IBM paid $11 billion for Confluent; Confluent's CTO says the data still refreshes every 12 hours

IBM▼−0.32% — IBM, down 0.32 percent today paid $11 billion for Confluent to put real-time data under AI agents ³. Five days after the Q3 print made half of IBM's software growth look bought, Confluent's own chief technology officer put the deeper problem on the record: Stephen Deasy told Frontier Enterprise↗ that parts of a company's data estate still refresh only every 12 hours, and the pipelines were designed so a person could review the output before anyone acted ¹. The stall is not in the data platform; it is in the architecture around it.

Software grew 8%, and UBS counts about 1% organic

Software revenue grew 8% to $7.8 billion in the quarter↗, with the Data line up 19% as reported and 20% at constant currency, which is where Confluent now sits ². Chief financial officer Jim Kavanaugh said about three-quarters of the software growth was net new business and about a quarter came from roughly $400 million of deals that slipped out of Q2 ².

The dissenting reading is UBS's: strip out that catch-up and the acquired revenue from Confluent and HashiCorp, and organic software growth is about 1% ². The acquired revenue stops counting as growth when it laps in March 2027 ². Whatever IBM built on its own is what remains after that date.

Deasy's diagnosis, and the chain it belongs to

Enterprises built operational systems and analytical estates and connected them with batch pipelines whose output landed in a lakehouse, was modeled, and waited for a human to review ¹. An AI agent breaks that arrangement because it decides continuously and talks to other software, so its data has to be current, governed and drawn from across the company, not sitting in a batch window ¹.

His example lands the human stake: an airline passenger should leave the plane already rebooked with a gate text, not find the cancellation on a departure board and queue at a desk where nobody has heard of them ¹.

Governance moved to the source, or the premium is unearned

Follow the consequences and the stakes for IBM's price come into view. Deasy argues for moving governance to the source, as close to where events are generated as possible, so every downstream consumer inherits it ¹. If IBM executes that shift, the Confluent purchase becomes the foundation of an agent stack enterprises actually trust, and the Data segment's 20% constant-currency growth is early evidence rather than a purchase receipt ².

If IBM does not rebuild the batch plumbing its customers run on, the acquisition sells into estates that stay twelve hours stale, and the buyer's premium rests on demand that has not arrived.

The verdict waits on January and the March lap

Krishna told analysts that enterprises are asking what the true return on their AI spending is, and some worry it is completely out of control with no real return ². Management quantifies 2027 only in January, when the Street has pencilled in roughly 7.5% software growth that UBS warns could fall below once Confluent's contribution laps ². The shares closed October 23 at $253.56, near the $250.45 200-day average and about 1.5% under the $257.22 consensus target, with this week's Microsoft▲+1.08% — Microsoft, up 1.08 percent today, Alphabet▼−0.35% — Alphabet, down 0.35 percent today and Amazon▲+0.44% — Amazon, up 0.44 percent today prints setting the read on enterprise AI budgets ².

IBM bought the right layer; the lap date will show whether it rebuilt what sits around it. Whoever turns a model's capability into something a person's workday uses is the one creating the value, and that work here is architectural, not commercial.