Steven Xi is a venture capital investor at Eastlink Capital with deep cross-border tech investing experience bridging Silicon Valley and China, having previously led China-related investments at Riverwood Capital and been an early investor in companies like Databricks, Uber, and MobileIron. He combines an unusually technical academic background in physics and econometrics with hands-on experience as a startup co-founder and corporate development executive, and was a pioneer as the first Mainland Chinese corporate finance associate at CSFB on Wall Street.

• Strong belief in the interdependence of AI and data infrastructure, emphasizing that effective artificial intelligence is impossible without a robust data layer. He highlights specific engineering achievements in this space, such as Databricks expanding into transaction processing and StreamNative’s work on serverless data streaming.
• Focus on the operational tooling and infrastructure required to build and scale Large Language Models. He actively promotes portfolio companies like Vellum for their work on agents and Modal for providing the compute infrastructure used to train models for major tech companies like Meta.
• A perspective that the competitive edge in AI is shifting from raw compute power to data quality. He suggests that the industry is moving away from simply having the biggest models and most GPUs toward prioritizing the pipelines used to generate proprietary, high-impact data, pointing to Scale AI and Mercor as key examples of this trend.
• Value placed on long-term relationships with founders, often stretching back to their academic roots. He frequently shares anecdotes about meeting entrepreneurs, such as the Databricks team or students at Stanford, years before they achieved massive success, highlighting his appreciation for the journey from research to large-scale enterprise.
• Attention to financial and operational metrics when evaluating the success of data companies. He compares growth rates, net dollar retention, and AI readiness between major players, specifically noting how Databricks stacks up favorably against competitors like Palantir and Cloudera.