Siarhei - stock.adobe.com

News

Couchbase eyes APAC growth as enterprises confront data problem

Having been taken private by private equity firm Haveli Investments, the NoSQL database supplier is pitching its newly launched AI data plane as the fix for the fragmented data layers

Aaron Tan
By
Published: 14 Jul 2026 10:30

In Japan and South Korea, artificial intelligence (AI)-powered avatars are being deployed to keep isolated elderly people company – and unlike most chatbots, they are built to remember last week’s conversation.

The avatars are developed by Agora, a real-time engagement platform provider, on top of Couchbase’s AI Data Plane, a unified data infrastructure layer for AI agents that the database supplier made generally available on 30 June.

“There’s a movement to deliver avatars to combat elderly isolation and loneliness in Korea and Japan,” said Michael Cronin, managing director for Asia-Pacific (APAC) at Couchbase. “They’re using the AI Data Plane to have those avatars built with a conversational memory, to build relationships with their users over time.”

The use case captures the pitch Couchbase is now taking across the region: that the biggest obstacle to enterprise AI is not the models, but the fragmented data infrastructure underneath them.

Cronin, a database industry veteran whose career spans DataStax, Yugabyte and Databricks, is about 18 weeks into the new role, and plans to relocate from London to Singapore by the end of summer. His territory, he said, stretches from Tokyo to Sydney, with each market bringing its own regulatory and customer nuances.

The one constant, he added, is pressure to deliver on AI. “From the board on down, the push to develop on AI is really high,” he said, noting that adoption in Asia is moving faster than in the West.

Yet most of those early efforts have foundered. “There was a mad gold rush to get into AI development at the onset, but name your analyst report from Gartner and Forrester – the stats say anywhere between 80% and 90% of AI development has failed to move into production,” said Cronin.

The culprit, he pointed out, is decades of point solutions. “What they wound up with is almost a Frankenstein-like monster of a fragmented data layer,” said Cronin. “The key to AI development is to move that data layer closer to the customer in a seamless way, so that AI can really shine.”

Memory, models and governance

Couchbase’s AI Data Plane brings together what it calls agent memory, a persistence layer that lets AI agents retain context across sessions; an agent catalogue for discovering and governing agent tooling; and a model context protocol (MCP) server for integrating agents with external systems. It runs across Couchbase’s Capella managed cloud service and self-managed deployments, and is complemented by Enterprise Analytics 2.2, which federates queries across Apache Iceberg-based lakehouses.

The agent memory capability is framework-agnostic and has been validated with the LangGraph, CrewAI and LlamaIndex orchestration frameworks, so development teams can switch frameworks without rebuilding their memory layer. IDC research director Devin Pratt, commenting on the launch, said moving from chat-style pilots to production agents is “a data problem, not just a model problem”.

Cronin said the governance piece is resonating in a region where AI regulation is tightening. “AI governance is becoming fairly strict,” he said. “There are the DPDP [Digital Personal Data Protection] regulations that are fairly new in India, and Singapore is quite advanced in AI development but also pretty strict when it comes to regulation. The agent catalogue allows enterprises to control and audit the actions of their AI usage.”

Keeping frequently used context in memory also cuts unnecessary inference calls and token consumption, he added, lowering the total cost of ownership of AI applications.

Outside the region, Cronin pointed to UK grocer Tesco, which consolidated caching, database, search and analytics workloads, including vector search, onto Couchbase after a sprawl of systems began eating into service-level agreements. “It limited their ability to introduce new functionality, including AI functionality, because there was no headroom in their architecture to expand or innovate,” he said.

Couchbase is hiring across the region to meet demand, said Cronin, adding: “I have headcount in Tokyo, Korea, India, Singapore, Australia – we’re hiring for all roles, including sales.”

It also leans heavily on local partners, which can help the company “accelerate and grow without growing headcount at the same time”.

A crowded field

The AI Data Plane is Couchbase’s first major launch since it was taken private by Austin-based private equity firm Haveli Investments in September 2025, after which BJ Schaknowski, formerly chief executive of healthcare software firm Symplr, replaced long-time CEO Matt Cain.

The deal was part of a broader wave of consolidation in the database market as suppliers reposition for AI workloads. IBM acquired DataStax in 2025, while MongoDB bought embedding model specialist Voyage AI to bolster its AI credentials.

MongoDB remains the largest independent rival by some distance, reporting revenue of $2.46bn for its fiscal 2026 – roughly 10 times Couchbase’s last disclosed annual revenue of about $210m before it went private. MongoDB has made a similar pitch, with chief executive CJ Desai recently touting an integrated offering for AI agents that combines search, vector search and embeddings in what he called a “single intelligent data layer”.

The hyperscalers pose a different kind of threat, with Amazon Web Services’ (AWS’s) DynamoDB, Microsoft’s Azure Cosmos DB and Google Cloud’s Firestore all offering overlapping NoSQL capabilities, and all three now bundling vector search into their database portfolios.

However, Cronin noted that the prevalence of hybrid and multi-cloud estates in the region blunts that competition. Tesco, for instance, self-manages Couchbase across Microsoft Azure and its own private cloud. “We don’t have a preference for any of those deployment models, and that’s what our customers and prospects like about us.

“The idea that they could have a single data layer that spans their private cloud and the hyperscalers, while looking like a single database to their customers and applications, is really attractive to our customers.

“From that perspective, we don’t feel that kind of pressure from the hyperscalers,” said Cronin. “In fact, we find them to be quite partner-friendly. We do quite a bit of work with the Google, Azure and AWS teams on collaboration and co-selling.”

Read more about data management in APAC

Read more on Database management