Databricks DAIS Meeting Minutes: Software Ecosystem Restructuring in the Age of AI Agents

Overview

This report is based on meeting minutes from Goldman Sachs analysts attending the Databricks Data + AI Summit (DAIS) in San Francisco, deeply exploring the evolution logic of the software ecosystem in the AI agent era. The core view of the report is that value in the software industry is gradually concentrating towards vendors that can support open standards, achieve seamless data ingestion, and provide a unified governance operating system layer. Databricks redefines the competitive barriers of software companies by promoting open ecosystems and accelerating innovation, dividing companies into those relying on innovation and product strength ('benign stickiness') versus those relying on high migration costs ('malicious stickiness'). The report points out that custom agent applications will become a key driver for future Total Addressable Market (TAM) for software, benefiting infrastructure and platform companies like Databricks, Snowflake, and Palantir.

Core views

The focus of software architecture is shifting. Key architectural concepts proposed by Databricks include: the importance of open standards to ensure customers maintain good data hygiene and possess a single source of truth; and an emerging market for custom agent applications, these applications are located in the gaps between traditional SaaS systems, supported by a unified operating system/ontology layer with governance and model control capabilities. The report believes that value in the software stack will increasingly accumulate in the hands of vendors capable of enabling these two architectures, primarily Databricks, Snowflake, and Palantir currently, potentially including Microsoft, ServiceNow, and application vendors finding headless monetization methods in the mid-term. Divide between 'benign stickiness' and 'malicious stickiness'. The report proposes a simple touchstone to assess the durability of software moats: 'benign stickiness' companies innovate at an accelerated pace in core differentiation, products are loved by customers; while 'malicious stickiness' companies innovate less in recent years, maintaining stickiness only because migration costs are high, usually accompanied by higher customer complaint rates or lower usage engagement. As AI makes migration easier, 'malicious stickiness' moats will be eroded. By lowering switching thresholds (e.g., OpenSharing protocol), Databricks actually raises industry standards, forcing competitors to increase R&D and M&A efforts to maintain competitiveness. Among Goldman Sachs covered stocks, Shopify and Cloudflare are viewed as typical 'benign stickiness' companies due to their high innovation speed and disruptive pricing strategies aimed at gaining long-term market share. Blurring boundaries between custom applications and infrastructure. Databricks' Ontology product and Palantir's Frontline Deployment Engineering (FDE) team model make building custom applications tailored to specific business use cases more feasible. These custom applications occupy a larger share of proxy software TAM than before. Although the cost of building applications has decreased due to coding tools, maintenance costs remain a pain point (ServiceNow previously quantified as 5-10x TCO of SaaS). Databricks directly reduces this maintenance Total Cost of Ownership (TCO) through technologies like Unity Catalog and LTAP, eliminating the need for fragile ETL processes. In the competitive relationship with Palantir, the two show more complementarity, but customers must weigh the relationship between outsourcing ontology building work and long-term locking risks. Breakthroughs in new product lines and database technology. Databricks announced SIEM (Security Information and Event Management) and CDP (Customer Data Platform) two 'Chapter 4' applications, commercializing practices already existing among their sophisticated clients. Although milestones in the CDP field may come sooner (because switching costs in marketing are lower than in security), Databricks still needs time to build domain expertise. More fundamental technical breakthroughs lie in LTAP (Lakehouse Transaction/Analytics Processing), which solves a 40-year problem in database engineering, unifying OLTP and OLAP in an open format. The new computing engine Reyden supports real-time lakehouse queries, achieving sub-100ms latency in standard benchmarks, with performance 16 times higher than existing dedicated real-time service stacks. These innovations have driven Databricks core revenue (excluding token passthrough) to accelerate growth over the past five quarters.

Analysis framework

The report adopts a 'Architecture Evolution + Competitive Landscape + Financial Validation' analytical framework. First, deconstruct the core products released by Databricks DAIS from a technical architecture perspective (e.g., LTAP, Ontology, Reyden), analyzing how they solve industry pain points (such as data silos, high maintenance costs, poor real-time performance). Second, introduce a 'Benign/Malicious Stickiness' qualitative framework, combine with the impact of AI on migration costs, re-evaluate the quality of software company moats, and map this logic to public market targets (such as Shopify, Cloudflare). Finally, verify the actual driving effect of technological changes on fundamentals by comparing the acceleration trends of Databricks and Snowflake revenue growth, as well as feedback from conversations with customers regarding custom application adoption rates. This logical chain from micro-product features to macro-competitive barriers, then to financial performance validation, constitutes the core analytical approach of the report.

Methodology notes

• Competition and Strategy FrameworkMoat / competitive advantage

  Benign and Malicious Stickiness

  The report divides software company moats into two categories: 'benign stickiness' relying on continuous innovation and customer love, and 'malicious stickiness' maintained solely by high migration costs and low substitutability. In the context of AI reducing migration costs, only 'benign stickiness' is sustainable.

• Industry/Industrial Analysis FrameworkUpstream/Downstream Transmission in Industry Chain

  Value of software stack concentrates upstream towards infrastructure

  With the rise of custom agent applications, software value no longer stays solely on surface SaaS applications but transmits upstream to infrastructure vendors (such as Databricks, Snowflake) that can provide unified data governance, ontology layers, and open standards.

• Industry/Industrial Analysis FrameworkSubstitution Effect Analysis

  AI reduction of migration costs erosion of traditional moats

  AI tools make data migration and system switching easier and lower cost, thereby weakening the 'malicious stickiness' moats built by traditional software companies relying on high switching costs, accelerating market competition and industry restructuring.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Databricks (Unlisted)

  Core beneficiary, reshaping software architecture through open standards and unified OS layer, accelerating revenue growth.

