SF1012 LSJ Databricks 1-15-2026.pdf

DEPARTMENT OF THE TREASURY

LIMITED-SOURCES JUSTIFICATION

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

1. Identification of the agency and the contracting activity:

The requiring activity is the Internal Revenue Service Enterprise Services, and the contracting activity is Internal Revenue Service Office of Chief Procurement Officer.

2.a Nature and/or description of the action being approved:

The purpose of this limited-sources justification is to obtain approval to award a Firm-Fixed Price contract with Databricks Federal LLC for the purchase of the brand name items described in section 3.a below. The period of performance will be for a base period of twelve

(12) months with two twelve (12) months option periods. See requisition no. for a copy of the performance work statement.”

2.b Name of the vendor: Databricks Federal LLC 2.c Requisition No.:

2.d Bridge contract: Yes No 2.e Brand name: Yes No 3.a A description of the supplies or services required to meet the agency’s needs:

The IRS is looking to procure a renewal of Databricks licenses and supporting services to continue Enterprise Data Platform (EDP) functionality.

The Internal Revenue Service (IRS) modernization, data, and cloud initiatives, requires a robust and scalable data Lakehouse to host analytical data. The IRS EDP serves as this foundation for advanced analytics, reporting, and artificial intelligence (AI) initiatives across the agency and will be populated through initial loads and incremental updates while maintaining multiple data layers to support diverse analytical and operational needs.

Currently, the EDP uses Databricks offerings to support activities that include, but are not limited to:

Hosting big data engineering pipelines for ingestion and processing of structured and unstructured data.

Enabling analytical workloads for advanced data science, machine learning, and business intelligence.

Supporting modernization of mission-critical IRS systems that drive taxpayer services, enforcement, and compliance.

Meeting new legislative requirements for timely, accurate, and secure data handling.

EDP is particularly critical to the IRS’s operational infrastructure because it:

Hosts and transforms enterprise data that powers the Unified API, which supports several taxpayer- and employee-facing applications.

Enables delivery of customer service, collections, and case management functions that drive the IRS’s day-to-day business operations, including direct taxpayer support and fraud detection services.

Ensures enterprise data is accessible, secure, and reusable for a wide range of business and compliance needs.

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

The proposed acquisition is to acquire the following Databricks products:

Background:

In 2022, the IRS Enterprise Data Platform (EDP) program conducted a thorough Alternatives Analysis (AoA) evaluating large-scale commercial data platforms capable of meeting IRS mission, security, and operational requirements. This analysis assessed multiple vendors and products across IRS requirements including:

Scalability and performance at IRS data volumes Cloud-native architecture suitability Security posture, including FedRAMP High readiness Unified analytics, data engineering, and AI/ML capabilities Governance, lineage, and access control maturity

Based on this comprehensive evaluation, the IRS selected Databricks as the enterprise platform for EDP.

Since that selection, Databricks has been fully operationalized within the IRS and is now embedded in mission-critical workflows. EDP continues to be the system of record for IRS taxpayer data and tax processing, ongoing Filing Season support, fraud detection and processing, implementation of new legislation, and data analytics at the IRS.

The amount of data has materially increased since the original AoA, further reinforcing the conclusions of that analysis; having conducted present market research (details below), that analysis still holds true.

Item no.

Item Description

1 AWS GOV Cloud Commit

Supporting large-scale data ingestion, ETL processing, automated data engineering workflows, data governance, business analytics, and machine learning models

2 AWS US Gov Mission Critical Support

Databricks support level designed for customers with high-stakes workloads requiring the fastest response times and proactive oversight

3 Support US Gov DSE Gold

Acts as the primary point of contact for support management, event coordination, and escalations

4 RSA/Principal Consultant

Technical professional service support

5 Senior Consultant Technical professional service support

6 Learning Subscription Platinum

Provides expert-led training for a variety of class sizes

7 Guided Success Subscription Platinum

Technical professional service support

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

3.b Includes IT: Yes No NAICS 518210 PSC DA10 3.c IGCE/Estimated dollar value:

The IGCE is $ inclusive of a base period of twelve (12) months with two twelve-month option periods.

