Turning down a 67 billion offer, the most "Versailles" financing deal is born.

Another trillion-level unicorn has emerged in the AI circle.

Recently, AI big data company Databricks announced the completion of $5 billion strategic financing (equivalent to about 336.89 billion yuan), and the company's latest valuation reached $1900 billion (equivalent to about 1.3 trillion yuan).

This is a rather crowded round of financing. More than 20 institutions participated in this round, led by existing shareholder Coatue, with Blackstone, Abu Dhabi AI investment platform MGX, T. Rowe Price Associates, Sixth Street Growth and other institutions participating. Among them, Sixth Street Growth is a new investor, founded by Alan Waxman, former Chief Investment Officer of Goldman Sachs.

This is also the second round of financing officially completed by Databricks in the past year. The company's current valuation of $1900 billion has increased by about 42% compared with the $1340 billion valuation at the time of financing at the end of 2025.

After this round of financing, Databricks has also ranked among the world's highest-valued unlisted technology companies, second only to a few enterprises such as OpenAI, Anthropic and ByteDance.

Interestingly, judging only from business, Databricks is the most "unattractive" one among this group of star companies — OpenAI and Anthropic are betting on large models, ByteDance has an Internet application ecosystem, while the data cloud service that Databricks focuses on is slightly boring, which mainly helps enterprises manage, process and access their own business data.

But in the AI era, who dares to underestimate the value of data infrastructure providers?

"We Could Have Raised 100 Billion Dollars"

In fact, it is not accurate to define Databricks as an AI big data company. Or to say, it is not accurate to define Databricks with any familiar "digital industry" business form we know.

Databricks originated from a scientific research project at the University of California, Berkeley, which tried to solve the problem of large-scale data processing efficiency. In 2013, in order to realize commercialization, the Spark R&D team co-founded Databricks — at this stage, Databricks was indeed a standard data service provider in the true sense.

But 13 years later, in Databricks' current self-positioning, the "Data" in its name is no longer the protagonist. In the description on its official website, it says its service target is "DATA+AI", and the product is no longer a traditional platform, but "Data Lakehouse".

This is very abstract, but judging from the financing rhythm, it is indeed very in line with the aesthetic of the current era. And while announcing the completion of the new round of financing, Databricks also released a somewhat "pretentious" detail to the outside world: actually we didn't plan to raise so much money at the beginning.

According to the previous news disclosed by Ali Ghodsi, co-founder and CEO of Databricks, the company originally only planned to raise 1 billion US dollars. But after media reports in June that Databricks was carrying out large-scale financing, investors eager to reach the deal quickly flooded Ghodsi's phone. Among them, the total amount of investment intentions expressed by the investors that Databricks had inspected reached 150 billion US dollars (equivalent to about 1.01 trillion yuan).

Unable to reject such generous offers, Databricks had to expand the scale of financing. In July, the company announced that it was carrying out a new round of financing. At that time, Databricks disclosed that its valuation reached 1.88 trillion US dollars, but did not announce the specific amount of financing.

Less than a month later, the answer was revealed: Databricks announced that it had raised 50 billion US dollars, and the company's valuation was further increased to 1900 billion US dollars. In other words, Databricks not only raised 4 billion US dollars more than the original plan, but also rejected 100 billion US dollars of hot money (equivalent to about 673.76 billion yuan).

Of course, this is not the first time Databricks has arrogantly rejected investors.

In the 13 years since its establishment, Databricks has completed 13 rounds of financing, almost maintaining the rhythm of one round of financing a year. Frequent financing has also led to netizens' teasing: Databricks has raised so much capital that the alphabet list is almost used up.

Since 2025 alone, Databricks has completed 4 rounds of equity financing, with cumulative financing exceeding 200 billion US dollars. Among them, the $10 billion J-round financing completed in January 2025 once set a record for the largest financing scale of private technology companies. It was also in this round of financing that Databricks rejected a huge amount of intended investment. According to Ghodsi's disclosure, the total capital that all intended investors were willing to provide at that time had reached 190 billion US dollars (equivalent to about 1.324 trillion yuan).

The capital lineup behind Databricks is also constantly changing with the development of the company.

In the early stage, Databricks' investors were mainly technology VCs such as a16z, NEA and DCVC.

As the company's business scale expands, Databricks has begun to attract industrial capital such as Microsoft, AWS, CapitalG under Google, Salesforce Ventures, NVIDIA, and Meta to participate in the investment. The business territory of investors has further extended from cloud computing and enterprise software to AI infrastructure and AI application ecosystem.

