Databricks is raising strategic capital at a $188 billion valuation.
The company said it has signed a term sheet for a new funding round led by existing investor Coatue, with additional new and existing investors expected to participate. Databricks expects the round to close later this summer.
The new capital will be used to accelerate the company’s AI strategy. The products Databricks specifically highlighted are Unity AI Gateway, Genie and Lakebase.
At first glance, this looks like another large AI financing story.
But the real significance is not simply that Databricks has become more valuable.
The more important question is how Databricks is redefining itself.
Databricks is no longer just a data analytics company.
It is no longer enough to describe it only as a lakehouse company.
The position Databricks is trying to occupy is clearer and more ambitious:
It wants to become the control layer for enterprise AI.
While OpenAI, Anthropic, Google, Meta, xAI and Mistral compete over foundation models, Databricks is aiming at the layer where those models are actually used inside companies: data access, permissions, cost control, security, business context and operational governance.
The question in AI is changing.
Which model is the smartest?
That question still matters.
But for enterprises, it is no longer enough.
Companies now ask different questions.
Where is our data?
Which data did the AI access?
Which model should we use for which workflow?
Who controls token spending?
Is the answer grounded in our business data?
Can an AI agent act safely on top of our databases and applications?
Databricks is moving into the center of those questions.
The Bottleneck in Enterprise AI Is Not Model Intelligence. It Is Context.
Databricks describes the problem facing enterprise AI as a “context gap.”
That phrase matters.
It means companies are not struggling only because models are insufficiently intelligent. Increasingly, the problem is that enterprise data is fragmented, disconnected from AI systems and difficult to govern.
Many companies have already experimented with generative AI.
They have used it for meeting summaries, document drafting, code assistance, customer support, internal search and report generation. But uploading a few files into a chatbot is very different from connecting AI reliably to the full operating reality of a company.
Enterprise data is complicated.
It lives in ERP systems, CRM systems, data warehouses, cloud storage, spreadsheets, Slack, email, departmental tools and legacy databases. Different teams own it. Different permissions apply to it. Some data is current. Some data is stale. Some is customer information. Some is financial information. Some is trade secret.
AI models are powerful, but they do not automatically understand that context.
Without context, AI gives generic answers.
Without permissions, AI cannot access the right data.
Without controls, AI may misuse sensitive information.
Without cost governance, AI spending can grow quickly.
Without trust, business teams cannot rely on AI answers for decisions.
Databricks’ “context gap” is therefore the practical wall that many enterprise AI programs hit.
The success of enterprise AI will not be decided by model calls alone.
It will be decided by data connection.
From Tokenmaxxing to Valuemaxxing
Databricks co-founder and CEO Ali Ghodsi captured the shift in one phrase.
“Enterprises are moving from tokenmaxxing to valuemaxxing,” he said. “They don’t want to burn expensive tokens on the smartest model for every task — they want the best outcome per dollar.”
This is one of the clearest descriptions of the next phase of enterprise AI.
In the early generative AI wave, many companies treated access to the strongest model as the main goal. Larger models, better benchmarks, longer context windows and more advanced reasoning dominated the conversation.
But once enterprises begin looking at actual bills, the conversation changes.
Not every task needs the most powerful model.
Simple classification may require only a smaller model.
Document summarization may be handled by a mid-tier model.
Complex legal, financial or scientific analysis may need a stronger model.
Internal data search may depend more on retrieval, permissions and citations than on raw model size.
Workflow automation may require reliability and predictable cost more than maximum intelligence.
Enterprises will eventually operate model portfolios.
They will not use one model for everything. They will need to choose the right model for the right task, compare performance and cost, control access, log usage and verify reliability.
This is where Unity AI Gateway becomes strategically important.
An AI gateway is not merely a connection tool.
In a multi-model enterprise environment, it becomes the gatekeeper for who called which model, with what data, under which policy and at what cost.
As AI adoption spreads, companies want model flexibility.
But the more freedom they have, the more governance they need.
Databricks is building for that paradox.
Databricks Wants to Become a Multi-AI Control Company
The three products emphasized in the funding announcement point to a coherent strategy.
Unity AI Gateway is about multi-AI governance. It helps enterprises manage AI access, usage and cost across multiple models and agents.
Genie is Databricks’ AI coworker for business data. It is designed to turn enterprise data into trusted answers and actions, rather than generic chatbot responses.
Lakebase is a serverless Postgres database built for AI agents. It targets a future in which agents do more than generate text; they read, write and act on application data.