4.a Identification of the authority being used (Check the appropriate box that applies).

FAR 8.405-6(a)(1)(i)(A) – An urgent and compelling need exists, and following the procedures would result in unacceptable delays.

FAR 8.405-6(a)(1)(i)(B) – Only one source is capable of providing the supplies or services required at the level of quality required because the supplies or services are unique or highly specialized.

FAR 8.405-6(a)(1)(i)(C) – In the interest of economy and efficiency, the new work is a logical follow-on to an original Federal Supply Schedule order or BPA provided that the original order or BPA was placed in accordance with FSS procedures in accordance with the applicable Federal Supply Schedule ordering procedures. The original order or BPA must not have been previously issued under sole-source or limited-sources procedures.

FAR 8.405-6(b) – Items peculiar to one manufacturer. (i.e., Brand Name specifications) 4.b Supporting Rationale:

The IRS EDP is a critical service data and analytics platform that provides a secure, scalable, and unified ecosystem for managing enterprise-level data pipelines, transformations, and analytics in support of IRS mission objectives and evolving legislative mandates.

The justification for Databricks is based on its unique technical capabilities, operational maturity, and demonstrated alignment with IRS’s current Treasury requirements:

a. Requirement: Unified Data and AI Lakehouse Platform

The IRS requires a unified and governed Lakehouse platform for taxpayer data consumption and analytics throughout the IRS; a single, governed environment supporting end-to-end analytics and machine learning. This is essential to the government as it reduces data complexity, eliminates data silos that typically occur in multiplatform/multivendor environments, and removes data redundancy.

This results in data engineering, machine learning, and business intelligence workflows in a single environment, unifying data engineering, supporting ingestion, transformation (via declarative pipelines), and orchestration, and provides a single data source for taxpayer data consumption and analytics.

This Lakehouse architecture provides a single, scalable platform where the agency unifies decades of siloed tax, compliance, and operational data; run advanced analytics and AI directly on that unified data;

and do so with the governance, security, and performance required for highly sensitive federal tax information, which in turn supports faster fraud detection, better compliance risk scoring with maintaining separate legacy data lakes and warehouses.

Databricks is the only platform that provides a unified and governed Lakehouse platform that directly addresses the IRS requirement for a unified and governed Lakehouse platform for taxpayer data consumption and analytics throughout the IRS; single, governed environment supporting end-to-end analytics and machine learning, and does so with Unity Catalog serving as the central governance layer for data, AI assets, and permissions across all workspaces. Unity Catalog gives the IRS fine-grained, consistent access control, lineage, and auditing over highly sensitive taxpayer data and ML models, x

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

while the underlying Delta Lake storage allows data to live once in open, Atomicity, Consistency, Isolation, and Durability (ACID)-compliant tables on cloud object storage, powering SQL analytics, batch and streaming data engineering, and large-scale AI/ML from the same foundation.

Alternative product analysis:

This requirement is only partially met. Snowflake centers on its proprietary storage layer rather than an open, object-storage–native delta architecture, which limits the IRS’s ability to treat its cloud data platform as a fully open, multi-engine Lakehouse spanning raw to curated data in one layer.

This requirement is only partially met. BigQuery tightly couples compute and storage inside Google Cloud and is optimized for SQL analytics, which makes cross-cloud openness, fine-grained control over open table formats, and large-scale Spark-centric AI/ML workflows harder to implement as a unified IRS Lakehouse.

