now be defined and governed directly within Unity Catalog. This semantic definition ensures consistency, trust, and clarity across the organization, making it easier for users to leverage consistent business semantics across all workloads, like dashboards, SQL applications, notebooks, AI models, and data engineering jobs. Upcoming integrations will extend support to BI tools like Tableau, Hex, Sigma, ThoughtSpot, Omni, etc. and observability tools like Anomalo, Monte Carlo and more.

Locating metrics within the database schema versus within a centralized catalog each has its own trade-offs and in this discussion, we won’t declare one approach definitively superior.

Curated Internal Marketplace for Data & AI Discovery: To further empower business users, Databricks is introducing a new, curated internal marketplace. This marketplace serves as a central hub for discovering the highest-value data, AI models, and AI/BI assets, all organized intuitively by business domain. Every asset in this marketplace is augmented with automated data intelligence, helping teams quickly find, trust, and act on the right data for their specific needs.

These enhancements mean that business users can access and leverage critical metrics and AI assets with greater ease, confidence, and consistency, fostering a more data-driven culture across the entire enterprise.

AI

Mosaic AI has emerged as the complete end-to-end AI agent lifecycle development, deployment and governance platform. It prepares data and builds, deploys, and evaluates models and agents with monitoring and guardrails.

Several enhancements in the AI space are included in this section. Data Intelligence Platform users can now access Google’s Gemini models natively to build AI applications securely. These models provide advanced reasoning and large context windows. Other enhancements include MLflow 3, which redesigned this popular OSS developer platform with the latest in monitoring, evaluation, and lifecycle management for Generative AI, Storage-optimized Vector Search, which can now scale up billions of vectors while delivering 7x lower cost, and Serverless GPU Compute for users to train models, run inference, or process large-scale data transformations.

This can be found here: https://www.databricks.com/blog/mosaic-ai-announcements-data-ai-summit-2025

Agent Bricks

Get ready to add another new term to your Databricks vocabulary: Agent Bricks. This is Databricks’ innovative no-code/low-code platform designed to automate the entire lifecycle of building, evaluating, and deploying domain-specific, production-ready AI agents that are deeply grounded in your enterprise data. Unique to the Databricks approach is their focus on making domain-specific evaluation benchmarks to score Agent accuracy, and continuously optimizing the Agent with the latest AI research techniques to ensure optimal performance.

Agent Bricks leverages cutting-edge innovations from Databricks’ Mosaic AI research, allowing users to create and deploy sophisticated AI agents using just natural language. Simply provide a high-level description of the task you want the agent to perform, connect your relevant enterprise data, and Agent Bricks automatically generates the necessary agent logic, benchmarks, synthetic data, evaluation judges, and optimization routines.

Core Capabilities of Agent Bricks:

• Automation: Agent Bricks intelligently selects the most appropriate underlying AI model and automatically generates any required domain-specific synthetic data to train and fine-tune the agent.
• Evaluation: It auto-generates comprehensive benchmarks and utilizes powerful Large Language Models (LLMs) as “judges” to rigorously evaluate and optimize the agent’s output for accuracy and relevance.
• Governance & Ops: Seamlessly integrated with MLflow 3.0 and Unity Catalog, Agent Bricks ensures robust governance, lineage tracking, and operational management of your AI agents, bringing them under the same control as your other data and AI assets.
• Multi-Agent Orchestration: For complex business workflows, Agent Bricks supports multi-agent orchestration. This includes a multi-agent supervisor with Model Context Protocol (MCP) support, enabling intricate scenarios like automated document analysis, compliance checks, and intelligent response generation.
• Cost Control: It intelligently balances performance and compute costs, delivering multiple optimized agent versions for you to choose from based on your specific requirements and budget.

Agent Bricks truly distinguishes itself through its extensive automation features, significantly accelerating the path from concept to production for AI agents. For instance, AstraZeneca successfully used Agent Bricks to deploy a knowledge extraction agent in just 60 minutes, achieving an impressive accuracy range of 60% to 80%. This highlights the platform’s ability to rapidly operationalize AI within the enterprise.

Genie

Databricks AI/BI Genie, is an intelligent, conversational AI assistant within the Databricks Data Intelligence Platform. Its primary goal is to empower business users (and other non-technical users alike) to interact with and derive insights from their data using natural language, rather than needing to write code, learn a BI tool or rely on experts to create insights on their behalf.

