Yongwin Fund completed a 14.869-million-yuan purchase of domestic computing servers in late September to expand its GPU cluster for deploying and serving models at the trillion-parameter scale and beyond, China Fund News and 21st Century Business Herald disclosed on November 2.
Tender records show the project carried a 15-million-yuan budget for 4 domestic servers, opened bids on September 21 with an uncontested candidate list on the 24th; the winner was Beijing Kelibang Information Technology. Yongwin told media its GPU cluster is already deployed and keeps expanding with usage.
The company gave two reasons for building its own capacity: keeping "data in-domain" to accumulate proprietary data assets, and securing infrastructure for rapid model and agent iteration. On the application side, Yongwin pushes AI across four core lines — research, risk control, marketing and operations.
Self-built compute has so far been the habit of large overseas asset managers; a domestic public fund putting a domestic GPU cluster on its procurement list is still novel. Wind data puts Yongwin's public fund AUM at 700.46 billion yuan as of Q3 2026, ranking 21st overall — a mid-to-large fund preparing its own compute for trillion-parameter models signals that domestic compute's cost and compliance thresholds have entered the affordable zone for asset managers.
Where data sensitivity meets model appetite, "domestic compute plus private deployment" is becoming the practical option in asset management; the actual utilization of these four servers will be a small but telling sample.