Silicon Data Launches First GPU Cost Index for AI

Silicon Data, a company that specializes in GPU market analytics, has released the first daily index to track the cost of the chips that are critical to AI. The index is called the Silicon Data H100 Rental Index. Founder and CEO Carmen Lee said the H100 Rental Index tracks hourly GPU costs, which she says is a first step toward transparency in a complex market. GPUs are key to training AI models because they can process thousands of tasks simultaneously. They can cost thousands of dollars apiece, and data centers use hundreds or even thousands of them.

Founded in 2024 and backed by DRW and Jump Trading, Silicon Data is developing a platform to standardize GPU prices with the goal of turning them into a tradable asset class. The index is compiled from data from multiple sources, including major cloud providers, and is normalized by GPU type, platform, and hardware configuration. Prices vary by brand, performance, and software specifications.

The index’s primary users are banks, hedge funds, and investment firms that finance the construction of GPU-powered data centers. While the index itself is not a tradable instrument, it serves as a benchmark, similar to stock indices, to help traders and AI developers assess the cost of computing resources. Lee emphasizes that the goal of the project is to create a transparent market and help participants understand the true cost of infrastructure.