Cloudera is pitching its Anywhere Cloud platform to Indian banks and regulators as a way to run AI across clouds and data centres without moving sensitive data.
Many companies can build an AI demo, but few can run it safely in production. At a customer and partner briefing in India on October 8, Cloudera said the fix is to take AI to the data instead of copying data to AI.
What is launching
Cloudera Anywhere Cloud was announced on August 19 at the company’s Evolve event in Singapore. It lets firms deploy data and AI services across public clouds, sovereign clouds, and private data centres through a single control plane, with general availability set for November 30.
CTO Sergio Gago said a deployment that once took two to three months can now go “from rack to AI inference” in 60 minutes. He said early partners such as ExxonMobil, IQVIA and Mastercard have tested it. The platform includes nine data services, an agentic copilot that turns plain-language requests into data workflows, and integration with Mistral models.
Why pilots fail
Gago said many generative AI projects stall at the compliance stage. Teams copy data into separate systems and build agents, but cannot show who accessed what or what it costs. He put the share of use cases that never reach production at “90-something per cent”.
The company’s answer is to let agents inherit the access rules, audit trails and lineage already in place. “Lineage is the missing piece on agentic workflows,” he said, because users must know where an answer’s data came from.
The banker’s view
Shashwat Kumar of DCB Bank said banks now want “decision-grade data”, meaning data that is both fresh and trustworthy. He argued that value comes from decisions, not models, and that even a strong model fails if its output does not reach the relationship manager in time.
On regulation, he took a counterintuitive line. RBI’s rules, he said, work like a strong brake that lets a car run faster because they give clarity. His approach is to keep steady, sensitive workloads on-premises and use the cloud for variable, low-sensitivity ones. “Everything everywhere is not what AI anywhere probably should be,” he said.
Sovereignty and lock-in
Gago tied sovereignty to business continuity. He said large language models are becoming critical infrastructure, and relying on a few providers from one country is a risk. He cited the recent suspension of access to Fable and Mythos for non-US users as an example of such limits.
His recommendation was “optionality”: keep using hyperscalers, but be able to move between clouds, sovereign providers and data centres with one click. He said that “the days of data hoarding are over” and that customers can bring open-standard data, such as Iceberg tables, and engines from rivals like Snowflake.
India’s role
Mayank Baid, Regional Vice President for India and South Asia, said about three-quarters of the country’s largest banks use the platform. So do most leading telcos, the GST Network and the Bombay Stock Exchange. Cloudera runs engineering centres in Bangalore and Chennai, and executives said the public sector and BFSI lead growth.
On acquisitions, Gago said the focus is on delivering the new platform, though the company is watching startups in the engines and databases space.