What does the IBM acquisition of Confluent mean for the future of streaming and Kafka?

On December 8th, 2025, IBM announced a definitive agreement to acquire Confluent in a deal valued at $11 billion. It is a massive moment for our industry. The acquisition was finalized on March 17th, 2026.

For some, this looks like a safe bet; a way for enterprise giants to finally "get" real-time data. But for those of us who have spent our careers in open source software and data infrastructure, it feels different. There’s a sense of wondering “when is the other shoe going to drop?”.

What does this mean for your roadmap, your budget, and your ability to build the next generation of AI-driven applications?

The $11 Billion Math: Why Prices May Rise

Let’s look at the numbers:

• The Multiple

  IBM paid roughly 10x trailing revenue, for a company growing at 21% with significantly negative operating margins; another way to look at it is a 34% premium on Confluent’s stock prior to the announcement (and nearly 50% over the October 2025 lows).

• The Profitability Gap

  Despite scaling to over $1B in revenue, Confluent still reported a net loss of nearly $295 million in 2025 (around -27% to -32% GAAP).

When a large corporation like IBM spends $11B of cash-on-hand, they are looking for a return on that investment. IBM have said they expect accretive EBITA at the end of year one and cashflow positive at the end of year two. There are only two paths to those results on those timelines: cost-cutting and increased monetization.

Cost cutting is part of any acquisition so let’s set that aside and take a look at what has transpired since IBM’s 2019 acquisition of RedHat. While history doesn’t repeat itself directly, it certainly rhymes.

The IBM Cycle

Contrary to other companies, IBM does generally avoid initial “sticker shock” price hikes for existing products. There were no major increases in RedHat licensing post acquisition, however, the economics of access changed dramatically:

• Gatekeeping free REHL

  IBM eliminated CentOS, the free and stable clone of RHEL. Access to the RHEL source code was restricted behind a subscriber-only portal. Customers were left with two choices: move to the less stable CentOS Stream or pony up for RHEL subscriptions.

• Shift to Consumption & Bundle Pricing

  IBM moved away from simple per-product licenses. RedHat software (non RHEL) is now in bundles called “Cloud Paks” resulting in customers paying for a stack of software rather than single products. Either pay higher a-la-carte pricing, or attempt to negotiate on a bundle of software you may not want.

• Shadow Price Increases

  While not reflected in direct product pricing, there are two additional areas of price increases behind the scenes. There’s no free support option or tier. The only access to any tier of support is through an active subscription. IBM is also known to be aggressive with software audits. All businesses want to be 100% in compliance, 100% of the time, with the commercial software they use. They don’t want to spend months of time, effort, and cost to prove compliance with no benefit to their own endeavors.

When a company spends $11 billion, they aren't looking for "community growth." They are looking for an Internal Rate of Return (IRR). That return can only come from a very short list of places: increasing prices, forcing product bundles, and cutting the "distraction" of community-driven innovation in favor of high-margin proprietary features.

I will cover this later, but Aiven is fundamentally different in our open source philosophy.

The "Streaming Tax" is Getting More Expensive

Running Kafka at scale has been plagued by a "Streaming Tax" that most teams just accept as the cost of doing business. But as your data grows, this tax becomes unsustainable. This is especially true for Kafka in the cloud.

Let’s look at the numbers:

• The Replication Tax: Default Kafka architecture requires 3x replication for durability. If you have 100TB of data, you’re paying for 300TB of expensive block storage.
• The Networking Tax: The silent killer. Most hyperscalers charge $0.01 to $0.02 per GB for data moving between Availability Zones (AZs).
• The Impact: In a typical high-availability Kafka cluster, with brokers in the same region but in three different availability zones, cross-AZ replication can account for 40% to 60% of your total infrastructure bill.

Kafka simply wasn’t initially designed to run natively in the cloud. This is NOT a knock against that initial design. Remember, the first open source release was in 2011! Kafka has become the de facto standard for event and stream processing. Here at Aiven, we’ve very publicly worked to dramatically improve the fundamental economics of Kafka in the cloud. Again… more to come in a bit.

