Martyn Rhisiart Jones

Segovia, 20th December 2025

Introduction

The data warehouse has undergone a profound transformation over the past decade. Expensive, rigid, on-premises systems were once built for batch reporting. These systems have now evolved into cloud-native, highly scalable platforms. They are designed to meet the demands of today’s data-driven organisations.

Modern data warehousing is no longer just about storing historical data for reporting. It now serves as a central and trusted foundation. It supports business intelligence (BI) and self-service analytics. It also provides real-time insights and supports machine learning. Additionally, it enables AI-powered applications. Three major forces have driven this shift:

• The explosion of data volume and variety (structured, semi-structured, and increasingly unstructured)
• The need for near-real-time decision-making rather than overnight batch jobs
• The rise of cloud computing, which enables elastic, cost-efficient, and fully managed architectures

Today’s leading data warehouses, such as Snowflake, Google BigQuery, Amazon Redshift, and Microsoft Fabric, deliver enterprise-grade performance. They offer strong governance. They integrate seamlessly with modern tools and workflows. All this is achieved while keeping costs predictable.

This guide outlines the core principles that define modern data warehousing. It highlights the key considerations organisations must weigh when designing or selecting a solution. It also raises an important question. Do these contemporary platforms still adhere to the foundational principles Bill Inmon established in the 1990s? Are they subject-oriented, integrated, non-volatile, and time-variant?

The short answer is yes. Modern data warehouses continue to embody Inmon’s vision of a trusted, integrated, historical source of truth for decision-making. However, the way these principles are implemented has evolved significantly to accommodate cloud-native elasticity, real-time data, and broader use cases.

You may be building a new data platform. Alternatively, you might be migrating from legacy systems or evaluating vendors. Whatever the case, you must understand these principles and considerations. This understanding will help you design a solution that delivers fast, reliable insights. The solution will remain future-proof and cost-effective.

Basic Principles and Considerations in Modern Data Warehousing (2025)

Modern data warehousing has shifted from rigid, on-premises systems to cloud-native platforms. These platforms are highly scalable and optimised for analytics, BI, AI/ML, and real-time insights.

Core Principles

1. Cloud-Native Architecture
   • Fully managed, serverless or near-serverless options.
   • Elastic scaling of compute resources independent of storage.
   • Pay only for what you use, with auto-suspend/resume capabilities.
2. Separation of Storage and Compute
   • Storage uses cheap, durable object storage (e.g., S3, GCS, Azure Blob).
   • Compute is virtualised and scales independently per workload.
   • Enables massive data volumes and concurrent queries without cost explosions.
3. Support for Diverse Data Types
   • Structured data (tables) + semi-structured (JSON, Parquet, Avro, etc.).
   • Increasing support for unstructured data via vector embeddings and semantic search.
4. ACID Transactions and Data Integrity
   • Full ACID compliance (Atomicity, Consistency, Isolation, Durability).
   • Reliable updates, deletes, and merges at scale.
5. Real-Time and Streaming Capabilities
   • Near-real-time ingestion and querying.
   • Native connectors for Kafka, streaming APIs, or change data capture (CDC).
6. Strong Governance, Security, and Compliance
   • Fine-grained access control (RBAC, row/column-level security, dynamic data masking).
   • Data lineage, cataloguing, auditing, anonymising, and encryption.
   • Compliance with GDPR, HIPAA, SOC 2, ISO 27001, etc.
7. AI/ML and Advanced Analytics Integration
   • In-database machine learning and statistical functions.
   • Vector search and semantic capabilities for LLMs.
8. Self-Service and Democratized Access
   • Business users access data via familiar BI tools (Tableau, Power BI, Looker).
   • Semantic layers for consistent definitions and metrics.

Key Considerations When Designing or Choosing a Modern Data Warehouse

1. Workload Patterns
   • High-concurrency BI/reporting ➡️ fast query performance and scalability.
   • Large-scale analytics and ML ➡️ strong in-database processing and AI features.
   • Real-time needs ➡️ low-latency ingestion and querying.
2. Cost Management
   • Pay-per-query or pay-per-compute-second models.
   • Use auto-suspend, query acceleration, and resource monitors to control spend.
   • Avoid unexpected bills by setting usage limits.
3. Data Volume and Growth
   • Must handle petabytes of data economically.
   • Storage should remain cheap and independent of compute.
4. Performance Optimisation
   • Clustering, partitioning, materialised views, and caching.
   • Query acceleration and indexing features.
5. Cloud Provider Alignment
   • Choose a platform that integrates well with your primary cloud (AWS, Azure, GCP).
   • Consider single-cloud vs. multi-cloud flexibility.
6. Team Skills
   • SQL-first platforms require minimal specialised training.
   • Avoid heavy dependency on niche skills unless needed for specific workloads.