Put these products together and the strategy becomes visible.
First, Databricks wants to control the enterprise data foundation.
Second, it wants to govern how AI accesses that data.
Third, it wants to provide the interface through which business users ask questions and take action.
Fourth, it wants to extend into the database layer where AI agents execute work.
In other words, Databricks is not trying to win primarily by owning the best general-purpose model.
It is trying to own the environment in which models can safely work inside companies.
That is a realistic strategy.
Not every enterprise needs to build its own frontier large language model.
But every enterprise needs its data to be usable by AI.
And every enterprise needs to control that usage.
Models may change.
Enterprise data and governance layers are much stickier.
That sticky layer is what Databricks is pursuing.
Why $188 Billion?
A $188 billion valuation is enormous.
It suggests that investors are not valuing Databricks merely as a traditional data-platform company. They are valuing it as a potential core infrastructure company for enterprise AI.
For the past few years, AI market attention has been focused heavily on model companies.
OpenAI, Anthropic, xAI, Google DeepMind and Meta have absorbed enormous attention and capital. But from the perspective of enterprise customers, models alone are not enough.
Companies do not buy models.
They buy outcomes.
Cost reduction.
Revenue growth.
Workflow automation.
Better customer service.
Faster decision-making.
Risk management.
More effective use of data.
To produce those outcomes, models must connect to enterprise data. That connection must be secure. Access must be governed. Answers must be traceable. Costs must be controlled.
This is where Databricks’ value grows.
Databricks already sits inside many enterprise data and analytics workflows. The company says more than 20,000 organizations use its platform, including 70% of the Fortune 500. Its homepage also says more than 60% of the Fortune 500 uses Databricks, showing how deeply it has embedded itself in large-company data environments.
Enterprise AI will ultimately run on enterprise data.
The place where that data lives becomes a strategic high ground.
The Snowflake Battle Will Intensify
Databricks’ strategy inevitably points to a deeper competition with Snowflake.
Both companies began from the data platform world. Both now want to become the operating layer for enterprise data and AI. Their technical histories and architectures differ, but the direction is increasingly similar: data, governance, AI applications and agents on top of a managed platform.
In the past, the competition was often framed as data warehouse versus lakehouse.
That frame is now too narrow.
The real questions are changing.
Who can connect enterprise data to AI more safely?
Who can let business users ask questions and get trusted answers more easily?
Who can provide the infrastructure for AI agents to execute work?
Who can manage cost and policy across multiple models?
Who can prove AI return on investment for enterprise customers?
Databricks’ focus on Unity AI Gateway, Genie and Lakebase shows where this competition is moving.
Data platforms no longer want only to store and analyze data.
They want to become the layer where AI understands data, answers questions and performs work.
The Databricks-Snowflake competition is evolving from data infrastructure into an enterprise AI operating-system battle.
In the Agent Era, Databases Matter Again
Lakebase is important because AI agents need more than text generation.
If an agent is going to perform real work, it must read data, store state, update records, interact with applications and continue workflows. For that, it needs a database.
An AI agent does not stop at writing a report.
It may need to check customer data, update an order status, record a decision, trigger a workflow or preserve the state of a task over time.
That requires reliable data infrastructure.
Lakebase targets this layer.
By offering a serverless Postgres database built for AI agents, Databricks is moving beyond storage and analytics into the execution layer of AI applications.
That is a significant shift.
As AI agents multiply, databases become more than back-end systems. They become the memory and ledger of agentic work.
What task is underway?
What data changed?
What decision was made?
What follow-up action is required?
If the model is the brain, the database is memory and the work record.
Databricks wants that record to live inside its platform.
The Enterprise AI Winner May Be the Company That Chooses the Right Model
Another important phrase from Ghodsi is “the right AI for the job.”
Enterprises will not use one model for everything.
They will use multiple models at the same time.
One workflow may use OpenAI. Another may use Anthropic. Another may use an open-source model or an internal model. High-security work may run in a controlled environment, while general tasks may use external APIs.
That creates a need for choice.
But choice creates complexity.
Which model can access which data?
Which model is most cost-effective for a particular workflow?
Which model produces the most reliable answer?
Which model complies with internal policy and regulation?
How should model spending be tracked?
A platform that solves these questions can become central to enterprise AI.
Databricks is positioning itself as a multi-AI platform rather than a company tied to one model provider. That is attractive to enterprise customers.
Companies do not want to be fully locked into a single model vendor. Prices, performance, security requirements, regulatory rules and geopolitical risks can all change.
