Event Streaming is Topping Out

There's no money left in real-time event streaming. A major consolidation wave is beginning.

Nov 03, 2025

event streaming heading towards the event horizon, and the death of Kafka (as a business)

I thought I’d capture in words a feeling a lot of people in the industry are experiencing, but nobody is publicly talking about.

The stream is oversubscribed. In a real sense - there are too many companies chasing the same relatively-small market of pub-sub/event streaming as well as the microscopic market of stream processing.

My prediction is simple - we will see a lot of consolidation. I believe we’re in the beginning of that consolidation. Here are the reasons why:

1. The Leader is Not Well 🤒

Confluent is undisputedly the leader in event streaming and stream processing. Confluent is not performing well.

The stock price is the best concise indicator. It IPO’d at $36, peaked at $91 and has been largely trading between $20-30 for the last 3+ years.

That is awful performance plain and simple.

Even if we ignore the $91 peak and call it bubbly times, as of writing it’s still -36% off from its IPO price. As someone who tracks the stock, I can’t express how stuck it feels between $20-30.

Why is the stock like that?

Confluent’s growth is topping out. Their cloud product, the golden child which grew fast, is decelerating rapidly.

Confluent **Cloud** YoY revenue growth, alongside quarterly revenue and quarterly nominal increases. You can see they’re quite visibly struggling to add more than $8-10M a quarter.

What’s worse is that this 24% growth rate is just for their cloud product. Another substantial part of the business comes from on-premise sales (Confluent Platform). That part, $125.4M for Q3/25, grew by just 14%.

The total is a 19.4% revenue growth rate.

2025 isn’t over yet and Confluent’s growth rate is already going into the teens. If the next quarters trend the same way, we may see single-digit growth by end of next year.

It’s no small feat to reach this revenue, but the market values you relative to its other opportunities. And the others are doing much better:

Confluent has $1.1B of ARR and is growing at 19.4%
Snowflake has $4.1B TTM and is growing at 29.2%
Databricks is at ~$4B and growing 50% (allegedly)
(an extreme, somewhat unrelated example) Azure has $75B of ARR and is growing at 39%

As you can see, the total addressable market for the products Confluent sells seems to be significantly smaller than what Snowflake/Databricks are selling.

These financials are the single best sign that streaming is topping out. You can’t fake them with a narrative.

The rest of the things I will present are more qualitative, but still interesting.

Confluent == Kafka

Confluent’s performance matters a whole lot with relation to Kafka because they are the creators of Kafka and the leaders in Kafka. Basically they ARE Kafka:

An analysis of 85%+ of open-source Apache Kafka contributions, from 2010-2025.

They don’t post a breakdown of product revenue, but recently shared Flink’s numbers. It has barely registered around $14M ARR. Out of $1B total, this is hardly going to be the savior of the company’s declining growth. Outside of Kafka and Flink, I don’t see other products that have the potential to make substantial money for the company.

The majority of their revenue, therefore, is allegedly coming from Kafka. When we look at it from that point of view, we start to see very reasonable explanations for the declining growth:

2. Competition is Hotter than Ever 🔥

Kafka is becoming a commodity. Everybody and their abuela has a Kafka service now.

AWS MSK has been out for 6 years now. It’s rumored to make more than Confluent’s entire business ($1B-$1.5B of ARR)1
Google Cloud Kafka (announced Jul 2024)
Oracle Kafka (announced Aug 2025)
Azure Event Hubs (has a Kafka API)
OVHCloud Kafka
Heroku Kafka
Canonical (the guys behind Ubuntu) Kafka (released Feb 2024)

And startups:

Aiven
Redpanda
StreamNative (Pulsar, but Kafka API)
AutoMQ
Bufstream
Instaclustr Kafka

It’s only the behemoths Snowflake and Databricks that still haven’t entered the space. But they most likely will. Snowflake almost bought Redpanda in early 2025.

These behemoths are in a good position, because they can wait to see how the market pans out and shop around for a deal. It’s a matter of time until they buy some of the above.

By the way, what I presented was the Kafka market. There are plenty of alternative competitors2 too:

Google Pub Sub
Azure Event Hubs
Cloudflare Pipelines
Redis Streams
s2.dev (small startup)
NATS

3. Kafka Prices Are Going Lower than Ever 📉

Kafka prices are seeing major deflation.

Kafka is available at every neighbourhood corner store now—that ubiquity alone drove prices down some ~6.5x.3

Unfortunately for the industry, the real price war started in 2023 when a new architecture showed you can instantly deflate Kafka prices by 10x+.