  Strengths

  Fast innovation speed, LTAP and Reyden tech breakthroughs solve industry pain points, open ecosystem attracts wide developers.

  Weaknesses

  Need to build domain-specific expertise (e.g., security, marketing), faces competition from Palantir and others in specific scenarios.

  Comparison

  Compared to Snowflake, Databricks is more forward-looking in AI-native architecture and real-time processing capabilities; compared to Palantir, provides a more open self-service platform rather than fully managed services.

  Risks

  Execution risk, implementation speed of domain-specific applications slower than expected.

• Snowflake (SNOW.US)

  Ecosystem beneficiary, as a complementary platform to Databricks, jointly driving custom application and data consumption growth.

  Strengths

  Strong data cloud foundation, forms duopoly with Databricks, benefits from overall data TAM expansion.

  Weaknesses

  Innovation speed in AI-native architecture and real-time processing slightly lags behind Databricks.

  Comparison

  Together lead data infrastructure market with Databricks, but Databricks leads in AI agent workflows.

  Risks

  Competitive intensification leading to market share loss, pricing pressure.

• Palantir (PLTR.US)

  Complementary beneficiary, leveraging Databricks and Snowflake as building blocks, shouldering engineering burden via FDE teams.

  Strengths

  Strong frontline deployment engineering capabilities, providing end-to-end ontology construction services for customers, extremely sticky.

  Weaknesses

  Customers may worry about long-term locking risks, pricing model weights early stage value highly.

  Comparison

  Complementary to Databricks, Databricks provides platform, Palantir provides services and solutions.

  Risks

  Customers switch to building in-house or choosing more open Databricks platform upon renewal.

• Microsoft (MSFT.US)

  Mid-term potential beneficiary, if successfully achieving headless monetization and leveraging its front-end ecosystem.

  Strengths

  Huge user base and front-end application advantages, Azure cloud infrastructure support.

  Weaknesses

  Has not shown leadership like Databricks in unified data governance and ontology layer yet.

  Comparison

  Possesses widest enterprise touchpoints, but needs to prove platform integration capability in AI agent era.

  Risks

  Internal product line conflicts, innovation speed slower than pure-play vendors.

• Shopify (SHOP.US)

  Representative of benign stickiness, winning long-term market share with high innovation speed and disruptive pricing strategy.

  Strengths

  Product innovation fast, pricing competitive, high customer satisfaction.

  Weaknesses

  Mainly focused on e-commerce, less generalizable than infrastructure vendors.

  Comparison

  As a paradigm of 'benign stickiness' at application level, similar to Cloudflare, distinguished from traditional SaaS.

  Risks

  E-commerce industry cycle fluctuations, intensifying competition.

• Cloudflare (NET.US)

  Representative of benign stickiness, utilizing low marginal cost advantage for aggressive pricing, expanding security business.

  Strengths

  Global network edge advantages, SASE and security product growth rapidly, strong cost control capabilities.

  Weaknesses

  Security brand awareness still being built, needs time to settle.

  Comparison

  Like Shopify, representative of high innovation, high customer love, moat is solid.

  Risks

  Security market competition intense, squeezing by large cloud vendors.

Key data

• Databricks ARR Growth ExpectationGrowth of 80% in first half of 2027 (approx. 65% excluding token passthrough)Reflects product-market fit and accelerated growth driven by AI
• Reyden Engine Performance ImprovementUp to 16 timesPerformance improvement compared to existing dedicated real-time service stacks
• Genie Ontology EffectivenessAgent accuracy improved by 30%, runtime reduced by 50%Improving AI agent performance in enterprise data through structured graphs
• Custom App Maintenance TCO Multiple5-10 timesQuantified by ServiceNow previously, production and maintenance costs of custom apps are 5-10 times that of SaaS; Databricks new technology aims to reduce this cost

Impact & implications

The report believes that the investment logic in the software industry is undergoing profound changes. Investors should focus closely on vendors that can reduce customer Total Cost of Ownership (TCO) through open standards and unified platforms, and support custom agent application development. Databricks' rise not only threatens traditional data warehouse vendors but also challenges vertical application software vendors by entering the SIEM and CDP fields. For Snowflake and Palantir, as beneficiaries or complements of the Databricks ecosystem, they will also share the dividends of custom application growth. Meanwhile, giants like Microsoft and ServiceNow are also expected to benefit in the mid-term if they can successfully achieve 'headless' monetization. Conversely, 'malicious stickiness' software companies that rely on high switching costs but lack continuous innovation may face valuation re-assessment and risks.

Risks

• Establishing professional experience and sales channels for Databricks in specific areas such as SIEM and CDP may take longer than expected.
• Customers may be unwilling to accept long-term locking risks from vendors like Palantir for outsourcing ontology building work.
• AI reducing migration costs may cause some 'malicious stickiness' software companies to lose revenue faster than expected, triggering industry price wars.
• Changes in macroeconomic environment may affect enterprises' willingness to spend on custom applications and AI infrastructure.

What to watch

• Customer adoption rates and revenue contribution progress for Databricks in SIEM and CDP fields.
• Acceleration of growth in Snowflake's AI-related workload and its competitive stance with Databricks.
• Palantir customer renewal rates and any signs of customers moving to building in-house or Databricks platforms.
• Specific progress and business model innovation by Microsoft and ServiceNow in headless application monetization.
• Continuous performance of 'benign stickiness' companies such as Shopify and Cloudflare in market share and pricing power.