Enterprise customers want a model-neutral control layer.
Databricks wants to occupy that layer.
The Lesson for Korean Companies: Data Control Comes Before AI Adoption
This announcement has direct implications for Korean companies.
Many organizations talk about AI adoption. They use generative AI, build internal chatbots and experiment with meeting summaries and automated reports. But many do not see meaningful business results.
The reason is not always that the model is weak.
Often, the data is not ready.
Departmental data is scattered. Permissions are unclear. Data quality is uneven. Some files are outdated. Official records and personal spreadsheets are mixed together. Security teams worry about external model use. Business teams do not fully trust AI answers. Executives ask about ROI, but token cost and business value are difficult to connect.
In that environment, simply adding another AI model will not solve the problem.
Korean companies need to start with data control.
Where is the company’s data?
Who can access it?
Which data can AI use?
Which model should be used for each task?
How should AI cost and business impact be measured together?
How can the company trace the basis and responsibility of AI-generated answers?
The “context gap” Databricks describes applies directly to Korean enterprises.
The bottleneck in AI transformation is not only the model.
It is data operations.
AI Governance Is Not Only a Compliance Issue. It Is a Cost Issue.
When companies hear the phrase AI governance, they often think first of regulation.
Privacy, security, compliance and audit readiness are all important.
But Databricks’ framing adds another point: governance is also cost control.
As AI usage increases, token spending can grow quickly. If every team uses high-end models for every task, costs become difficult to manage. If departments adopt different tools independently, duplication grows. If the company does not know who uses which model and why, cost optimization becomes impossible.
AI governance is therefore not just a control mechanism.
It is a management tool for AI ROI.
Companies need to decide which tasks require expensive models and which do not.
They need to separate work that can be automated from work requiring human approval.
They need to connect model usage to business outcomes.
They need to manage AI spending by department, workflow and use case.
They need to balance security risk with cost efficiency.
From this perspective, Unity AI Gateway is not only a security product.
It is an economic control layer for AI usage.
The CFO, CIO and CDO will increasingly ask the same questions.
How much are we spending on AI?
Which business outcomes are we getting?
Could a cheaper model produce the same result?
Is sensitive-data access being governed?
Databricks is turning those questions into a platform opportunity.
The Funding Round Is a Preview of the AI Acquisition Race
Databricks also said the new capital is expected to support future AI acquisitions and deepen AI research.
That suggests the company may become more aggressive in buying AI-related technology and teams.
The enterprise AI platform battle requires many pieces.
Model operations.
Evaluation.
Security.
Data connectivity.
Agent development.
Application generation.
Cost management.
Search.
Knowledge graphs.
Workflow automation.
No company can build all of this internally at the same speed the market is moving.
Large platforms will therefore acquire promising startups to fill product gaps quickly.
That is another reason the new capital matters.
Enterprise AI infrastructure is moving fast. Customers are shifting from experimentation to production. In that environment, a missing capability can become a lost market opportunity.
Databricks will likely use its capital to expand the platform, acquire missing capabilities and respond to Snowflake and Big Tech competitors.
The $188 billion strategic funding round is both defensive and offensive.
Defensively, it gives Databricks capital to avoid falling behind in the enterprise AI platform race.
Offensively, it gives the company firepower to pursue AI governance, agent infrastructure and data-control markets more aggressively.
The Winner of the AI Era May Be the Company That Moves Data
Databricks’ announcement shows where the AI industry is heading.
The first phase of AI competition was model-centric.
Who could build the biggest model?
Who could score highest on benchmarks?
Who could generate the most natural answer?
Those questions still matter.
But the enterprise market is now asking different questions.
Who can make company data usable by AI?
Who can govern multiple models across workflows?
Who can control AI cost?
Who can inject business context into AI answers?
Who can provide the data infrastructure that lets AI agents act?
Databricks is saying that it wants to be the answer to those questions.
The $188 billion valuation is not just a private-market price tag.
It is a signal of where investors believe the next enterprise AI battlefield lies.
AI models will continue to change.
The name of the smartest model may change.
But enterprise data, governance, cost control and business context are much harder to replace.
Databricks is aiming for that sticky layer.
The real AI platform of the enterprise era may not be the company that offers one more model.
It may be the company that lets enterprises use their own data safely, route work across multiple models cost-effectively and give AI agents the infrastructure to perform real work.
Databricks’ $188 billion funding push shows that this battle has already begun.