WarpStream pioneered the diskless Kafka architecture where instead of storing data on local broker disks and replicating between them, this job is outsourced to S3. It comes at a larger latency (slower), but can completely eliminate networking costs in Kafka.4

This was an incredibly major development because a well-optimized Kafka deployment has 80% of its costs coming from cross-AZ networking. In other words, $176k out of a $215k annual bill can go to networking.

Further, it’s precisely latency-insensitive workloads that drive this high volume network bandwidth. In other words, the workloads that don’t care about latency are the ones that drive the highest networking costs. They are therefore ripe for a cost reduction through this architecture.

This architecture was gated behind proprietary products for 2023/2024, but in 2025 Aiven began releasing it to open source Apache Kafka.

Not only that, but Aiven was also the first to unlock both fast topics and slow topics in the same distribution. Previously, you had to commit to one or the other, which slowed adoption as users inevitably have mixed workloads.

Once Diskless gets accepted and merged into the open-source project (ETA ~20275 would be my guess), savvy operators will be able to reduce their Kafka networking costs by literally 97% (!).

👉 see the calculations in Diskless Kafka is a big deal

4. Stream Processing Is NOT Making Money 🦗

OK, so Kafka is most certainly not going to be a revenue growth driver going forward. Can companies make up the difference from stream processing?

No. Definitively.

Confluent, for their whole lifetime, failed to significantly monetize stream processing.

~2016: they released Kafka Streams. They tried to commercialize it by marketing event-driven architectures (EDA)
~2020: they tried simplify stream processing by adding a SQL interface on top with ksqlDB
~2023: they abandoned ksql, bought a Flink startup and went all in on Flink
~2024: they market the hell out of Flink for AI Agents (similar story to microservices)
Q3 2025: out of a ~$1.1B annual recurring revenue, Flink makes just ~$14M ARR6 (1.25%)

While this is disastrous for Confluent, it’s even more brutal for the other stream processing startups in the space.

If the world’s best stream processing experts inside a company with massive distribution7 and capital can’t figure out how to monetize stream processing properly— I don’t know who can.

Take Responsive.dev, a company founded by and consisting of top-tier ex-Confluent talent8. It was a boutique Kafka Streams startup, focused on making streaming apps more reliable and scalable. It recently completely pivoted away from streaming and into observability.

5. All The Real-Time Companies Are Small 🐁

Outside of Confluent, who’s left in the general real-time space?

🎯 Timeplus
🎯 RisingWave
🎯 Materialize
🎯 Deltastream
🎯 Ververica
🎯 Factor House
🎯 Conduktor
🏃‍♂️ Redpanda (pivoted to AI Agents Oct 2025)
💰 Arroyo (acquired Apr 2025)
💰 Decodable (acquired Sep 2025)

Most of these companies are miniscule.

Redpanda is the biggest and most established of the bunch. They are well-known, have won the hearts of many developers and are selling into a proven billion-dollar market (Kafka).

After 6 years of operations, Redpanda is rumored to have barely crossed $20M revenue in 2025.9

If Redpanda can’t meaningfully catch up to Confluent’s Kafka revenue, and Confluent’s stream processing revenue is any indicator of the size of the real-time processing market — what hope is there for the others?

The way I see it, these companies have four choices:

pivot into a larger market and keep growing
get acquired
go bankrupt

Pick one.

The one standout that struck it big was WarpStream.

They started their company in 2023 and got acquired in 2024 for $220M. Confluent’s panic paired with Warpstream’s skill, luck and positioning resulted in an exit the whole streaming world envies.

It is very unlikely this gets replicated again.10

6. The Industry Is Waking Up 🥱

Slowly but surely, people are waking up to the fact that streaming has topped out. They’re either losing interest or pushing for simpler alternatives. You can hear it if you listen closely:

Listen to the Conferences 👂

Current, the conference formerly known as Kafka Summit, seems to be losing steam. Here are a few pointers:

Redpanda was banned from it.
The London conference (May, 2025) showed signs of shrinking.

I spoke to people who thought it was shrinking in size

Confluent was seen giving out free tickets on social media 26 days before the flagship New Orleans conference (Oct 2025). A ticket that people paid $500-1000 for started being given out for free a full month before the conference. What does that tell you?

Listen to the Pundits 👂

Here are the thoughts of two people11 with multi-decade experience in stream processing and data engineering - Chris Ricommini and Ryanne Dolan:

Yaroslav Tkachenko

shared his thoughts from the May 2025, Current London conference:

Finally, I feel like the data streaming industry is still in a tough spot. The growth is slow, and the sales cycles are long.
One person I spoke with said that “80% of the companies in the Expo hall will be dead in two years”.
I don’t want to believe them, but it might be true.

