On June 18, 2025, startup Compute Labs announced a collaboration with enterprise-level AI cloud service provider NexGen Cloud to launch a $1 million "Public Vault", tokenizing industrial-grade GPU assets for investors with an expected first-year annual return of 30%, paid in USDC stablecoin.

According to the introduction, the project will split and tokenize high-performance NVIDIA H200 GPUs, enabling investors to participate in enterprise-level AI computing infrastructure revenue sharing with an investment starting from hundreds of dollars, without directly managing hardware or bearing technical operation risks. Each H200 GPU has a market price of around $30,000, primarily used for AI model training and inference.
Compute Labs stated that this is the first time real AI computing power returns are directly returned to investors in stablecoin form, aiming to break the monopoly of giant cloud providers like AWS in AI infrastructure. Funds will be raised through NexGen Cloud's investment subsidiary InfraHub Compute and used to purchase GPU equipment, which will then be leased to data center operators with idle space, with rental income minus energy and hosting costs returned to investors.
In terms of specific operations, the project uses a token + Non-Fungible Token model for computing power ownership registration and revenue distribution, with Compute Labs uniformly collecting a 10% fee covering asset tokenization, management, and revenue processing.
Nikolay Filichkin, Business Director of Compute Labs, stated that target cooperation partners include small and medium-sized data centers, essentially "mom-and-pop shops" in the data center field. Additionally, the project has received investments from institutions such as Protocol Labs, OKX Ventures, CMS Holdings, and Amber Group.
Youlian Tzanev, Co-founder and Chief Strategy Officer of NexGen Cloud, stated that this model, by assigning a clear and tradable value to each GPU computing cycle, is expected to drive the AI market from over-speculation towards a supply-demand-driven rational pricing.