This requirement is not met. Amazon Redshift is fundamentally a data warehouse service that excels at structured, batch analytics but relies on external integrations for data lake patterns, so it cannot by itself provide the IRS with a single, Lakehouse layer for streaming, ML, and open-format governance overall tax data.

b. Requirement: Open Source and Interoperability

The IRS requires an opensource and interoperable platform that natively integrates with other technologies, currently or as part of the future state technology ecosystem of the IRS so that the IRS is not locked into any particular cloud environment or other technology platform. Databricks integrates with all major cloud platforms and has a proven record of contributing major technologies to the open-source community, including Apache Spark, Delta Lake, and MLflow. This aligns with federal directives promoting open standards, vendor neutrality, and technology transparency. Its open architecture ensures interoperability with multiple data storage and compute environments, including AWS, Azure, and Google Cloud, should the IRS change cloud vendors. Databricks also supports data sharing across platforms, clouds, and formats via Delta Sharing.

The IRS needs to remain open source, and interoperable, to integrate legacy on-prem and mainframe systems with modern cloud platforms, in order to avoid being locked into any single vendor, and ensure that tax data can securely move and be analyzed across different technologies and clouds as mission needs change. Opensource formats and engines (such as open table formats and Spark-based processing) let the IRS keep standards-based forms that multiple tools can use over time, instead of binding core records to one proprietary stack. Cloud Interoperability allow workloads like fraud analytics, AI models, and compliance reporting to run where they are most secure, while still maintaining unified governance over highly sensitive taxpayer information.

Databricks is the only data analytics platform built on the opensource format Delta Lake, which provides ACID transactions on top of open-source Parquet files. This prevents vendor lock-in, as data remains in the IRS’s own cloud storage (AWS Simple Service Storage (S3) / Azure Data Lake Storage (DLS) / Google Cloud Storage (GCS), for example) in an open format. As such, it natively integrates with other current technologies so that the IRS is not locked into any particular cloud environment or other technology platform.

Alternative product analysis:

This requirement is only partially met. Snowflake is a proprietary Software as a Service (SaaS) platform with its own storage layer and ecosystem, which reduces the IRS’s ability to rely on fully open formats and engines and limits multicloud interoperability outside Snowflake’s supported environments.

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

This requirement is only partially met. BigQuery is tied to Google Cloud and optimized for its native services, so while it increasingly supports open formats, the IRS would still face compatibility constraints in running the same open-source engines and governance patterns across multiple clouds and environments like AWS and Azure.

This requirement is not met. Amazon Redshift is an AWS-centric (not cloud interoperable) data warehouse that depends on AWS-native services and patterns, meaning the IRS would have to bolt on, rather than natively inherit, consistent open-source tooling and multicloud interoperability for core tax data and analytics workloads.

c. Requirement: Advanced unified analytics, data engineering, and AI/ML capabilities

The IRS requires a data and analytics platform with the ability to provide native support for ML and AI workflows to analyze trends in fraud detection model scoring for fraud case selection and other tax compliance activities as well as augmenting fraud analyst capabilities through context-aware LLMs.

These capabilities are essential to the IRS’s modernization efforts, enabling predictive analytics, fraud detection, detecting tax fraud at ever-increasing rates, and intelligent automation of both internal and taxpayer-facing tax processing tasks. Native support is important for simplification of the EDP ecosystem and minimizes the product set needed for ML and AI data analysis.

The IRS needs unified analytics, data analytics, and AI/ML capabilities so it can bring together all taxpayer, third-party, and operational data into a single analytical ecosystem, instead of fragmented views spread across systems. With unified analytics, the agency can spot patterns that indicate fraud or non-compliance earlier and more accurately. Advanced analytics and AI/ML give the IRS the ability to continuously learn from the full history of tax and operational data, and rely less on potentially out-of-date static reports, so risk models and fraud detectors improve over time. They also enable anomaly detection on streaming data, so the IRS can intervene earlier and reduce false positives.