(note it wasn’t me he spoke to, lol)

In a post titled “Kafka: The End of the Beginning”, Chris Ricommini also echoed his general thoughts on the market following that quote from Yaroslav:

I was speaking with someone just yesterday that said something similar: it’s the same people, same companies, and same technology. There are no new ideas and no new users.
This person said they would not attend future Current conferences; it wasn’t worth it; they didn’t learn anything new.
Yaroslav’s post also says that Redpanda was banned from the conference. I don’t know the details, nor do I really care. But if true, it’s hard not to read this as an indication of scarcity and fear, not one of abundance.
[…]
I’ve been thinking about this stagnation for a while. I don’t believe we’ve solved everything. It’s still very difficult to write and deploy stream processing jobs, for example; it honestly feels like it’s gotten worse, not better. I’d like to believe that these challenges are not fundamental—that there are better ways.
[..]
Meanwhile, the stream processing side of the house seems somehow both worse and better. I am no fan of Flink, but it has won the stream processing race. Unspoken amongst vendors, the true winner is really just regular old Kafka consumers and producers.

Listen to the People👂

For lack of a better word, there is a “Simplification” movement forming post the ZIRP hangover of the industry. Few examples:

Kafka’s 80% Problem - Aiven recently admitted most Kafka workloads are small (< 10MB/s or even < 1MB/s) and that Kafka has overhead costs that are too high for such modest workloads
Flinks’s 95% Problem - Tinybird points out Flink is over-engineered complexity for the majority of use cases it claims to solve

Look no further than the reception I get from my own posts.

My recent post where I claimed Postgres can be used to cover a majority of Kafka and queue use cases went viral and generated >60k blog views.
My tweet citing both 80%/95% articles went viral too

This is all heavily pointing to the fact that there seems to be large agreement forming around the fact that the industry is overcomplicated with relation to what the majority of users need.

Why Are We At The Event Horizon?

It’s human nature to want to point the finger at somebody. I wouldn’t attribute this malperformance to bad management of any single company.

I don’t think the state we’re in today is due to a lack of talent or bad strategy.

I think there simply isn’t a market big enough for everybody. Regardless of how bright and resourceful you are, it’s an uphill battle to sell something people do not want.

What’s Next?

Widespread consolidation. It’s time to bundle up.

As that person in the conference allegedly said - “80% of the companies in the Expo hall will be dead in two years”.

I think more acquisitions will happen in an effort to bundle these products with other general data infra. For example, Confluent was reportedly in talks of being acquired recently. I believe this will happen with high likelihood.

The Decodable and Arroyo acquisitions I don’t believe were done from a position of strength either. I assume they were necessary — they realized independent growth wasn’t a winning strategy and somebody saw they can bundle them with their offering nicely.

For the companies that haven’t gone for sale (yet), a pivot is likely in the works.

A great recent example is Redpanda pivoting to AI Agents.

What About Kafka? 🥺

Apache Kafka isn’t going anywhere, but the companies that manage it are.

I don’t see any real competition that can replace Kafka. I just see a lot of overextended VC-funded companies that can’t justify their growth to investors anymore.

Open source Kafka will continue to exist, and with some luck, maybe continue to thrive. There is a real risk that a wide gap in the market may be left when Confluent gets acquired and the fallout from that. I don’t see many players in a position to fill such a theoretical gap.

Maybe it’s time for a leaner, more sustainable startup around Kafka? DM me if interested.

Why Am I Telling You This? 🥸

As “The Kafka Guy”, I am the last person who should be sharing such negative views about the industry.

But being the Kafka guy doesn’t mean I have to shill things I don’t believe in. I’m just using my platform to share what I believe is the truth.12

Before some vendors try to discredit me - I’m not trying to be doomerist for clicks. I really, truly believe this.

I think what’s said here is pretty obvious for anybody who isn’t deep down in the echo chamber drinking the kool-aid and suffering from sunken-cost bias. If you don’t agree - set a reminder and revisit this piece in two years.

I also hold no ego in these predictions. In fact, I’ll be happier if I’m wrong. I make money off the same industry that’s in decline.

In the end, I’m all for the industry and don’t want to see it go away.

The less we bullshit each other and the sooner we face reality, the faster we can get over this bump. ✌️

Disagree? Comment why.
A lot of numbers were shared here, so please correct me if I got anything wrong.