As generative AI technology rapidly evolves, Taiwanese companies are quickly moving from early pilot phases to full-scale operation and deployment. However, while companies are pouring resources into purchasing expensive computing equipment, the development gap in underlying storage infrastructure is quietly becoming the biggest stumbling block to the long-term return on investment (ROI) of AI.
Recognizing the need for AI infrastructure upgrades, Hitachi Vantara is promoting its VSP One virtualization storage platform and advanced storage and object storage solutions. Hitachi Vantara Taiwan General Manager Jiang Weiyi and Taiwan Chief Technology Advisor Lin Qizhen further analyzed the "infrastructure bottlenecks" and "data governance risks" faced by enterprises in the process of scaling up AI, and further dissected the current status of industry adoption in the Taiwan market.
The predicament of underutilized computing power: avoiding turning expensive GPUs into "supercars on muddy roads".
According to Hitachi Vantara's "2025 Data Infrastructure Status Report," a staggering 99% of surveyed Taiwanese companies have already implemented AI, with 70% achieving initial success. However, only 29% of these companies believe they are ready to realize the long-term return on their AI investments.
Faced with this huge gap between expectations and reality, Lin Qizhen pointed out the key problem: enterprises spend heavily on purchasing powerful GPUs, but neglect the synchronous upgrade of their data storage infrastructure. He vividly compared it to buying a top-of-the-line supercar, ready to take it for a spin, only to find a muddy road in front of you. When the underlying storage architecture cannot provide sufficient data access throughput, the expensive front-end GPUs will often be idle, and their computing power cannot be effectively utilized.
Jiang Weiyi also emphasized that the development of AI is definitely not a bubble; it is substantially driving huge storage demand and accelerating the replacement of existing old equipment that cannot handle the massive workload of AI.
Taiwan is a leader in AI adoption in four sectors: government, finance, healthcare, and manufacturing.
Observing the current pulse of the Taiwan market, Jiang Weiyi analyzed that the areas with the fastest pace of data application and AI adoption are mainly concentrated in four sectors: government agencies, finance, healthcare, and manufacturing. Although these four industries face different pain points, their needs for "high performance" and "high compliance" in the underlying architecture are surprisingly similar.
• Government agencies:With the advancement of smart cities and convenient public services, government agencies need to integrate vast databases across departments. The core challenge lies in breaking down data silos and achieving data integration and AI-assisted decision-making while ensuring national-level cybersecurity.
• Financial industry:As a highly regulated industry, the financial sector has extremely stringent requirements for system availability when adopting AI for risk management, anti-money laundering, and precision marketing. Enterprise-grade high-end storage can ensure that these critical services are not interrupted due to downtime or performance fluctuations.
• Medical industry:Precision medicine and AI-assisted diagnosis (such as medical image interpretation and gene sequencing) have led to an explosive growth in data. Medical data is not only massive in size, but also involves extremely sensitive patient privacy and is highly dependent on storage environments with de-identification capabilities.
• manufacturing:As a global hub for technology manufacturing, Taiwan's high-tech factories are actively utilizing AI for production line defect detection, yield analysis, and predictive maintenance. These applications require processing massive amounts of sensor and image data in extremely short timeframes, placing extremely high demands on storage system throughput.
The Battle Against Unstructured Data: Automatic Sorting and De-identification to Create a Cybersecurity Protection Network
Beyond performance bottlenecks, data governance is a hidden time bomb in the era of large-scale AI deployment. Surveys show that 41% of enterprises indicated that the high complexity of their data environments makes identifying cybersecurity incidents more difficult; meanwhile, only 11% of enterprises are considered to have high data maturity, with the majority still operating under decentralized data management.
Lin Chi-chen further explained that in the current enterprise environment, "unstructured data," including documents, videos, and sensor logs, constitutes a large portion of the data, which is precisely the nutrients that AI machine learning craves. However, if this data is not properly managed, or if sensitive data, including personally identifiable information, is directly fed into AI models for training, the company will face huge compensation claims and reputational damage in the event of a data breach.
"Data must be refined to extract useful and valuable content in order to drive the advancement of AI." To this end, Hitachi Vantara's VSP One unified data platform provides a complete and concrete solution:
• Breaking down data silos:By integrating the VSP 360 management interface, enterprises can centrally manage data distributed across the ground, edge, and cloud on a single platform, and easily assign data weights.
• Storage of objects with a "brain":The new VSP One object storage supports industry-first native Amazon S3 Tables functionality and Apache Iceberg, helping enterprises build modern data lake warehouses. Its built-in automation mechanisms can improve data quality, perform decentralized processing, and remove personal data, ensuring that the value of AI is realized without crossing regulatory red lines.
• Ultimate usability and sustainable energy efficiency:For mission-critical environments, the VSP One architecture achieves "eight nines" high availability (equivalent to only 0.3 seconds of downtime per year). Furthermore, the device incorporates a smart energy-saving mechanism that automatically reduces the operating efficiency of components such as fans during periods of low usage, helping businesses save electricity and reduce carbon emissions.
Data refinery is the ultimate foundation for monetizing AI.
From pilot projects in a single scenario to large-scale deployment across the entire company, the maturity of "data governance" will determine the level of AI development.
In AI infrastructure, storage systems are no longer just "data warehouses" that simply store files, but must evolve into "data refineries" with intelligent analysis, automatic sorting, tagging management, and security protection.
Hitachi Vantara, through its VSP One unified data platform, not only addresses the physical performance bottleneck of idle GPU computing power but also lays the foundation for a robust "data highway" through a powerful unstructured data governance mechanism. Only by acknowledging and investing in resilient data infrastructure can enterprises truly unlock long-term business value in this AI wave.
