Brickster Newsletter

The $134B Power Play: Why Databricks is saying "No" to Wall Street

Databricks has the numbers to IPO tomorrow. It is staying private anyway.

On June 4, 2026, Databricks CEO Ali Ghodsi called 2026 "a terrible year to go public," ruling out an IPO this year. Plenty of people were waiting on a marquee debut to thaw the market, so the call stung. But it was a choice, not a stumble: Databricks has the revenue and the cash to list whenever it wants, and it is choosing to wait.

The timing math is simple. 2026 is crowded: super unicorns like OpenAI, SpaceX, and Anthropic are expected to absorb over $200 billion in public-market liquidity, and there's little upside to competing for the same investors. As of February 2026 Databricks was running at a $5.4 billion annualized revenue run-rate, growing more than 65% year over year, with AI products alone past $1.4 billion (roughly a quarter of revenue). A company posting those numbers doesn't need the public markets to prove it is real.

Staying private also buys room for expensive bets that take years to pay off: the $1.3B MosaicML acquisition, model serving, Unity Catalog as the governance layer for AI agents. None of those pay off on a public-market timeline.

The Foundation

Databricks is founded by the original creators of Apache Spark, aiming to simplify big data processing.

Challenging the Warehouse ($6.2B)

Series F funding as Delta Lake gains adoption and Databricks sharpens its pitch against the cloud data warehouse.

Hyper-Growth ($38B)

Series H values the company at $38B. Databricks expands aggressively into unified data and machine learning services.

AI Acquisition ($43B)

Acquires MosaicML for $1.3B to integrate Generative AI directly into the Lakehouse, raising at a $43B valuation.

Super-Unicorn Status ($134B)

A Series L round of more than $4B (Dec 16) values Databricks at $134B, one of the most valuable private software companies in the world.

Postponing the IPO

Reaches a $5.4B revenue run-rate (with $1.4B in AI products). Ali Ghodsi announces they will wait out a crowded IPO market.

For customers, the delay is good news. Companies prepping for an IPO often trim costs, raise prices, or slow risky R&D; to make the financials look clean for Wall Street. Databricks is under none of that pressure, so there is no reason to expect a sudden repricing or a feature freeze while it gets its house in order.

The real question is how long private can last. Snowflake is already public, and the big cloud providers keep shipping their own AI analytics. For now the math protects Databricks: at a 65%+ growth rate, staying private costs it nothing it can't make back. The risk shows up only if that growth slows.

Conversational BI

Genie gets serious, and gets a price tag

Databricks pushed Genie deeper into production conversational intelligence this week, and the community started doing the math on what it actually costs. The pricing conversation is the tell: Genie has crossed from demo to line item.

• Scaling conversational intelligence with Databricks Genie · community
• Databricks Genie pricing, explained · community

Agent Tooling

The agent toolchain fills in

A new open AI dev kit, an agent-centric certification, and a clean recipe for tracing any agent through OpenTelemetry, MLflow, and Unity Catalog. The scaffolding for building production agents on Databricks is arriving faster than the agents themselves.

• databricks-solutions/ai-dev-kit · project
• Trace any AI agent with OTel, MLflow, and Unity Catalog · video

Data Engineering

Lakehouse Federation war stories

Two from the trenches: a silent failure where Federation "loses" tables whose names contain a space, and a practitioner who ripped out a Postgres to Kafka to DMS to S3 pipeline and replaced the whole thing with Lakeflow. The kind of detail you only get from people running it in anger.

• When Lakehouse Federation "loses" your tables: the space-in-name trap · community
• Replacing a Postgres → Kafka → DMS → S3 pipeline with Lakeflow · community

One feed, reordered for your role

Stop tab-hopping. /feed merges every stream into one, ranked by what you actually work on.

Databricks ships across a dozen surfaces a week. Keeping up used to mean bouncing between our News, Releases, Videos, Projects, and Community pages and stitching the picture together yourself. The new /feed does the stitching. No account needed: pick your role and all five streams merge into a single feed, reordered so the topics you work on float to the top.

A data engineer sees Lakeflow and Delta items first. A platform lead sees Unity Catalog and governance. Pick one of six roles in a click. Or sign in to describe a custom role in your own words, and we map it to the topics that fit. Narrow it further by stream, time window, or topic. It's the antidote to the same information overload the IPO market is drowning in, just pointed at your week instead of Wall Street's.

Reorder for your role

✓ Data Engineer  ML / Data Scientist  Platform / Infra  BI  Custom role

🤖 Got a Databricks question?

The Brickster Assistant searches our full archive and answers with citations.

Enjoying Brickster Weekly?

Forward this to a colleague who works with Databricks.