Grayscale AI Compute ETF begins trading under ticker GCPU

Grayscale, the firm best known for turning Bitcoin into a Wall Street product, is now doing the same thing for raw computing power. The Grayscale AI Compute ETF started trading on September 18 under the ticker GCPU, giving investors a way to bet on the GPU capacity that powers artificial intelligence without buying a single share of Nvidia or renting a single server rack.

The fund launched at $37.76 per share with roughly $12.46 million in assets under management. That’s modest by ETF standards, but the product’s architecture is what makes it interesting: GCPU doesn’t hold equities. It holds GPU compute futures contracts.

A futures-based play on compute, not companies

The fund invests primarily in standardized futures contracts tied to GPU-based compute capacity. The strategy involves front-month contracts and roll mechanisms, meaning the fund continuously transitions from expiring contracts into newer ones.

The fund is actively managed, which gives Grayscale’s team discretion over contract selection and roll timing. That’s a meaningful distinction from passive index-tracking funds, where the strategy is mechanical.

Why compute instead of companies

Of course, futures-based products come with their own complications. Roll costs can eat into returns over time, a phenomenon commodity investors know painfully well from the era of contango-plagued oil ETFs. The performance of GCPU will depend not just on where compute prices go, but on the shape of the futures curve and how efficiently the fund navigates contract rollovers.

Grayscale’s expanding product universe

The AI Compute ETF sits under the Grayscale Funds Trust and trades on a US exchange like any traditional ETF, with standard creation and redemption mechanisms.

The $12.46 million in initial AUM puts GCPU firmly in the small-fund category. For context, ETFs typically need to cross $50-100 million in assets to be considered commercially viable over the long term.