DevNews

Databricks Raises $5B at $190B, Aiming It All at Agents

On this page
  1. The round
  2. What the money is for
  3. Why an agent platform ships a database
  4. Reading the valuation honestly
  5. Sources and further reading

Databricks closed a $5 billion round at a $190 billion valuation on August 13, 2026, led by Coatue, six months after a round that valued it at $134 billion. The numbers behind it are the interesting part: a revenue run rate above $7 billion growing more than 80 percent year over year, more than 1,000 customers past $1 million in annual recurring revenue, and positive adjusted free cash flow over the trailing twelve months. The money goes to three products, and all three answer the same question. Databricks is betting that the hard part of enterprise AI is no longer the model, it is the data plumbing and the bill.

The short answer

Databricks closed a $5 billion round at a $190 billion valuation on August 13, 2026, led by Coatue with Blackstone, MGX, T. Rowe Price and new investor Sixth Street Growth. The company reports a revenue run rate above $7 billion, growing more than 80 percent year over year, with positive adjusted free cash flow over the trailing twelve months. Funds go to Lakebase, a serverless Postgres past a $100 million run rate, Genie, and Unity AI Gateway. The previous round, in February, valued the company at $134 billion.

$5Braised, at a $190 billion valuation
$7B+revenue run rate, up more than 80% year over year
1,000+customers above $1 million in annual recurring revenue
Answer card summarising the Databricks round of August 13, 2026: $5 billion raised at a $190 billion valuation led by Coatue, a revenue run rate above $7 billion growing more than 80 percent year over year, and more than 1,000 customers past $1 million in annual recurring revenue.
The round in one card. Source: the Databricks press release of August 13, 2026 and reporting from CNBC and Bloomberg the same day. PNG

Funding announcements mostly tell you what a company wants to be believed about it. The useful signal is further down, in which products the money is attached to.

The round

On August 13, 2026, Databricks closed a $5 billion round at a $190 billion valuation. Coatue led, with Blackstone, MGX, T. Rowe Price and new investor Sixth Street Growth, plus a long roster of returning backers that includes Andreessen Horowitz, Thrive Capital, Goldman Sachs Alternatives and Temasek.

Six months ago the same company was valued at $134 billion. That is a 42 percent step up in half a year, and the operating numbers published alongside it are what makes the step defensible or not. Revenue run rate is above $7 billion, growing more than 80 percent year over year in the second quarter. Adjusted free cash flow is positive on a trailing twelve month basis. More than 1,000 customers are above $1 million in annual recurring revenue and more than 100 are above $10 million.

That last pair is the one we would read first. Growth rates can be bought; a thousand accounts each spending seven figures is harder to manufacture and tells you the product is embedded rather than trialled.

Comparison chart of Databricks figures published on August 13, 2026: a $190 billion valuation, $5 billion raised, a revenue run rate above $7 billion, Lakehouse annualised revenue above $1.5 billion and Lakebase run rate above $100 million.
The published figures, in billions of dollars where applicable. Source: the Databricks announcement of August 13, 2026. PNG

What the money is for

Three products, and they are named specifically rather than gestured at.

Lakebase is a serverless Postgres database built for agent workloads, already past a $100 million revenue run rate. Genie is an assistant that answers questions against a company's own business data. Unity AI Gateway governs which models are used across an organisation and controls what they cost.

Chief executive Ali Ghodsi's framing is that enterprises do not just want AI that talks, they want agents working across their business that remember context, deliver accurate answers and execute work without blowing through budgets. Set that sentence next to the product list and it lines up exactly: memory is Lakebase, accurate answers on business data is Genie, and not blowing through budgets is Unity AI Gateway.

Meanwhile the older core keeps growing. The Lakehouse warehousing business is past $1.5 billion in annualised revenue with growth above 100 percent year over year, which is an unusual number for a mature product line at that size.

Why an agent platform ships a database

The Lakebase piece is the one worth pausing on, because at first glance a data warehouse company shipping a Postgres looks like a distraction.

