Cloudera has launched Anywhere Cloud, a modular data and artificial intelligence (AI) platform built to run across public clouds, private datacentres, sovereign infrastructure and air-gapped networks from a single control plane.
The supplier took the wraps off the platform at its Evolve 2026 conference in Singapore this week, with oil and gas giant ExxonMobil among the design partners that have spent the past few months using it.
“Enterprise AI has outgrown the public cloud-only model,” said Leo Brunnick, chief product officer at Cloudera. “Organisations shouldn’t have to choose between innovation and control.”
The intent behind Anywhere Cloud is to allow cloud-style provisioning of data and AI services without moving the data. The platform breaks Cloudera’s data services into modules that customers deploy from an Anywhere Cloud marketplace: data engineering, streaming, a Trino-based lakehouse engine, workflow orchestration and Cloudera AI.
The marketplace also carries partner engines that Cloudera has certified, such as graph query engine PuppyGraph and a range of vector databases, and a software development kit lets customers add engines of their own.
Much of what’s under the hood of Anywhere Cloud came from acquisitions. Taikun, the Kubernetes management firm Cloudera bought in August 2025, supplies the compute layer, while lineage and cataloguing capabilities from Octopai, acquired in November 2024, feed into a unified data fabric. An earlier deal, for operational AI specialist Verta in June 2024, added model management capabilities.
Compute where the data is
ExxonMobil joined the design programme a couple of months ago. Anvesh Koripella, an enterprise architect at the company who has run large Cloudera estates for a decade, said the first trial installation involved a slew of manual steps and took about six hours.
“When we did it last week, it came down to around 65 minutes, from the click of a button to get all the four experiences out and ready to go,” he said, referring to Cloudera’s AI, data engineering, data warehousing and Data Flow services.
The bulk of ExxonMobil’s data comes from sensors, gas meters and analysers at refineries and oil wells around the world, said Monika Thulasi, a system architect at the firm. The company is looking to implement a centralised data lakehouse holding structured and unstructured data, along with streaming analytics to support decision-making.
“Our aim is to move from batch-processing oriented data to streaming-oriented data,” Thulasi said, adding that Anywhere Cloud puts ingestion, streaming, analytics and AI capabilities “all together in a single pane of glass”.
Once the data infrastructure is in place, Thulasi hopes to use AI to spot faults in plant equipment, push alerts to a dashboard and work through root causes and fixes with engineers kept in the loop.
“Say, for example, there’s an image that shows some issues with the plant. We want to monitor it and send data to the dashboard so we can take immediate action,” she said. The aim is to move engineers “from working on day-to-day operations to high-value work”.
Anywhere Cloud runs on its own, but it can also plug into existing Cloudera Data Platform (CDP) deployments to query assets. Cloudera has committed to supporting CDP version 7.3.2 until 2032. “We designed Anywhere Cloud to be a standalone solution that does not need CDP,” Brunnick said. “However, it’s designed to sit side by side with your CDP and data services – and use the data that you have.”
Fell on deaf ears
Cloudera’s CEO, Charles Sansbury, said the company had spent two years talking to customers about sovereignty, security and cost. “And to be quite direct, the cost discussion fell on deaf ears,” he told CNBC’s Squawk Box Asia in Singapore. “In the past year, the volume has gone up dramatically.”
What changed, he said, was that companies moving workloads from testing into production were surprised by the costs. A fraud detection application that runs around the clock at a large bank should operate on an organisation’s own hardware, he said, as doing so on cloud-based infrastructure can be “three or four times as expensive”.
“There are other workloads that have peaky computing demands or are short-term in nature, need to be spun up and spun down very quickly, and those are better run on cloud infrastructure,” Sansbury added.
Anywhere Cloud is meant to let companies lay a control plane across workloads and match each workload to the infrastructure that fits, whether the priority is security, governance or cost. Sansbury put the time and cost of building it at three years, three acquisitions and almost $1bn in research and development.
Databricks takes a different route
Cloudera is not alone in going after data that enterprises won’t or can’t move due to sovereignty and cost pressures. Databricks announced its own storage ecosystem in June 2026, built on an open-source protocol called OpenSharing that lets storage suppliers expose on-premises data to Databricks’ Unity Catalog without copying it.
While the outcome is similar, their approaches differ. Databricks reaches into on-premises storage from a control plane that still runs in its own cloud, whereas Cloudera can be installed in the customer’s environment.
Lian Jye Su, chief analyst at Omdia, said that focus on sovereignty and control should land well in Asia-Pacific, where enterprises want a firm grip on their data and AI infrastructure and are looking to maintain operational sovereignty against a complicated geopolitical landscape. The marketplace matters too, he said, letting buyers pick best-of-breed components rather than commit to a single supplier’s stack.
Su was less convinced that Cloudera has closed the gap on AI tooling. Databricks retains an edge in agentic AI developer experience and infrastructure performance, he said, and has moved faster to embrace generative and agentic AI.
“Cloudera is not trying to be the fastest or most innovative in the room,” said Su, adding that its focus is on bringing comprehensive AI development and governance to on-premises, edge and sovereign data that must stay put.
Given what each is built for, Su expects enterprises to run both.
The launch comes a week after Cloudera published survey findings from 1,500 enterprise architects, cloud infrastructure leads and data architects, of whom 95% said they had delayed or cancelled AI initiatives over the past year because of governance, compliance or regulatory problems. Most (84%) also reported higher infrastructure costs driven by AI workloads.
Two-thirds had moved at least some AI workloads from public cloud back to private cloud or on-premises infrastructure over the past year, with Asia-Pacific close behind at 64%. In Singapore, 43% of organisations have shifted AI workloads out of public cloud, while a further 36% are evaluating the move.
Leave a Reply