At the same time, global asset management institutions such as BlackRock, Fidelity and T. Rowe Price, as well as investors with national capital backgrounds such as GIC, Temasek and MGX, have also joined Databricks' shareholder list.

"Infrastructure Provider" in the GPT Era

If you want to know why Databricks can be so arrogant, you still have to go back to the somewhat abstract "Data+AI" business mentioned earlier.

In the era of large AI models, whether data is rich, accurate, timely and can be effectively called largely determines whether AI can be transformed from technical capabilities into commercial value. And Databricks is a company that provides data infrastructure services.

In the early days of its establishment, Databricks mainly provided users with a cloud-based data processing platform based on Spark, so as to simplify the large-scale data processing process. In the following ten years, the company's business gradually expanded from big data processing to scenarios such as data analysis and machine learning, and built a "data lakehouse" architecture.

Enterprises will generate a large amount of business data in their daily operations. These data need to be collected, sorted and stored, and also need to be further used for analysis, real-time processing and machine learning. In the past, different data processing needs were often carried by different systems, and enterprises needed to transfer data between multiple systems and bear the corresponding maintenance and management costs. Databricks' Lakehouse architecture attempts to integrate these requirements into a unified data processing platform.

If the business only stays at this level, Databricks may at most become an excellent enterprise software company. But after 2023, Databricks began to redefine its position.

After GPT set off the boom of large models, enterprises' enthusiasm for embracing AI rose rapidly. But for most enterprises, the real implementation of AI is not as simple as accessing a large model. How to activate the data accumulated over the years within the enterprise and let AI truly understand the enterprise's business is a more realistic problem. This has also become a new opportunity for Databricks.

Based on the capabilities accumulated around enterprise data for many years, Databricks began to further extend its business to enterprise AI scenarios. In 2023, the company acquired MosaicML for about $1.3 billion. At that time, MosaicML was already a highly concerned startup in the generative AI field, with its core capability of helping enterprises train and deploy large models using their own data. After that, Databricks further supplemented its capabilities in data governance, data access, AI data processing and databases through acquisitions of companies such as Okera, Arcion, Lilac, Tabular and Neon.

Through a series of business supplements, Databricks has gradually become an important data infrastructure platform in the process of enterprise AI implementation, and finally formed the abstract "data lakehouse". On the official website, Databricks explains the meaning of "data lakehouse" as follows:

A data lakehouse built on open source and open standards simplifies your data assets by eliminating the silos that historically complicated data and AI.

At present, Databricks serves more than 20,000 institutions around the world, and large enterprises such as Adidas, Bayer, Mastercard and Unilever are all on the customer list. Among them, more than 1000 customers have an annualized consumption scale of more than 1 million US dollars on Databricks, and more than 100 customers have an annualized consumption scale of more than 10 million US dollars.

The strategy of taking high-profile actions in AI-related businesses has indeed raised the valuation ceiling of Databricks.

The most intuitive example is that at the node of 2026, perhaps fewer and fewer people will compare Databricks with Snowflake.

Snowflake, founded in 2012, is an enterprise data service provider that was sought after by capital at the same time as Databricks. In September 2020, Snowflake took the lead in listing on the New York Stock Exchange, and the valuation corresponding to the IPO issue price was about 33.5 billion US dollars. In fiscal year 2021 (as of January 31, 2021), Snowflake's revenue was about 590 million US dollars. Databricks' annual recurring revenue in 2020 was about 425 million US dollars, and its valuation after completing a new round of financing at the end of 2019 was about 6.2 billion US dollars, which was only about one fifth of Snowflake's IPO valuation.

After the GPT era, the two companies ushered in different development paths.

According to the latest revenue data disclosed by Databricks recently, the company's annualized revenue reached about 7 billion US dollars, a year-on-year increase of more than 80%. In the past 12 months, the company has continued to achieve positive adjusted free cash flow. The annualized revenue of the company's core product "Lakehouse" exceeds 1.5 billion US dollars, a year-on-year increase of more than 100%. In contrast, Snowflake's revenue in fiscal year 2026 was 4.47 billion US dollars, with a year-on-year growth rate of 29%, and its current market value is about 113 billion US dollars.

Databricks has already surpassed Snowflake in terms of revenue scale. The valuation of 1900 billion US dollars also means that the capital market is giving Databricks higher growth expectations based on its position in the AI ecosystem.

The wealth creation story of AI is still going on, and the next batch of Databricks may already be on the way.

This article is from WeChat official account "Dongsi Shitiao Capital" (ID: DsstCapital), written by Li Xinting, and published with authorization from 36Kr.