NextFin

From ‘Dataslows’ to Dataflows: The Gen2 Performance Revolution in Microsoft Fabric

Summarized by NextFin AI
  • Microsoft announced significant improvements for its Dataflow Gen2 component at FabCon Europe, enhancing performance and reducing costs for data transformation tasks.
  • The new tiered pricing model lowers costs by over 50% for extended workloads, charging 12 CUs per second for the first 10 minutes and 1.5 CUs thereafter.
  • Performance enhancements include the Modern Evaluator and Partitioned Compute, which accelerate execution and improve processing efficiency, with the Modern Evaluator achieving a 30% reduction in execution time.
  • These advancements position Dataflows Gen2 as a viable alternative to code-first solutions, promoting broader adoption of Microsoft Fabric as a unified data platform.

NextFin News - At the recent FabCon Europe conference held in Vienna in early January 2026, Microsoft announced substantial performance and cost improvements for its Dataflow Gen2 component within the Microsoft Fabric platform. Dataflows Gen2, a no-code/low-code ETL (extract, transform, load) tool, enables users to ingest and transform data from over 100 built-in connectors using a Power Query GUI, storing outputs as delta tables in OneLake for downstream consumption by Fabric engines such as Spark, T-SQL, and Power BI.

Historically, Dataflows Gen2 faced criticism for high Capacity Unit (CU) consumption compared to code-first alternatives like Fabric notebooks and T-SQL scripts. Previously, Dataflow Gen2 executions were billed at a flat rate of 16 CUs per second, leading to significant costs for longer-running data transformations. Microsoft’s new pricing model introduces a tiered CU billing structure: the first 10 minutes of execution are charged at 12 CUs per second, with subsequent time billed at a reduced 1.5 CUs per second. For example, a 20-minute Dataflow run now costs 8,100 CUs versus the prior 19,200 CUs, representing a potential CU saving of over 50% for extended workloads.

Complementing pricing changes, Microsoft introduced two key performance enhancements: the Modern Evaluator and Partitioned Compute. The Modern Evaluator is a new query execution engine built on .NET Core 8, designed to accelerate Dataflow execution, improve processing efficiency, and enhance scalability and reliability. Partitioned Compute enables parallel execution of transformation logic segments, reducing overall processing time by overcoming the previous sequential execution bottleneck. Currently, Partitioned Compute supports connectors such as ADLS Gen2, Fabric Lakehouse, Folder, and Azure Blob Storage.

Independent benchmarking conducted by industry experts using a dataset of approximately 29 million records (2.5 GB) across 50 CSV files stored in SharePoint demonstrated the tangible benefits of these innovations. Four scenarios were tested: Dataflow Gen1 (legacy Power BI dataflow), Dataflow Gen2 without optimizations, Dataflow Gen2 with Modern Evaluator enabled, and Dataflow Gen2 with both Modern Evaluator and Partitioned Compute enabled.

Results showed that Dataflow Gen2 with Modern Evaluator outperformed all other configurations, achieving a 30% reduction in execution time compared to Dataflow Gen1 and approximately 20% faster than unoptimized Dataflow Gen2. CU consumption metrics from the Capacity Metrics App corroborated these findings, indicating that the Modern Evaluator reduced CU usage by more than 50% relative to Dataflow Gen1. Interestingly, Partitioned Compute’s impact was scenario-dependent and less pronounced in this test, suggesting further exploration is needed to optimize its benefits across diverse workloads.

These advancements address the longstanding trade-off between ease of use and operational cost/performance in no-code data integration tools. By significantly lowering CU consumption and accelerating execution times, Microsoft is positioning Dataflows Gen2 as a viable alternative to code-first solutions, expanding its appeal to citizen developers and enterprises seeking scalable, cost-effective data transformation capabilities within Fabric.

Looking ahead, the evolution of Dataflows Gen2 is likely to catalyze broader adoption of Microsoft Fabric as a unified data platform. The integration of delta tables in OneLake facilitates seamless downstream analytics and AI workloads, while ongoing enhancements in query execution and parallelism promise further performance gains. As Microsoft continues to refine pricing models and feature sets, organizations can anticipate improved total cost of ownership and faster time-to-insight, critical factors in competitive data-driven environments.

Moreover, the modular architecture of Dataflows Gen2, combined with the Modern Evaluator’s .NET Core foundation, opens avenues for future integration with AI-powered optimization and adaptive workload management. This aligns with Microsoft’s broader strategy to embed AI capabilities across Fabric, enhancing automation and intelligent data processing.

In conclusion, the Gen2 performance revolution in Microsoft Fabric’s Dataflows marks a significant milestone in democratizing data engineering. By bridging the gap between no-code simplicity and enterprise-grade efficiency, Microsoft is empowering a wider spectrum of users to harness the full potential of their data estates with agility and cost discipline.

According to the detailed analysis published on Towards Data Science on January 13, 2026, these innovations not only reduce operational costs but also improve scalability and reliability, setting a new standard for cloud-native data integration platforms.

Explore more exclusive insights at nextfin.ai.

Insights

What are the main features of Dataflows Gen2 in Microsoft Fabric?

How did Dataflows Gen2 evolve from its previous versions?

What are the key technical principles behind the Modern Evaluator and Partitioned Compute?

What recent changes has Microsoft made to the pricing model for Dataflows Gen2?

How has user feedback influenced the design and functionality of Dataflows Gen2?

What are the current market trends for no-code data integration tools?

What recent performance benchmarks have been published regarding Dataflows Gen2?

What future developments can be expected in Microsoft Fabric's Dataflows Gen2?

What challenges does Dataflows Gen2 face in competing with code-first solutions?

How does Dataflows Gen2 compare to traditional ETL tools in terms of cost and performance?

What are the limitations of the Partitioned Compute feature in Dataflows Gen2?

How are delta tables utilized in Dataflows Gen2 for data storage?

What impact does the integration of AI capabilities have on Dataflows Gen2's future?

What feedback did industry experts give regarding the performance of Dataflows Gen2?

How does Dataflows Gen2 improve scalability and reliability for users?

What are the implications of Microsoft’s tiered CU billing structure for users?

What historical criticisms were directed at Dataflows Gen2 before the updates?

How does the modular architecture of Dataflows Gen2 support future integrations?

What role do citizen developers play in the adoption of Dataflows Gen2?

Search
NextFinNextFin
NextFin.Al
No Noise, only Signal.
Open App