Unblocking the Malaysia AI Scale Stall: Building the AI-Ready Data Strategy That Works

Malaysia’s digital economy is poised for a future where AI drives productivity gains across every sector, with projections indicating a 25.5% contribution to GDP by the end of 2025. The Malaysia Digital Economy Blueprint (MyDIGITAL) promises a 30% uplift in productivity across all sectors by 2030 through digitalization.

However, most organizations are stuck at the starting line, with engines revving but going nowhere. The problem? Data gridlock.

“Productivity isn’t everything, but in the long run, it’s almost everything,” says Dickson Woo, general manager and technology leader of IBM Malaysia, sizing up the real issue. “That’s because productivity is about more than just cost savings or efficiency — it’s about creating self-funding mechanisms that enable investment in innovation that drives growth and prosperity.”

The numbers are stark: 57% of enterprise data globally isn't AI-ready. In Malaysia, where manufacturing contributes 29.1% to GDP and grew 3.7% in the first quarter of 2025, this friction translates to billions in unrealized potential. Companies are drowning in data silos, wrestling with fractured lineage and burning budget on infrastructure that can’t scale.

The 2025 IBM CEO Study found that inadequate technology ranks among the top barriers to innovation, yet “65% of CEOs say allocating more budget flexibility for digital opportunities is critical for long-term growth and innovation,” Woo points out.

It creates a familiar paradox: AI promises transformation, but the very data foundation required to power it remains locked in legacy systems, scattered across clouds, and governed by patchwork policies that make compliance a nightmare.

The lakehouse that actually works

The open data lakehouse offers an architectural shift that can actually solve these challenges. IBM watsonx.data is built on an open lakehouse architecture optimized for governing data and AI workloads, with capabilities including querying, governance, and support for open data formats, supporting multiple query engines like Presto, Spark, Db2, and Netezza on a single governed platform.

“AI and Automation have helped IBM unlock extreme productivity gains across the company, and we are on track to reach USD4.5B in savings by the end of 2025,” says Woo. For Malaysian enterprises, the promise is similar: through workload optimization, organizations can reduce data warehouse costs by up to 50% by matching the right [AI] engines to the right workloads.

“Internal IBM benchmarks show up to 60% cost savings compared to alternatives like Databricks Photon,” Woo continues. IBM watsonx.data with Presto C++ and query optimizer delivered better price performance compared to Databricks' Photon engine, with equal query runtime at less than 60% of the cost, derived from public benchmarks.

Making data actually AI-ready

A major trend in today’s enterprise data landscape is the rise of unstructured data, which is now becoming vital for machine learning. However, storing and managing unstructured data poses a challenge for many companies that have fine-tuned their data infrastructure for structured data.

The integration of DataStax Astra DB, a NoSQL and vector database built on Apache Cassandra, into the watsonx stack tackles the unstructured data question head-on. Astra DB enhances the vector capabilities of IBM watsonx.data, built for elastic scalability and predictable performance, supporting mission-critical workloads with near-zero latency. Automated ingestion and vectorization transform PDFs, images, and logs into AI-ready formats. Retrieval-augmented generation (RAG) pipelines enable production GenAI without the usual data preparation nightmares.

For Malaysia’s financial services and manufacturing sectors, this matters enormously. Budget 2026’s focus on high-value manufacturing includes outcome-based incentive frameworks linking tax relief to innovation metrics. Companies need to move fast, but they can’t sacrifice governance. IBM watsonx.governance provides automated AI governance with agent monitoring, risk management and regulatory compliance, featuring end-to-end AI governance with automated workflows that ensure AI systems meet ethical guidelines while scaling.

More importantly, the approach also drives AI democratization through the use of open-source technologies and low-code tools. “Apache Iceberg, Presto, and Cassandra, combined with low-code tools such as Langflow, empower non-experts and citizen developers to build AI applications,” says Woo. It reduces reliance on scarce expertise while enabling collaborative development across skill levels.

The phased escape route

For enterprises trapped in costly legacy architectures, the transition path matters as much as the destination. “Inadequate technology is one of the top barriers in innovation,” notes Woo, adding that “75% of CEOs say they'll have to take more risk than their competition to maintain a competitive advantage.”

IBM's own CIO organization faced fragmentation due to data scattered across seven different data lakes built on various technologies, resulting in access issues, inconsistent governance, and rampant data duplication that increased operational costs.

The winning strategy? Start by attaching existing data sources to watsonx.data without migration. Utilize open formats, such as Apache Iceberg, to unify data across silos. Apply governance and lineage tools to ensure compliance. Then, gradually optimize workloads using fit-for-purpose engines. This minimizes disruption while unlocking value from systems you’re already running.

Producer, not just consumer

As Malaysia aims to become an AI solutions producer, not just a consumer, spooling open the vendor lock-in becomes an imperative. IBM watsonx.data uses open table formats like Apache Iceberg to facilitate collaboration and eliminate vendor lock-in, enabling users to share a single copy of data across multiple AI engines. This vendor-neutral approach enables interoperability, which is critical for commercializing “Made in Malaysia” AI products with enterprise-grade features.

DataStax Astra DB supports multi-region replication across cloud providers, replicating data across three availability zones within each deployed region to ensure high uptime and maintain data integrity. For Malaysian enterprises expanding regionally, such an architecture maintains consistent governance across diverse regulatory zones while ensuring high availability and reliability.

With manufacturing driving economic growth, Budget 2026 backing high-value digital infrastructure, and CEO confidence in AI-driven transformation at an all-time high, the window for building AI-ready data foundations for Malaysian enterprises is open. But it won’t stay open forever.

Image credit: iStockphoto/jamesteohart