Here’s what Databricks Genie does:

• Natural Language Querying: You can ask questions about your data in everyday language.
• Instant Visual Answers: Genie doesn’t just give you raw data; it understands your intent and returns a combination of text summaries, tabular data and visualizations. With every response, Genie also provides an explanation of how it arrived at the answer to give users confidence in that output.
• Deep Research Capabilities: Beyond simple queries, future enhancements to Genie will allow it to perform “Deep Research.” This new capability enables Genie to move beyond basic descriptive answers to address “why” and “how” questions (e.g., “Why are sales down?” or “How can we increase our pipeline?”). Leveraging the latest advances in LLM research, Genie can now tackle complex questions by producing a research plan and analyzing multiple hypotheses in parallel before summarizing the results..
• Contextual Understanding: Powered by Databricks Intelligence, Genie learns the semantics of your organization’s data, understanding business terms and relationships to provide more accurate and relevant responses.
• Access to Metrics and KPIs: Integrated with Unity Catalog Metrics, Genie can directly leverage defined business metrics and KPIs, ensuring consistent answers based on your organization’s single source of truth.
• Integration within Databricks One: AI/BI Genie is a core component of Databricks One, the new simplified user experience designed to democratize access to data and AI capabilities for all business users.

In essence, Databricks Genie aims to break down the technical barrier to data analysis, making it easier for anyone to find, understand, and act on insights from their data using the power of conversational AI.

81% of Databricks customers use Genie.

Apps

Databricks and Snowflake are both converging on a crucial strategic goal: collocating applications as close to the data as possible. Minimizing data movement allows for faster insights, better performance, and enhanced security. However, while their approaches to achieving this are similar in objective, they diverge significantly in their target audience and underlying architecture.

• Snowflake’s Native Apps are primarily aimed at ISVs (Independent Software Vendors), allowing them to build and monetize applications directly within the Snowflake Data Cloud. These apps, like Relational AI, run inside Snowpark Container Services (SPCS), which leverage Kubernetes containers. While powerful, this containerized approach can sometimes impose specific restrictions on how applications are built and deployed.
• Databricks Apps, in contrast, are designed to empower business users (and internal developers) to build and deploy their own internal, domain-specific applications securely inside the Databricks Data Intelligence Platform. Databricks envisions these apps becoming the “Visual Basic of data apps,” providing a low-code, accessible way to create powerful tools.

Databricks Apps are designed to offer unparalleled flexibility for developers:

• Choice of Frameworks: Developers can leverage popular frameworks like Streamlit, Dash, Shiny, and Gradio. Interestingly, Databricks estimates it hosts more Streamlit implementations than even Snowflake, despite Snowflake owning Streamlit.
• Choice of Tools: The platform supports integration with leading development and low-code tools such as Retool, Posit, Superblocks, and Lovable.
• Node.js Support: Databricks Apps now also support Node.js, expanding the range of programming languages developers can use.

At DAIS 2025, Databricks highlighted the power of Databricks Apps by announcing Superblocks as a key launch partner. In a compelling demo, a natural language prompt was used within Superblocks to rapidly create a sophisticated supply chain application. This app features interactive widgets seamlessly integrating with various Databricks services like Databricks SQL, Lakebase, AI/BI, MLflow, and AI/BI Genie. With a single click, the application was deployed as a Databricks App, with the final step being a straightforward configuration of permissions and authorizations, demonstrating the ease of operationalization.

This approach emphasizes that Databricks Apps are not just about hosting code; they’re about providing a highly integrated, governed, and flexible environment for organizations to build tailored, data-driven applications that live right where their data and AI capabilities reside.

Summary

The Databricks Data + AI Summit 2025 unmistakably signals a fascinating acceleration toward a unified future of data and AI. Databricks is aggressively expanding beyond its traditional strengths in data engineering and data science, now offering comprehensive platforms for analytics and generative AI. Yet, this expansion intensifies its rivalry with Snowflake. Our deeper dive into their respective strategies reveals that while the destination appears increasingly converged, the paths they’re charting for customers are fundamentally distinct.

The broader story isn’t about features alone — it’s about platform consolidation, ecosystem control, and strategic vision. In this AI-first era, the stakes have never been higher.

However, we believe this isn’t a zero-sum game. As the cost of intelligence plummets, the demand for data will skyrocket, presenting both platforms with an equal opportunity to thrive and flourish.