Databricks Went From a Berkeley Lab to $190 Billion. Is It Worth It?

Good morning, Mason here,

Today's issue is another deep dive. One company, start to finish.

In 2013, a group of Berkeley researchers asked Ben Horowitz for about $200,000 to keep working on their open-source project.

He offered them $14 million instead.

Thirteen years later, Databricks is worth $190 billion, has passed a $7 billion revenue run rate and is growing more than 80 percent a year.

It's worth more than Snowflake. And it still hasn't gone public.

So is any of this worth the price?

Read time: 8 mins

SPONSORED OF THE DAY

DIT AgTech Doesn’t Need to Sell Hardware to Make Money.

One installation can generate years of recurring supplement revenue. 

provides the hardware free to ranchers through a three-year nutrition plan. The hardware gets installed once. The supplements keep generating revenue year after year. 

Then there’s the second paycheck. 

Methane reduction is becoming an economic opportunity for agriculture. 

Methane-reducing nutrition create carbon credits, giving ranchers another income stream while creating an additional revenue opportunity for . 

That means one connected livestock operation generates recurring supplement revenue, with carbon revenue as additional upside. 

We win when ranchers win. 

The model is already working in Australia, one of the world's toughest livestock environments, with 500+ units deployed and 370,000 head on the platform. 

Now DIT AgTech brings that proven model to the massive U.S. livestock market. 

For investors, the opportunity is getting in before the model scales into its next market.

^{𝘐𝘯 𝘮𝘢𝘬𝘪𝘯𝘨 𝘢𝘯 𝘪𝘯𝘷𝘦𝘴𝘵𝘮𝘦𝘯𝘵 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯, 𝘪𝘯𝘷𝘦𝘴𝘵𝘰𝘳𝘴 𝘮𝘶𝘴𝘵 𝘳𝘦𝘭𝘺 𝘰𝘯 𝘵𝘩𝘦𝘪𝘳 𝘰𝘸𝘯 𝘦𝘹𝘢𝘮𝘪𝘯𝘢𝘵𝘪𝘰𝘯 𝘰𝘧 𝘵𝘩𝘦 𝘪𝘴𝘴𝘶𝘦𝘳 𝘢𝘯𝘥 𝘵𝘩𝘦 𝘵𝘦𝘳𝘮𝘴 𝘰𝘧 𝘵𝘩𝘦 𝘰𝘧𝘧𝘦𝘳𝘪𝘯𝘨, 𝘪𝘯𝘤𝘭𝘶𝘥𝘪𝘯𝘨 𝘵𝘩𝘦 𝘮𝘦𝘳𝘪𝘵𝘴 𝘢𝘯𝘥 𝘳𝘪𝘴𝘬𝘴 𝘪𝘯𝘷𝘰𝘭𝘷𝘦𝘥. 𝘋𝘐𝘛 𝘈𝘨𝘛𝘦𝘤𝘩 𝘩𝘢𝘴 𝘧𝘪𝘭𝘦𝘥 𝘢 𝘍𝘰𝘳𝘮 𝘊 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘚𝘦𝘤𝘶𝘳𝘪𝘵𝘪𝘦𝘴 𝘢𝘯𝘥 𝘌𝘹𝘤𝘩𝘢𝘯𝘨𝘦 𝘊𝘰𝘮𝘮𝘪𝘴𝘴𝘪𝘰𝘯 𝘪𝘯 𝘤𝘰𝘯𝘯𝘦𝘤𝘵𝘪𝘰𝘯 𝘸𝘪𝘵𝘩 𝘪𝘵𝘴 𝘰𝘧𝘧𝘦𝘳𝘪𝘯𝘨, 𝘢 𝘤𝘰𝘱𝘺 𝘰𝘧 𝘸𝘩𝘪𝘤𝘩 𝘮𝘢𝘺 𝘣𝘦 𝘰𝘣𝘵𝘢𝘪𝘯𝘦𝘥 𝘩𝘦𝘳𝘦: https://bit.ly/4bzuWCi​}

THE DEEP DIVE

Databricks co-founder and CTO Matei Zaharia, who started Spark as a PhD student. Photo: DKP415 via Wikimedia Commons (CC BY-SA 4.0)

The story starts with Matei Zaharia. Around 2009, as a PhD student in UC Berkeley's AMPLab, he started building Spark, a much faster way to crunch huge piles of data than Hadoop, which was the standard back then. It went open source in 2010 and caught on fast.

In 2013, Zaharia and a handful of others from the lab, including professor Ion Stoica and Ali Ghodsi, turned it into a company. They weren't thinking big. Ghodsi has said they asked Andreessen Horowitz for about $200,000 to code for a year. Ben Horowitz came back with $14 million. Fortune reported that first round valued Databricks at about $47 million.

Ghodsi's own path was unlikely. Born in Iran in 1978, he was still a little kid when his family escaped to Sweden. There he figured out coding on a Commodore 64. Later came a PhD in Stockholm, then a visiting spot at Berkeley in 2009. Stoica ran the company first. Ghodsi took over as CEO in January 2016.

Here's the catch with Spark, though: it was free. Anybody could download it. So Databricks had to sell something people couldn't just get for nothing, and that problem shaped everything it did after.

Zaharia open-sources Unity Catalog on stage at the 2024 Data + AI Summit. Photo: MockingBird45 via Wikimedia Commons (CC BY-SA 4.0)

Version one wasn't fancy. You'd hand Databricks your data, it would run Spark on it in the cloud, and you'd skip hiring people to keep the thing alive. The big break came in late 2017, when Microsoft announced Azure Databricks. Once it launched the next spring, Microsoft's sales team was selling Databricks too.

From there it just kept adding layers. Delta Lake cleaned up the mess inside data lakes. Around 2020 came the "lakehouse" pitch: why keep a data lake and a data warehouse when one system could do both? Databricks SQL followed, aimed right at the warehouse business. Yes, Snowflake's business.

AI came next. Databricks bought MosaicML for about $1.3 billion in 2023, put out its own open model, DBRX, in 2024, and now sells tools like Agent Bricks for building AI agents on top of company data. In 2025 it paid about $1 billion for Neon, a database startup that became its Lakebase product.

The bet under all of it is that AI models are turning into a commodity, but every company's own data isn't. Whoever holds that data gets paid every time a company runs AI on it. As Ghodsi put it in August, "Enterprises don't just want AI that talks."

A Databricks booth at a 2023 tech conference. Photo: Raysonho via Wikimedia Commons (CC0)

Databricks mostly doesn't sell licenses. Customers pay for how much computing they use, so revenue grows on its own as companies push more data and more AI through the platform.

It also gives a lot away. Spark, Delta Lake, MLflow and Unity Catalog are all open source. That pulls developers in for free, and plenty of them end up at companies that pay.

And it's generating cash now. Databricks says it has produced positive adjusted free cash flow over the last 12 months. At this size, growing 80 percent a year, that's rare.

The weak spot is cost. Databricks runs on Amazon, Microsoft and Google's clouds, and AI work is expensive to serve. Sacra put its gross margin at about 80 percent in mid-2024, and the AI products are adding more pressure on top of that.

Ghodsi gives the keynote at the 2026 Data + AI Summit. Photo: MockingBird45 via Wikimedia Commons (CC BY-SA 4.0)

Over 20,000 organizations are on Databricks now. That includes 70 percent of the Fortune 500. Some of them spend a lot: north of 1,000 customers are on pace to pay it more than $1 million a year, and over 100 are past $10 million.

Those customers keep spending more. Databricks' net revenue retention was reported above 140 percent in fiscal 2024, meaning existing customers spent about 40 percent more than the year before. Snowflake's latest number is 126 percent.

The newer products are pulling their weight too. Its data warehouse business passed a $1.5 billion run rate and is growing more than 100 percent a year. Lakebase passed $100 million. And Fox signed on as a multiyear customer this month.

The fine print: these are all company-reported numbers. Databricks is private, so no auditor has signed off on them the way one would for a public company.