It is not, and the reason is access patterns. A warehouse is built for large scans over columnar storage, run occasionally, with seconds of latency being fine. An agent working through a multi step task does the opposite: many small transactional reads and writes, state carried between steps, at whatever rate its loop runs rather than at human speed. That is an OLTP workload, and the honest answer to an OLTP workload is Postgres. Serverless matters because agent load is spiky in a way that provisioned capacity handles badly.

The argument for putting it inside the same platform rather than beside it is that agent state and analytical data otherwise live in two systems with a sync between them, and that sync is where correctness goes to die.

The cost control product points at the same pressure from the other side. Nobody built a spend governor for models in 2024 because nobody needed one. They do now, in a market where a model vendor can double its cache pricing overnight and another can put a four month expiry on its list price. Model costs became a line item that moves without warning, and a platform that abstracts which model runs where is selling insurance against exactly that.

Reading the valuation honestly

A $190 billion valuation on a $7 billion run rate is roughly 27 times revenue. That number only works if the growth rate holds, so the 80 percent is doing most of the load bearing, and 80 percent at this scale is genuinely hard.

The figures we would track over the next year are the boring ones. Positive adjusted free cash flow means the business is not dependent on this round to operate, which is a different position from most of the companies raising large rounds this year. Customer concentration at the $10 million level tells you how much of that run rate rests on a small number of relationships. Those two move slowly and are hard to dress up, which is precisely why they are worth watching.

Sources and further reading

Frequently asked questions

What are the actual numbers in this round?

Databricks raised $5 billion at a $190 billion valuation, announced on August 13, 2026 and led by Coatue with Blackstone, MGX, T. Rowe Price and new investor Sixth Street Growth, alongside a long list of returning backers including Andreessen Horowitz, Thrive Capital, Goldman Sachs and Temasek. The company reports a revenue run rate above $7 billion, growing more than 80 percent year over year in the second quarter, with positive adjusted free cash flow on a trailing twelve month basis. The previous round, six months earlier, valued it at $134 billion, so this is a 42 percent step up in half a year.

Where is the money actually going?

Into three named products, all of them aimed at running agents against enterprise data. Lakebase is a serverless Postgres database built for agent workloads, already past a $100 million revenue run rate. Genie is an assistant that answers questions against a company's own business data. Unity AI Gateway governs which models get used and controls the resulting spend. The Lakehouse warehousing business, the older core, is past $1.5 billion in annualised revenue and growing more than 100 percent year over year, which is unusual for a product at that scale.

Why does an agent platform need its own Postgres?

Because an agent's access pattern does not look like an analyst's. A warehouse is built for large scans over columnar data, run occasionally, with latency measured in seconds. An agent doing multi step work issues many small transactional reads and writes, needs to hold state between steps, and does it at a rate set by a loop rather than by a human. That is a transactional workload, which is what Postgres is for, and serverless matters because the load is spiky and unpredictable. The pitch is keeping that state next to the analytical data rather than syncing between two systems.

What does Ali Ghodsi say the problem is?

His framing in the announcement is that enterprises do not just want AI that talks, they want agents working across the business that remember context, deliver accurate answers and execute work without blowing through budgets. That sentence maps one to one onto the three products: memory and context is Lakebase, accurate answers on business data is Genie, and not blowing through budgets is Unity AI Gateway. Whether or not you buy the platform, the diagnosis is worth taking seriously, because the cost control clause is the one that shows up on invoices rather than in demos.

How should we read a valuation like this?

Carefully, and with the operating numbers in front of the headline. A $190 billion valuation on a $7 billion run rate is roughly 27 times revenue, which only makes sense if the growth rate holds, and 80 percent year over year at this scale is the claim doing the work. The figures worth watching are the ones that are harder to engineer: positive adjusted free cash flow, more than 1,000 customers above $1 million in annual recurring revenue and more than 100 above $10 million. Concentration and cash generation tell you more about durability than the valuation multiple does.