Interview: Cloudera bets on hybrid AI growth in JAPAC

Cloudera sees a growing opportunity in Asia-Pacific as banks and other large enterprises reconsider where they run artificial intelligence workloads, with cloud costs, data sovereignty requirements and regulatory constraints pushing hybrid architectures higher up technology agendas.

The data platform provider is betting that Cloudera Anywhere Cloud, alongside its new applied AI capability, will help it capture spending as organisations move beyond AI pilots and focus on deployment, governance and return on investment.

Enterprises are also reassessing earlier cloud-first strategies. Customers increasingly want to move workloads between public cloud, private infrastructure, data centres and edge environments without redesigning applications for each location.

Hybrid shift

"Cloudera Anywhere Cloud is fundamentally different. We've made three key acquisitions. The last one, Taikun, basically provides us this foundation layer to be able to containerise all of our offerings and services and be much more modular. Our current platform is more monolithic, and with the new platform, the ability to write once and deploy anywhere is the biggest key. You can move or bring AI to your data, whether it's at the edge, in the data centre or in the public cloud," said Brian Rosso, Chief Revenue Officer, Cloudera.

The architecture is designed to remove infrastructure dependencies that previously complicated hybrid deployments. Individual services can be containerised and deployed independently while remaining under a common governance and security layer.

That approach is particularly relevant in Asia-Pacific, where large financial institutions are revisiting architecture decisions ahead of wider AI deployments.

"All the banks are relooking at how to rearchitect, and I think that's a huge opportunity. In APAC, it is probably a little bit different from the US, where customers are a little bit more conservative about an all-out cloud-first approach. At the same time, they would also like to have a hybrid solution to make sure that for certain workloads that make sense to go on cloud, they can do that. A lot of customers are trying to build a hybrid architecture," said Remus Lim, Senior Vice President, JAPAC, Cloudera.

Infrastructure choices are increasingly being made workload by workload rather than under a blanket cloud-first policy. That creates an opening among organisations that have moved significant estates to public cloud but want some AI processing under their direct control.

"We think it's a huge opportunity for us because when you look at Cloudera, we've been talking about hybrid for the last probably four or five years. It has not been a popular discussion, at least in the US. We never believed that enterprises would be 100% cloud. For those companies that have moved, that's where we're seeing a significant opportunity. It's not that they're always repatriating workloads, but they are setting up these private AI, sovereign AI factories within their data centres," added Rosso.

Sovereign demand

Sovereignty is also becoming central to infrastructure decisions, particularly in regulated industries and government, where organisations want more control over where data is processed, who manages it and how AI systems access it.

Cloudera's approach extends beyond data residency. Private and air-gapped deployments can also address security, privacy and governance requirements while retaining a cloud-style operating model.

"It's not necessarily just data residency or where the data resides. It's also the exposure because you are now leaving your data for someone else to manage. Security and privacy are definitely big concerns. There are customers like governments that, for sure, want to have this sovereignty where they want to put their cloud on the private cloud. We have an air-gapped environment, so it is not just a matter of where the data resides, but also the security and guardrails around the content or data that is residing there," added Lim.

Those concerns are converging with the economics of AI. Organisations that used public cloud infrastructure for proofs of concept can face a different cost profile once applications move into continuous production.

Generative AI adds another variable because inference and token consumption can make spending difficult to predict. Cloudera is positioning data-centre inference as an option for organisations seeking greater control over utilisation and infrastructure costs.

"When you take that pilot to production, you start realising this is always on. We're spinning the metre. Token costs are going out of control. From a CFO perspective, this is why we've seen so much movement back towards the data centre. You started out in the cloud doing your POCs because of the speed, scalability and ease of use. However, like everything, there is a cost involved," added Rosso.

Cloudera is also working with Nvidia on AI inference, embedding Nvidia NIM microservices into its inference service. The aim is to improve performance while allowing customers to run inference within their own data centres.

Applied AI

Cloudera has also established an applied AI team comprising data scientists and Forward-Deployed Engineers, or FDEs, who work directly with selected customers to turn AI use cases into deployed systems.

Created only several months ago, the capability differs from conventional customer success and pre-sales support. Cloudera plans to deploy these specialist resources selectively, including across Asia-Pacific.

"I think technology is there, but the challenge is, from a business point of view, you need to map it into your environment. You need to set up processes. You need to embrace technology and even change management. With the implementation of agents, the workflow has changed. It's not just implementing technology. When we say applied AI, we are basically going in to help customers look at how to drive use cases," added Lim.

The programme is intended to accelerate adoption and increase the likelihood that AI investments move beyond experimental deployments. It also reflects pressure on technology executives to demonstrate returns after a period of extensive pilots and AI spending.

"For us, it is about how we accelerate use-case adoption. It's complicated. It's not easy. You have all the data, security, governance and lineage. How do we deliver outcomes faster, cheaper and more successfully within organisations? What we've found, given the fast pace of new technology, is that you do have to be close. You have to embed someone that understands your platform and understands the AI applications," added Rosso.

Partners will continue to play a major role in implementation, including infrastructure vendors, systems integrators and AI model providers. Cloudera is designing its architecture to support different processing engines, AI frameworks and models rather than requiring customers to adopt a single technology stack.

AI budgets

Lim expects AI-related spending to remain one of the largest sources of incremental technology investment among major Asia-Pacific enterprises.

The opportunity extends beyond budgets explicitly labelled as AI. Large-scale agentic applications also require data infrastructure, compute capacity, governance and supporting platforms.

The scale of financial services organisations in markets such as India could make those requirements substantial. Large banks can serve customer bases comparable with the populations of several countries, creating significant demand for compute and data infrastructure as AI systems enter production.

"I think every company you talk to has a budget for AI. Everybody is going after their budget, so in my opinion, that is the biggest budget. One of the largest Indian banks has a huge amount of compute to drive some of their biggest use cases. They tell us their bank subscribers alone are the population of a few countries. Imagine the scale they have. To drive some of these agentic AI use cases, the amount of GPU power that's required and the platform that is going to support their application is going to be huge," added Lim.

Cloudera manages about 30 exabytes of data across its customer base, according to Rosso, with a significant portion remaining in enterprise data centres.

"We always talk about AI, and that's great, but it's the data that is the most valuable. As long as Cloudera is focused on providing that data platform that can be run in hybrid and at the edge, being able to provide that one true hybrid data platform with an agentic control plane, unified data governance, and being able to interoperate with other technologies is our focus," added Rosso.