Databricks provides native integrated support for machine learning and generative AI workflows, including model development, training, deployment, and monitoring. Databricks model serving simplifies the hosting and model inference for high-quality ML/AI applications and agents on IRS data, specifically addressing areas of financial and reputational risk. This includes support for multi-agent systems to deliver higher-quality results with less manual debugging.

Databricks uniquely addresses the IRS’s requirement for native support for unified, robust data analytics, and advanced AI/ML capabilities by consolidating all three on a single, open Lakehouse platform where teams work directly against one governed source of tax data. The platform supports high-performance SQL and dashboarding through AI/ Business Intelligence (BI) Genie, large-scale data engineering (batch and streaming through Spark Declarative Pipeline), and full lifecycle machine learning (through MosaicML and MLOps) on the same Delta Lake storage layer, eliminating the silos and data movement that typically separate BI, data science, and operational analytics. This integrated environment enables the IRS to design, train, and operationalize AI/ML models with model serving and Agent Bricks for fraud detection and compliance risk scoring directly alongside core analytics workloads.

Alternative product analysis:

This requirement is only partially met. Snowflake’s architecture is primarily optimized for SQL-based analytics. Additionally, Snowflake lacks a built-in equivalent to MLflow. Without native experiment tracking, model versioning, and integrated model serving, the agency would be forced to purchase and integrate a third-party MLOps tool. Snowflake is strong for SQL-based BI but requires additional services and integrations for large-scale feature engineering, streaming, and notebook-driven ML, making truly unified analytics and AI/ML workflows more fragmented for the IRS.

Acquisition Title: “EDP Databricks Acquisition” This acquisition is conducted under the authority of the Multiple Award Schedule Program (see FAR 8.401)

Treasury Standard Form – 1012 (Rev 12/24)

Prescribed by Treasury: DTAP 1008.405-6 Tracking No.

This requirement is only partially met. BigQuery excels at serverless SQL analytics but pushes most advanced ML and data science to separate tools, so end-to-end AI/ML pipelines are not natively unified around the same engine and governance layer.

This requirement is not met. Amazon Redshift is a data warehouse for SQL reporting, so advanced feature engineering, real-time streaming, and large-scale ML must be offloaded to other non-native AWS services, preventing the IRS from having a single, integrated platform where analytics and AI/ML all run natively on the same governed lake of tax data.

d. Conclusion As detailed above, the IRS has conducted market research to identify products that could meet the agency’s needs, comparing different vendor products on functional and technical capabilities.

Databricks is the correct platform to fulfill the IRS’s requirements for a secure, scalable, and AI-enabled data platform. Databricks’ alignment to EDP’s technical requirements provides the highest-value, lowest-risk path forward to secure the 2026 Filing Season and beyond. Therefore, it is recommended that the IRS proceed with Databricks under an exception to fair opportunity, in alignment with federal acquisition regulations and the agency’s modernization goals.

5. A determination by the ordering activity Contracting Officer that the order or BPA represents the best value consistent with FAR 8.404(d):

The Contracting Officer has determined that the task order represents the best value consistent with FAR 8.404(d) and lowest overall cost alternative (considering price, special features, administrative costs) to meet the Government’s needs based on market research. This task order will be awarded against DataBrick’s GSA Contract 47QTCA25D00D3.

6. A description of the market research conducted among schedule holders and the results or a statement why market research was not conducted:

Market Research was conducted by researching previous and existing contracts with similar requirements, as well as requesting a Quote from the vendor to verify reasonable pricing for the technical requirements as stated in the Performance Work Statement (PWS). It was also determined no other internal IRS contracts could fulfilled the requirements of this effort.

7.a Any other facts supporting this justification:

N/A 7.b Attachments: Yes No

8. A statement of the actions, if any, the Requiring Activity and Contracting Activity may take to remove or overcome any barriers that led to restricted consideration before any subsequent acquisition for the supplies and services is made:

For any subsequent acquisitions for similar products, the Government will seek competition to the maximum extent possible.

9. Certifications and Approvals: