Confluent Makes Lightning Tables Generally Available on AWS, With Latency Figures From Its Own Internal Testing; an IDC Analyst Asks for Benchmarks

Overview

Confluent, which describes itself as an IBM company, said on November 4, 2026 that Lightning Tables, a read-only serving layer for Apache Kafka data, is generally available on AWS, according to IBM’s newsroom. Confluent counts it as part of Query, the fifth pillar of its Data Streaming Platform. The release labels the performance figures in the announcement as coming from Confluent’s internal testing, and an IDC analyst said to TechTarget that it “would be nice to see Confluent publish benchmarks and reference architectures for Lightning Tables.”

What Confluent Announced

According to the IBM newsroom release, Lightning Tables on Confluent Cloud durably stores and materializes Kafka data into tables and keeps them current as changes arise. It can be queried through SQL and a direct API, and Confluent says it serves applications without separate ETL pipelines, databases or caches. Confluent’s Q4 launch blog post adds that fresh context can also be exposed to AI agents through the Real-Time Context Engine over the Model Context Protocol (MCP), and that Lightning Tables supports BYOK, backward-compatible schema evolution, and usage and performance metrics.

The release lists three capabilities under Query: Tableflow, which makes streaming data available to lakehouses and warehouses as open table formats; the Real-Time Context Engine; and Lightning Tables. The blog post also describes Snapshot Queries and Streaming Queries in its Query section. It names six lead items for the release: Lightning Tables, Managed Model Inference, Schema Registry updates, multi-tenancy for Confluent Private Cloud, Confluent Gateway and Tableflow updates.

What the Performance Figures Rest On

The IBM newsroom release gives “approximately 100 milliseconds from new data creation to serving, with query latency of 10-100 milliseconds and throughput of 1,000+ queries per second” and attributes them to Confluent internal testing. These are the vendor’s figures, and the release does not say how the tests were run.

Confluent’s blog post gives the same query latency and throughput but words data freshness differently: it says changes are available for serving “within hundreds of milliseconds.” The text of the blog post that the Herald read does not use the phrase internal testing. Neither Confluent page explains the difference between “approximately 100 milliseconds” and “within hundreds of milliseconds,” and the Herald does not reconcile them.

On the Herald’s own reading of the five pages cited in this article, none gives a Lightning Tables price, a maximum table size, or the data set, cluster configuration and workload behind the figures.

According to TechTarget, Stewart Bond, an analyst at IDC, suggested that Confluent provide evidence of how well Lightning Tables and its other new features work, saying: “It would be nice to see Confluent publish benchmarks and reference architectures for Lightning Tables.” Sean Falconer, Confluent’s vice president of product for AI products and strategy, said the company will “keep pushing on how much data Lightning Tables can handle and on overall query throughput,” per the same article.

What an Analyst Said

TechTarget reports that Bond called the new features “moderately” significant additions, and it describes Lightning Tables as the most valuable for Confluent customers. Bond said: “Lightning Tables stand out because they change the read path,” but “speed and ease of use alone will not advance the real-time imperative.” He noted an IDC survey from earlier this year that, as TechTarget describes it, found data trust and quality, rather than latency, to be the top technical barrier to real-time data initiatives.

The same article reports that Bond noted remaining gaps: to deliver a consistent governed view, TechTarget writes, Confluent needs to add capabilities such as open semantic standards and ontologies, agent governance including lineage, and more widespread hybrid connections. It also reports that Bond noted Confluent’s streaming data capabilities are the most complete across brokering, stream processing, data serving and context and governance, with Microsoft perhaps its closest rival, and it quotes him: “Microsoft is closest on event-to-semantics-to-action with Fabric Real-Time Intelligence and Fabric IQ.” Of Confluent, he said that what it lacks in coverage “it makes up for in scale” via the IBM acquisition.

The Streamhouse Label

Confluent frames the launch around the Streamhouse, which the IBM release calls a new type of data architecture. The term predates this launch. On September 15, 2026, Aiven, Confluent, Redpanda, StreamNative and Ververica announced the Streamhouse Working Group and an open definition with three attributes: real-time, production-native and decentralized. The announcement says the name was coined by Ververica, that organizations can build a Streamhouse with open-source technologies, commercial products or both, and that no single project, component or vendor portfolio constitutes one on its own. The IBM release says Confluent’s Data Streaming Platform provides a complete, unified foundation for building a Streamhouse.

In a TechTarget article published September 22, 2026, Donald Farmer, founder and principal of TreeHive Strategy, said of the definition: “The streamhouse definition provides a vendor-neutral blueprint.” Kevin Petrie, an analyst at BARC U.S., said at that point that the Streamhouse website documentation was “pretty sparse so far” and would benefit from a more detailed reference architecture based on customer implementations.

What Is Not Settled

• Cloud coverage. The general availability announcement covers AWS. TechTarget reports that Confluent will focus on expanding capabilities to all public clouds; none of the cited pages gives dates for Lightning Tables on other clouds.
• IBM acquisition date. TechTarget writes that IBM acquired Confluent in December 2025. The Herald previously reported that IBM completed the acquisition on March 17, 2026. TechTarget does not say which event it means, and this article does not reconcile the two.
• Independent testing. None of the cited pages reports a Lightning Tables benchmark by a party other than Confluent.