Running AI against data a company refuses to move used to be Cloudera's argument alone. The company built its whole fiscal year 2026 story around that idea. Today it turned the idea into a product.

Cloudera Anywhere Cloud does not ask a company to move its data anywhere. It sends the AI and data tools to wherever the data already sits: the public cloud, the company's own data centers, or systems built to keep data inside a country's borders. That is the sovereignty case in one sentence: the workload moves to the data, so the regulated data never has to cross a border to get used. All of it runs from one control panel, and an AI assistant handles the setup. Tell it what you need in plain language, and it builds the workflow (Cloudera, Aug. 2026).

Cloudera calls itself the only company that can do this. Five other major vendors are making a similar case about their own platforms this year. Databricks made almost the same promise about its own platform ten weeks earlier. Before either of them, Amazon Web Services turned on a live sovereign cloud region in Brandenburg, Germany, in January, built so customer data never leaves the European Union (CX Today, 2026). Oracle has sold the same idea longer, through OCI Sovereign Cloud, Oracle Alloy, and Compute Cloud@Customer (Oracle, 2026). IBM previewed its own Sovereign Core that same month, built to run AI workloads under strict data-residency rules (Network World, 2026). Google Cloud expanded its own sovereign AI lineup in April, adding tools that let its models reach a customer's data without moving it (Google Cloud, 2026). Cloudera and Databricks are the latest two entrants into a category several major vendors are already building toward.

The regulatory clock explains part of why all of this happened this year. For most of 2026, every high-risk AI system in the European Union faced new compliance obligations starting August 2, and vendors built toward that date. The European Union deferred those obligations at the last minute: the law pushing the deadline to December 2027 and August 2028 entered into force July 27, 2026, twenty-three days before Cloudera's announcement. The Act's transparency rules still took effect on the original August 2 date. The deferral did not undo a year of vendor investment built around keeping data in place, and buyers still have GDPR and a growing list of national data-residency rules to satisfy regardless of what the AI Act itself requires.

The Clearest Comparison: Databricks

Among the field, Databricks is the closest architectural match. In June, Databricks announced something similar: the Databricks Storage Ecosystem. It runs on an open-source tool called OpenSharing, which connects Databricks to a company's own storage, on-site or in a private cloud, without copying any data out. Hundreds of customers had already asked for this. The storage partners involved manage more than two zettabytes of data between them (Databricks, 2026).

Cloudera's pitch works the same way. Workloads run in the public cloud, in sovereign systems, or in private data centers. No migration. One set of governance rules. All of it built on Apache Iceberg, an open format for storing data, and a shared catalog that tracks where everything lives.

Cloudera and Databricks landed on the same design for the same reason every vendor above did. Moving regulated data across borders is difficult, and enterprise buyers have increasingly asked vendors to bring the platform to the data instead.

Where Cloudera and Databricks Differ

Cloudera's chief product officer, Leo Brunnick, said it in one line: "Enterprise AI has outgrown the public cloud-only model" (Cloudera, Aug. 2026). Cloudera Anywhere Cloud runs on Apache Iceberg and the Polaris catalog, both open standards, plus shared APIs. That means other software can read the same data in an open format.

One example: PuppyGraph already builds on top of Cloudera's setup. It queries the same Iceberg data as a knowledge graph, no ETL, no copying anything. Cloudera highlighted the partnership in its release (Cloudera, Aug. 2026).

The catalog covers part of the platform.

The layer above it, the piece that decides what runs where, is where Cloudera and Databricks compete.

At Cloudera, an AI assistant decides where a workload runs, what gets set up, and how the rules apply everywhere the platform touches. That piece is Cloudera's own, built and run by Cloudera. Databricks draws the line in about the same place. Its catalog, Unity Catalog, is open, other tools can read it. But the software that decides what an AI agent does with the data stays inside Databricks' own products (Databricks, 2026).

Open data formats solved one problem: moving data freely between systems. Enterprises spent a decade fighting for that. The decision-making layer above it is the newer problem: customers still cannot see or change most of it. Both companies are still working out how to answer that question in public.

What the Buyer Gets to Decide

IXEN.ai, a Cloudera partner in Singapore, said it plainly in its own launch quote: teams want room to experiment, and they want to know the data stays under their own control (Cloudera, Aug. 2026). That trade-off is what keeps most AI projects from ever reaching production. Cloudera has the numbers to back that up. In its own Data Readiness Index, 73 percent of information technology leaders said infrastructure problems had held back their work (Cloudera, Apr. 2026).

Cloudera Anywhere Cloud, like the Databricks Storage Ecosystem before it, covers the infrastructure half of that problem. The contract covers the rest: whether a different AI system, built in-house or by a competitor, can read and write the same data without going through the vendor's own copilot first.