IBM's $11 billion Confluent buy ships a product that deletes ETL, cache and database layers
Confluent's first Current conference since IBM closed the $11 billion purchase shipped Lightning Tables, which serves live Kafka data with no separate database, pipeline or cache. IBM's October call put Confluent bookings up about 20%, while roughly half of IBM's software growth is the acquisition itself.
Richard Tangthree-minute read
The most consequential thing shipped at Confluent's Current 2026, the company's first conference since IBM−0.85% — IBM, down 0.85 percent today closed its $11 billion purchase in March, is a deletion 1 2. Lightning Tables, generally available on AWS this week, turns Kafka from a transport pipe into a queryable store of the current state of a business: an event lands, a table updates, an application queries it directly. No ETL pipeline, serving database or cache sits in between, with new data served in about 100 milliseconds in Confluent's own testing. 1 2
Agents act on stale context, and the fix deletes paid layers
A warehouse refreshed overnight cannot serve an agent that has to act now. Confluent's example is an airline rebooking agent that assembles an itinerary valid in three separately updated systems and wrong in reality; a Streamhouse keeps one continuously current picture and serves it to agents over the Model Context Protocol, the shared interface AI tools increasingly speak. 1
The replaced parts are specific: the pipelines, caches and serving databases that today sit between an event and a decision. IBM sells the subtraction itself, promising data products "without brittle batch pipelines or custom integration sprawl." 3 IDC's October survey ties the spending to the mechanism: agents reach production where real-time coverage is complete. 4
Agents reach production where real-time coverage is complete
Multiple agents in production climb from 8% to a majority with full coverage
Data
| Share of organizations | |
|---|---|
| Agentic apps in production, limited real-time integration | 8% |
| Agentic apps in production, comprehensive real-time integration | 50% |
| Real-time data movement extended to all workflows | 16% |
Bookings are real; the headline growth is half bought
IBM's third-quarter call, on October 21, gave the deal its first numbers 5. Data revenue grew 20%, supported by Confluent bookings also up about 20%. 5 Software grew 9%, and roughly half of that was purchased: Confluent's revenue moved onto IBM's books when the deal closed, leaving organic software growth near 4%. 2 6 The organic question we flagged in October has a partial answer: demand is visible in bookings, while the revenue line is still mostly accounting 2 6.
The load-bearing claim is vendor-tested, and that is testable
The skeptical read of the week, from BigDATAwire, is fair: the speed and savings claims rest on Confluent's internal testing, and collapsing layers deepens dependence on one supplier. 7 Shaun Clowes, who ran Confluent's product before the sale and now runs Confluent and data at IBM, answers the openness worry with Red Hat and HashiCorp, and says the rebuilding lands on integrators: "The people that can harness their data fastest are the ones who will win." 8
Accenture−1.94% — Accenture, down 1.94 percent today, already moving about 80% of its data over Kafka, says the architecture cut its time-to-insight from months to days. 4 1
An IBM shareholder is now long the deletion of the middle of the data stack. Every cache and pipeline Lightning Tables retires is budget that consolidates toward IBM's data line, and bookings say the migration has started. Whether the price earns out is not yet in any disclosure. The decidable part is the work in between, wiring a company's events into its decisions; the integrators doing that work are being paid for it, and the skill is learnable.
Get Platforms by email, free
The cloud, data and developer software AI is built on. In our daily email, every weekday.