Leading Modern Data Warehouse Platforms (2025)

Platform	Strengths	Best For
Snowflake	Ease of use, strong governance, separation of storage/compute	Enterprise BI, self-service analytics
Google BigQuery	Serverless, fast analytics, built-in ML	Large-scale analytics, Google Cloud users
Amazon Redshift	Cost-effective, Spectrum for querying S3, zero-ETL	AWS-centric, large BI workloads
Microsoft Fabric	Integrated with Power BI, unified analytics	Microsoft ecosystem, end-to-end analytics

A Fundamental Question

So, does a modern data warehouse still follow Inmon’s integration, subject-oriented, non-volatile, and time-variant principles?

The shortest answer is yes. The modern Data Warehouses still follow Inmon’s core principles, but with important modern adaptations

Bill Inmon, often called the “father of the data warehouse,” defined four foundational principles in the 1990s:

1. Subject-oriented
2. Integrated
3. Non-volatile
4. Time-variant

Modern cloud-native data warehouses (Snowflake, BigQuery, Redshift, Fabric, etc.) still adhere to these principles in spirit. Still, their implementation has evolved significantly due to cloud architecture, real-time data, and broader use cases.

How Modern Data Warehouses Align with Inmon’s Principles

Principle	Inmon’s Original Definition (1990s)	How Modern Data Warehouses Implement It (2025)
Subject-oriented	Organised around business subjects (e.g., customers, sales, inventory) rather than applications	Still true: modern warehouses are typically structured around subject areas. They use dimensional modelling like star or snowflake schemas. Semantic layers such as dbt and Looker are applied. Data products are also employed.
Integrated	Data from multiple sources is cleansed, standardised, and unified into a consistent view.	Still core: ETL/ELT pipelines, data catalogs, and governance tools (e.g., Snowflake Horizon, BigQuery Data Catalog) ensure integration. Master data management and golden records remain essential.
Non-volatile	Data is read-only; once loaded, it is not modified (historical data preserved)	Mostly true: modern warehouses support updates/deletes (ACID transactions), but core analytics data is typically append-only or slowly changing dimensions (SCD). Real-time updates are handled via streams or delta processing without losing history.
Time-variant	Data is time-stamped and historical; supports “as-of” reporting	Strongly true: timestamped tables, slowly changing dimensions (Type 2 SCD), temporal queries (e.g., Snowflake’s TIME TRAVEL, BigQuery’s SYSTEM_TIME), and time-series support preserve historical views.

Key Evolutions and Adaptations in Modern Warehouses

While the principles remain foundational, modern implementations differ from Inmon’s original vision in these ways:

1. Non-volatile is less rigid
   • Inmon envisioned a purely read-only, historical repository.
   • Modern warehouses support complete ACID transactions, allowing updates, deletes, and merges (e.g., for GDPR compliance or corrections).
   • However, most analytical workloads still treat data as append-only or use change tracking (streams, CDC) to preserve history.
2. Real-time and streaming data
   • Inmon’s model was batch-oriented.
   • Modern warehouses handle near-real-time ingestion and querying (Snowflake Streams, BigQuery Streaming, Redshift Streaming) while maintaining historical integrity.
3. Cloud-native elasticity
   • Inmon assumed fixed on-premises hardware.
   • Today’s warehouses scale compute and storage independently, making them far more flexible while preserving the logical principles.
4. Self-service and semantic layers
   • Inmon relied on IT-controlled ETL.
   • Today, semantic layers and data modelling tools like Erwin, Sparx, Power Designer, dbt, Atlan, and Looker provide integrated views. They allow business users to access these views without deep technical knowledge.

Summary

To summarise, modern data warehousing has evolved dramatically. It has shifted from the more rigid, expensive, on-premises systems of the past. These systems are now cloud-native and highly scalable. They are optimised for today’s data-driven needs. These platforms now serve as the trusted foundation for business intelligence, self-service analytics, real-time insights, machine learning, and AI applications. They are driven by exploding data volumes. There is also the demand for near-real-time decision-making and the power of cloud computing.

The core principles of modern data warehousing include:

• Fully managed, elastic cloud-native architectures with decoupled storage and compute
• Support for diverse data types (structured, semi-structured, and unstructured)
• Full ACID compliance and strong governance, security, and compliance features
• Real-time ingestion and querying capabilities
• In-database AI/ML integration and vector search
• Self-service access through semantic layers and familiar BI tools

Key considerations for designing or selecting a modern data warehouse include workload patterns and cost management. Other factors are data volume and performance optimisation. It should also align with cloud providers and be supported by the team’s skills. Leading platforms in 2025, Snowflake, Google BigQuery, Amazon Redshift, and Microsoft Fabric, offer enterprise-grade performance and predictable costs.

A fundamental question remains: do these modern systems still follow Bill Inmon’s four original principles of the data warehouse? These principles are subject-oriented, integrated, non-volatile, and time-variant. The answer is yes. Contemporary cloud-native warehouses continue to embody Inmon’s vision of a trusted, integrated, historical source of truth. However, they implement these principles with significant modern adaptations, such as controlled mutability, real-time streaming, cloud elasticity, and self-service capabilities.

Modern data warehousing ultimately prioritises fast, reliable insights at scale. It ensures strong governance and accessibility for both technical and business users. At the same time, it maintains the foundational goals of integration, consistency, and historical integrity that Inmon first defined.

Many thanks for reading.