From Two Hard Drives to an Investment Thesis
Last August, Leto Bao, a former ByteDance employee, bought two Seagate large-capacity hard drives on Pinduoduo for his quant trading platform. After delivery, he noticed the same model kept rising in price—multiple adjustments within a week, only up. For a standardized industrial product, this was abnormal. Using price tracking tools, he confirmed the trend across Seagate and Western Digital models, pointing to a systemic shortage rather than temporary promotions.
Tracing the Cause: AI Storage Demand
Investigation revealed that AI model training and inference generate enormous data requiring long-term, low-cost storage, primarily on large-capacity HDDs (not SSDs). Cloud giants like Microsoft, Amazon, and Google are buying nearline enterprise drives in bulk. Seagate's HAMR technology boosts per-drive capacity, aligning perfectly with data center needs. With limited production capacity, enterprise orders squeezed retail supply, explaining the Pinduoduo price hikes. Seagate's quarterly report showed 39% revenue growth and record gross margins, signaling the market began pricing storage as an AI supply chain component. Bao bought 500 shares at ~$150, sharing his rationale internally.
Using 13F Filings to Validate Institutional Interest
Before adding more, he needed institutional confirmation. U.S. 13F filings disclose quarterly holdings of managers with over $100 million AUM. He waited for two consecutive quarters of upward trends. In H2 2024, STX had ~800 holders; by Q2-Q3 2025, holders surged to 1,200+, with new filers increasing each quarter. While market value growth to $45.6 billion partly reflected price appreciation, the breadth metric (number of institutions) confirmed sustained professional interest. He then scaled up via LEAPS calls on $STX and $SNDK.
Results and Methodology
From $150 to $965, Seagate returned over 6x, briefly becoming the S&P 500's top gainer. The initial 500 shares yielded ~$400,000 paper profit. Bao's playbook: (1) notice everyday anomalies (price hikes, shortages), (2) plot data to separate trend from noise, (3) ask if the demand is long-term structural and find the key listed beneficiary, (4) validate with institutional filings over multiple quarters. He emphasizes survivorship bias—past failures exist but were not shared. This is not investment advice; readers should always do their own research.