Databento at NVIDIA GTC 2026: Highlights
Since the late 1990s, systematic trading firms have largely focused on two main challenges: reducing latency across the trading stack and improving the accuracy of their forecasting models. At NVIDIA GTC 2026, Christina Qi spoke about why data access—not compute—now determines competitive advantage in next-generation algorithmic trading. Christina is the CEO of Databento and previously co-founded a quantitative hedge fund that traded billions of dollars daily.
In her talk at NVIDIA GTC, Christina examines how data delivery, storage, and networking are evolving to support GPU-accelerated trading workflows and how Databento's partnership with NVIDIA is helping firms streamline access to market data at scale.
In an effort to reduce execution times from milliseconds to nanoseconds, engineers pushed into low-level software and hardware advancements, including microwave networks, FPGAs, ASICs, and increasingly specialized networking infrastructure. For years, competitive advantage depended on execution speed and the ability to process and respond to market data faster than anyone else.
Over the same period, forecasting methods evolved from linear regression to more complex machine-learning approaches, including gradient boosting and, eventually, deep learning trained on GPUs. Today, most leading firms have access to significant GPU compute. What remains difficult is accessing the volume and depth of market data required to train, test, and deploy those models effectively, as well as delivering real-time data that's reliable enough for production environments.
Databento makes it simpler and faster to access institutional-grade financial data. With Databento's APIs and no-code browser platform, users can instantly access petabytes of market data—and only pay for what they use. As an official distributor of data from over 75 trading venues, Databento's servers are hosted in each exchange's colocation facilities for low-latency, high-fidelity data capture directly from the source.
Both real-time and historical data are available in over 15 formats, including L1, L2, and L3, with nanosecond-resolution timestamps. Databento has pioneered many industry firsts, including usage-based pricing and the delivery of full-exchange L3 feeds over the internet.