Compute Labs

Compute Labs is a Solana-native protocol that transforms enterprise-grade GPU hardware into yield-bearing digital assets, giving retail investors direct financial exposure to AI compute infrastructure that was previously accessible only to institutional buyers. Founded in early 2024 and headquartered between Los Angeles and Seattle, the project builds what it calls the "AI-Fi" category: a merger of real-world asset tokenization and decentralized finance applied specifically to the compute economy powering large-scale AI model training and inference.

The Problem Compute Labs Addresses

Access to high-performance computing infrastructure — NVIDIA H100s, H200s, and newer B-series GPUs — has historically been limited to hyperscalers and well-capitalized data centers. These machines generate substantial revenue by leasing compute capacity to AI companies running workloads, but individual investors have had no direct path to capturing that income stream without either buying physical hardware outright or investing indirectly through public equities. Compute Labs argues that tokenization solves this by putting fractional GPU ownership on-chain and routing lease revenue directly to token holders.

Compute Tokenization Protocol

The core technical product is the Compute Tokenization Protocol (CTP), which converts physical GPUs deployed in partner data centers into two types of on-chain tokens: GNFTs and miniGPU tokens.

GNFTs (GPU Non-Fungible Tokens) represent full ownership of a specific deployed GPU unit. They are implemented as SPL22 tokens on Solana — a hybrid token standard that can behave as both non-fungible (uniquely identifying one physical machine) and fungible (tradeable in secondary markets). Each GNFT carries ownership rights and a direct claim on that GPU's leasing revenue.

miniGPU tokens are the fractional equivalent, issued at a 100:1 ratio to GNFTs by default (though vaults may adjust this ratio). They allow smaller investors to hold a proportional slice of a GPU's income without committing the capital required for a full unit. Holders of a GNFT can split it into miniGPUs through an on-chain Solana program that mints the fractional tokens while locking the underlying GNFT, making the two forms interconvertible.

Revenue is generated by leasing the underlying GPU capacity to companies running AI workloads. Income from those leases is distributed to GNFT and miniGPU holders in USDC, creating a transparent, performance-linked yield stream. Compute Labs has cited 20–50% annual yields depending on GPU utilization rates, though actual returns vary with market demand for compute.

Token metadata — including GPU model, specifications, and deployment details — is stored on-chain, while associated images and provenance records are stored on Arweave for decentralized permanence. Both token types use Solana's SPL Token Program and Metaplex's Token Metadata Program, benefiting from Solana's sub-second finality and low transaction fees for yield distribution and secondary trading.

GPU RWA Vaults

The primary investment vehicle is the GPU RWA Vault: a structured pool of deployed GPUs that accepts investor capital during a sale period and issues corresponding GNFTs or miniGPUs. The inaugural vault was backed by NVIDIA H200 GPUs running live AI workloads, with a $1 million initial deployment. Subsequent vaults targeted the B200 and GB200 (NVL) series as NVIDIA's more recent hardware generations became available. The vault structure handles GPU procurement, data center deployment, compliance, and operational management on behalf of investors, leaving token holders with a passive yield position.

Native Token

Compute Labs planned the launch of a native fungible token (referred to as CLAB or COMPUTE in public materials) for Q3 or Q4 2024. This token was designed to facilitate ecosystem transactions — including purchasing and renting compute resources within the protocol — and to serve as a governance and incentive layer. Final tokenomics and launch parameters were not fully documented in available public sources at the time of research.

Funding and Investors

In April 2024, Compute Labs closed a $3 million pre-seed round at a $30 million fully diluted token valuation, announced publicly in July 2024. The round was oversubscribed by approximately 2x and structured as a Simple Agreement for Future Tokens (SAFT). Protocol Labs — the organization behind the Filecoin decentralized storage network — led the round.

Additional institutional participants included OKX Ventures, HashKey Capital, Blockchain Coinvestors, CMS Holdings, Amber Group, P2 Ventures (formerly Polygon Ventures), and MH Ventures. The angel investor cohort added considerable ecosystem credibility: Sandeep Nailwal (co-founder of Polygon), Illia Polosukhin (co-founder of NEAR Protocol, also a co-creator of the Transformer architecture underlying modern large language models), and Austin Federa (then Head of Strategy at the Solana Foundation) all participated. Following the pre-seed close, Compute Labs was in the process of raising a seed round at a materially higher valuation.

NVIDIA Inception VC Alliance

Compute Labs is incubated by NVIDIA's Inception VC Alliance program, which provides early-stage AI startups with discounted GPU supply, go-to-market support, and direct access to NVIDIA business units and original equipment manufacturers. This relationship is operationally significant for a protocol whose core product requires reliable access to enterprise GPUs at competitive prices.

Team

The company was co-founded by CEO Albert Z. and CTO Xingfan Xia. Xia has written publicly on the technical implications of open-source AI developments, including the DeepSeek model release in early 2025. The team numbered approximately five people at founding, with pre-seed proceeds earmarked for expanding the core development and operations staff.

Position in the Solana Ecosystem

Compute Labs occupies a distinct niche at the intersection of Solana's real-world asset (RWA) sector and its broader DeFi ecosystem. By building on Solana, the team gains access to fast settlement for USDC yield distributions, a liquid secondary market for SPL tokens, and a DeFi composability layer that can integrate GNFT and miniGPU tokens into lending protocols, yield aggregators, and liquidity pools over time.

The project is frequently mentioned alongside io.net and Aethir as part of a wave of decentralized compute infrastructure projects that emerged in 2024. Compute Labs differentiates on product design: where io.net aggregates spare GPU capacity into compute marketplaces for buyers, Compute Labs focuses on the investor side — creating yield-bearing ownership instruments backed by dedicated, enterprise-grade hardware running contracted workloads.

Longer-term, Compute Labs has indicated plans for multi-chain expansion to NEAR Protocol and Monad, and intentions to extend the tokenization model beyond GPUs to quantum and edge computing hardware as those markets develop.