Is ASEAN’s AI Choking on Its Own Data?

Octavian Tanase doesn’t mince his words when it comes to the ongoing data fragmentation — the bane of today’s model infrastructure. “Fragmentation is the enemy of accuracy,” says the chief product officer at Hitachi Vantara, referencing findings from the company’s State of Data Infrastructure Survey.

The survey numbers are stark: Only 30% of enterprise data is structured. The other 70%? Messy, unstructured, and largely unusable for the AI systems companies are desperately trying to deploy. This is no storage crisis. It’s more of a data infrastructure crisis hiding in plain sight.

The impact on ASEAN will also be greater. Why? Because 42% of Asian enterprises already consider AI critical to their operations, five percentage points higher than the global average. But AI accuracy in the region sits at just 32%. This means the gap between ambition and capability in the region is wider.

“While this isn’t a problem unique to ASEAN,” Tanase explains, “the region's rapid pace of AI adoption makes this execution gap more visible.” Translation: Everyone has messy data. But when you’re moving faster than everyone else, the crash is louder.

The dark data catastrophe

Indonesia offers a glimpse of how deep the problem runs. Dark data — information that exists but remains unclassified, untagged, and unusable — accounts for 24% of enterprise data in the country. That’s more than double the global average of 10%. Companies are sitting on mountains of information they can’t even see, let alone use.

Trust follows close behind. Indonesian enterprises report some of the lowest confidence levels in AI outputs worldwide. When your training data is invisible or incoherent, your models hallucinate, make costly mistakes, and erode confidence across entire organizations.

The paradox deepens when you consider scale. Data storage requirements across Asia are projected to grow 123% in the next two years. Companies are drowning in information while simultaneously starving for insights. “Simply adding more capacity is not enough,” Tanase warns, “because it increases cost and energy use without addressing long-term efficiency.”

This is amplified by the multicloud sprawl challenging ASEAN enterprises, which makes control harder.

The USD120B question

External projections estimate that AI and generative AI could contribute approximately USD120 billion to the GDP of the ASEAN-6 economies by 2027. But that windfall only materializes if companies can actually operationalize their data.

This is where data infrastructure stops being a back-office concern and becomes a strategic weapon. The companies that figure out data reduction, unified storage architectures, and real-time orchestration will capture disproportionate value.

To directly address the sprawl, Tanase notes: “That is why we built VSP One to unify block, file, and object storage into a single platform, eliminating the need to manage silos across different environments.” Hitachi Vantara’s VSP One includes data reduction technologies like compression and deduplication to cut storage overhead, and synchronous and asynchronous replication to safeguard availability across sites.

Security amplifies the urgency. Nearly half of Asian enterprises cite it as their top concern, and for good reason. Agentic AI — autonomous systems that can make decisions and execute tasks — raises the complexity to a new level. “Autonomy without oversight is a risk,” Tanase says flatly. “Agentic AI will only scale responsibly if enterprises invest as much in controls, auditability, and safeguards as they do in performance.”

The efficiency imperative

Cloud costs are rising so rapidly that efficiency has become a board-level priority across ASEAN. The companies seeing the fastest ROI aren’t the ones with the most storage but those actively optimizing how they use it.

To balance scalability and sustainability, Hitachi Vantara designed its VSP One Block NVMe-QLC systems to deliver up to 14.1 petabytes of effective capacity, with guaranteed data reduction ratios of 2:1 to 4:1. Furthermore, “The systems are also ENERGY STAR® certified, combining high-density QLC drives with power efficiency to lower energy consumption per terabyte,” Tanase states.

Data reduction technologies, thin provisioning to eliminate wasted capacity, and replication engines that rebalance workloads across environments are no longer nice-to-haves. They're the difference between capturing that USD120 billion opportunity and burning budget on infrastructure that can’t deliver.

What agentic AI actually demands

The next wave makes today’s challenges look quaint. Agentic AI doesn’t wait for batch processing or overnight data pipelines. It expects near-real-time orchestration, instant workload agility, and governance that operates at machine speed.

The challenge for enterprises is that “What they deploy today may not fit tomorrow’s needs, and over-investing upfront can lock them into rigid infrastructure,” Tanase cautions.

To solve this, the Hitachi iQ M Series was built with a modular, right-sized architecture that adapts as AI workloads grow. “By integrating NVIDIA GPUs with high-performance networking and storage, the iQ M Series provides the accelerated compute needed for demanding use cases like retrieval augmented generation (RAG), fine-tuning, and inferencing,” he adds. This system supports mixed, multi-modal workloads with 100% availability.

Tanase identifies three specific breaking points for ASEAN infrastructure teams: data orchestration across silos, workload agility that can scale instantly rather than incrementally, and governance that can audit autonomous decisions in real time. The Model Context Protocol (MCP) — a technology that can enable agents to connect with sovereign data sources through a common standard — is emerging as critical infrastructure here, though adoption remains uneven.

The human dimension everyone ignores

Lost in the infrastructure conversation is what happens to people. The World Economic Forum projects AI and information processing trends will create 11 million jobs and displace 9 million by 2030 — the largest net impact of any technology trend. ASEAN enterprises building AI infrastructure today are also reshaping their workforce tomorrow.

“ASEAN enterprises must think beyond siloed experimentation,” Tanase argues. “They need to combine better structuring and governance with proactive reskilling so that both their technology and their people are prepared for AI.”

“For ASEAN infrastructure teams, the goal is architectures that can support real-time autonomy while still guaranteeing trust, transparency, and security,” he emphasizes.

The window is narrow. ASEAN’s AI adoption curve is steeper than anywhere else, which means both opportunities and failure modes arrive faster. Companies that solve the 70% problem — turning unstructured chaos into AI-ready datasets — will dominate their markets. Those that don’t will continue to wonder why their AI investments never paid off.

Image credit: iStockphoto/Doucefleur