Why "Data Ownership" is the Only AI Strategy

We are moving into a world of Agentic AI. This isn't just about chatbots; it’s about autonomous agents that can plan, reason, and execute tasks.

To build these agents, your data infrastructure needs to be the "nervous system" of your company. It needs to provide:

• Freshness: Agents making decisions on 10-minute-old data are useless. You need sub-second latency.
• Context: Agents need to be fed a real-time stream of what is happening across your entire business - not just what’s in a single vendor's silo.
• Independence: If your AI’s "memory" (your data stream) is locked into a proprietary fork of Kafka, you are no longer in control of your innovation.

Open-source models and frameworks already outperform their proprietary counterparts in many areas because they allow for Private AI - running models within your own trust boundary. Your data streaming layer must follow the same principle. You shouldn't have to ask a vendor for permission to access your own data architecture.

The Aiven POV: Open Source is Always Best

First, let’s revisit how Aiven is fundamentally different when it comes to open source software. We have a very simple philosophy: 100% Upstream Open Source.

We don't do "Open Core." We don't have a "Community Edition". When we build a breakthrough like KIP-1150, we build it in the community.

Why do we do this? Because your data is YOUR data! If you run on Aiven, you are running on the same open source Apache Kafka that the rest of the world uses. Full stop. There is no gatekeeping of features or forced bundles. We don’t force you to purchase a support plan or contract; of course those are available with high SLAs should you need them! You stay because we offer the best operational rigor and the lowest TCO - not because we’ve trapped your data.

A Technical Breakthrough: The "Diskless" Future

Now let’s revisit that “streaming tax”. At Aiven, we haven't been sitting on the sidelines. We believe the way to beat the "Streaming Tax" isn't through better enterprise sales - it’s through better engineering and architecture.

Our team, including Apache Kafka maintainers like Josep Prat have been a driving force behind KIP-1150 (Diskless Topics).

This is a fundamental redesign of how Kafka works in the cloud:

• Stateless Scaling: By writing data directly to object storage (like S3 or GCS) instead of relying on local broker disks, we decouple compute from storage.
• Eliminating the Streaming Tax: KIP-1150 allows for a "leaderless" design where all brokers can interact with the same data in object storage. This eliminates both the Replication Tax and Network Tax. There is no need for multiple replicas of active data or expensive cross-AZ replication traffic.
• 97% Cost Reduction: In some high-throughput scenarios (1GiB/s), moving to a diskless architecture can reduce annual infrastructure costs from $1.8M to just $20K.

This isn't a proprietary feature hidden behind open-core. KIP-1150 has been officially accepted into Apache Kafka. It is a refining of the existing open-source Apache Kafka framework to make it truly cloud-native. It turns Kafka into a utility that scales with your business, not your bill.

Zero Risk: The Confluent Takeout Program

I know what most of you are thinking: "Moving sounds great, but the 'Double Bubble' will kill my budget."

It’s the classic trap. You need to leave a platform that has become too expensive or too slow, but you can’t afford to pay for two platforms during the 2-6 month migration period.

We’ve decided to fix that. We have launched the Confluent Migration Program to make your transition risk-free and cost-neutral:

• Financial Bridge: Depending on your current spend, we provide 2 to 6 months of free service to cover your migration. We effectively zero out the "double billing" period.
• Direct Collaboration: This isn't a "do it yourself" kit. You get direct access to our engineers to ensure your cutover is seamless.
• Enterprise Perks: For larger migrations, we offer increased support programs, optimization/adoption programs, and a seat on our Customer Advisory Board (CAB) to help Aiven continue to deliver what your business needs.

Final Thoughts

If you are tired of the "Streaming Tax," if you are worried about your roadmap being "IBM-ified," or if you just want to see how much you could save by moving to a truly open, "inkless" architecture, let’s talk.

We are hosting a technical webinar on March 25th where Josep Prat and I will walk through the KIP-1150 architecture and the specifics of the Takeout Program. I’d love to see you there.

Register for the Future of Kafka Webinar