A recently launched cryptocurrency, initiated in late May 2026, has spurred significant demand for Graphics Processing Units (GPUs). Unlike traditional cryptocurrencies that utilize SHA-256 hashing, this new digital asset requires processing matrix multiplications, a computational task fundamental to artificial intelligence models. A prominent inference provider has already integrated this technology, marking a pivotal moment for proof-of-useful-work (PoUW) entering the production phase. This development aligns with projections of a convergence between cryptocurrency mining and AI compute capabilities.
Key Takeaways
- Pearl (PRL), a new Layer-1 blockchain, employs proof-of-useful-work (PoUW) consensus. Its security relies on miners performing large-scale matrix multiplication using Nvidia GPUs, rather than traditional hashing algorithms.
- The mainnet for Pearl was launched on April 27, 2026. The project was developed by Pearl Research Labs and is based on peer-reviewed cryptographic research, distinguishing it from meme-driven cryptocurrencies.
- An exclusive partnership with Together AI has resulted in a discounted inference endpoint powered by Pearl, representing the first significant real-world validation of its PoUW model.
- The primary driver for the network’s utility is paid demand for computation. Currently, a substantial portion of the “useful work” involves inference tasks that have not yet been commercially contracted, effectively shaping the network’s proof-of-work into an AI-centric format.
- Profitability is rapidly declining. The estimated daily revenue for an RTX 5090 GPU has approximately halved within weeks due to an increase in network difficulty.
Understanding Pearl: Proof-of-Useful-Work and the MatMul Advancement
Proof-of-useful-work, an established concept aiming to replace the energy-intensive hashing of cryptocurrencies with computation yielding tangible benefits, has seen limited success until now. Pearl represents a serious advancement in this domain. Instead of SHA-256, the computational work securing the Pearl network involves matrix multiplication (matmul), a core operation in the linear algebra used extensively in AI training and inference. Pearl terms its consensus mechanism proof-of-useful-work, implemented via a process known as NoisyGEMM.
From a technical standpoint, miners execute noisy matrix multiplications on GPUs. The resulting commitments are then hashed and secured on-chain using zero-knowledge proofs for efficient verification. The underlying blockchain architecture is robust, forking from Bitcoin’s well-established btcd codebase. It is released under a permissive open-source license and includes a full node, wallet, light client, and a dedicated GPU miner. The innovation lies primarily in the work function, not the core infrastructure.
The design intentionally mirrors Bitcoin, including a maximum supply of 2.1 billion PRL (precisely 100 times Bitcoin’s 21 million supply), a decreasing block reward schedule, and block confirmation times averaging around two minutes. Pearl differentiates itself from previous PoUW attempts through its foundation in verifiable mathematics, specifically the 2025 cryptography paper “Proofs of Useful Work from Arbitrary Matrix Multiplication” by Komargodski, Schen, and Weinstein. This paper proposes a scheme with minimal overhead, estimated at approximately 1+o(1) compared to a standard matrix multiplication. This suggests that securing the blockchain adds negligible computational cost beyond the underlying matrix multiplication task itself, offering a dual output: chain security and useful computation.
Details can be found on the website : hashrateindex.com