Why “old” data is the new gold in the age of AI (www.techradar.com)

Seagate SVP Melyssa Banda argues that “old” data is the new strategic gold for AI: organizations should stop treating historical records as expendable archives and instead see every byte as potential training signal. With generative AI turning raw information into actionable intelligence, long-timespan datasets provide the context models need to go from plausible to precise — improving anomaly detection, predictive maintenance, trend analysis, transparency and regulatory auditability across finance, healthcare and manufacturing. Deleting historical data isn’t just short-term cost cutting; it erases future model improvements and institutional memory. Technically, most retained data in large deployments still lives on HDDs (about 87%) because modern AI workloads demand high-capacity, durable drives optimized for sustained throughput rather than pure latency. To control OPEX, organisations are moving to tiered storage: hot data on high-performance systems, cold or archival data on cost-optimized tiers (and complementary media where appropriate), plus smarter lifecycle policies that treat data as dynamic capital. The implication for ML teams and infra architects: invest in scalable, flexible storage and metadata/audit systems so historical data remains accessible, traceable and usable — otherwise you undermine model performance, explainability and regulatory compliance.