Databricks co-founder and CEO Ali Ghodsi. Photo: DKP415 via Wikimedia Commons (CC BY-SA 4.0)

The valuation chart is kind of absurd. About $47 million in 2013. $2.75 billion in early 2019, when it passed $100 million in recurring revenue. $28 billion in 2021. $62 billion in December 2024 on a $10 billion round. More than $100 billion in August 2025, $134 billion that December, and $190 billion this August after a $5 billion round led by Coatue.

Revenue kept up. Databricks did about $1.6 billion in the fiscal year that ended in January 2024. It hit a $4.8 billion run rate at the end of 2025 and passed $7 billion this summer, with growth speeding up from about 55 percent to more than 80 percent.

It's borrowing too. In January it reportedly added $1.8 billion in debt, taking its total borrowing capacity to about $7 billion.

What it hasn't done is go public. In June, Ghodsi told Bloomberg TV, "I just think this is a terrible year to go public," with SpaceX, Anthropic and OpenAI all chasing IPO money. So that $190 billion price is set by a small group of private investors, not the market.

A Snowflake booth at a 2023 tech conference. Photo: Raysonho via Wikimedia Commons (CC0)

For years the rivalry was Databricks versus Snowflake. Snowflake went public in 2020 and was the bigger name. Not anymore.

Snowflake expects about $6.1 billion in product revenue this fiscal year, up 36 percent, and it's worth roughly $115 billion. Databricks has a bit more revenue on a run-rate basis, more than twice the growth, and a valuation about 65 percent higher.

It got personal sometimes. In June 2024, Databricks announced it was buying Tabular, a startup Snowflake had also reportedly bid for, right in the middle of Snowflake's own conference. Ghodsi said it paid over $1 billion. Bloomberg later reported close to $2 billion.

The bigger threat might be its own partners. Microsoft, Amazon and Google all host Databricks, and all three sell competing data products. Same goes for OpenAI and Anthropic. Each signed a $100 million deal with Databricks, and each would happily take its enterprise customers if it could.

PREMIUM

📰Join Founderscrowd Premium

Join Founderscrowd Premium to get:

• Unique investment opportunities

• Our detailed valuation analysis (why this entry point makes sense)

• The specific risks to watch before IPO

• How to structure your position ($25K minimum)

• Real-time alerts when secondary tranches open

• Step-by-step process to invest

Premium members also get first dibs on allocation — the difference between getting in and getting locked out.

$100/month. Cancel anytime. No commitment.

This is where founders and serious investors actually allocate capital.

Join Premium

THE VERDICT: IS $190 BILLION TOO MUCH?

Six parts. It comes down to two arguments.

The bull case is strong. In two years revenue has almost tripled. Growth got faster, not slower. Cash is coming in. And Databricks is already inside most big companies at a moment when AI is making their data worth more.

The bear case is real too. The price is set privately. Margins are under pressure from AI costs. The numbers aren't audited. And its biggest suppliers are also some of its biggest competitors.

The math. At $190 billion, investors are paying about 27 times Databricks' current revenue run rate. Snowflake trades at around 19 times the product revenue it expects this year. So Databricks costs about 40 percent more per dollar of revenue, for more than twice the growth.

The test: say growth cools to 50 percent next year, then 40, 30 and 25. That gets Databricks to about $24 billion in revenue by 2030. Give it 8 times sales, roughly what a mature software company gets, and you're at about $190 billion. Put differently, today's price assumes four more years of strong execution and then a grown-up multiple.

For investors: Don't just look at the growth rate.

Look at how much of it the price already assumes.

Mason's take: Compared to most of the private giants I write about, this one's easier to defend. Databricks generates cash, grows faster than its public rival and is already everywhere. I think $190 billion is a fair price if it keeps executing. The risk is that nobody gets a real public check on these numbers until the IPO, and by then most of the upside could already be gone.

Founderscrowd Newsletter | Friday, October 23, 2026 | The Databricks deep dive

$47M in 2013. $190B in 2026. An IPO in 2027?