Author: CoinW Research Institute
Abstract
Pearl Research (PRL) has risen more than 60% in the past 7 days and over 250% in the past 30 days, with its circulating market capitalization now climbing to approximately $650 million, making it a popular project in the recent AI crypto sector. Through its Proof of Useful Work (PoUW) mechanism, Pearl attempts to enable GPUs to participate in blockchain mining while performing AI computations, achieving dual returns from AI service revenue and PRL rewards. Its collaboration with Together AI has preliminarily validated its commercial application potential, but the proportion of genuinely paid computation, business revenue, and token value capture mechanism still remain to be verified. Meanwhile, PRL's current circulating ratio is only about 16.4%, and continuous mining emissions may bring selling pressure. The short-term price action reflects the market's optimistic expectations for AI compute mining, while long-term value depends on whether real AI demand can continue to grow and support the current valuation.
1. PRL Surges Over 250% in 30 Days, AI Compute Mining Becomes a New Market Hotspot
Recently, Pearl Research (PRL) has become one of the standout performers in the AI crypto sector. According to CoinGecko data, as of October 14, PRL was priced at approximately $1.95, up about 64% over the past 7 days and nearly 261% over the past 30 days, with a circulating market capitalization of approximately $650 million. There is no single event that can fully explain this rally; the market's focus is on the novel compute utilization model proposed by Pearl. Traditional cryptocurrency mining requires consuming large amounts of computational resources to maintain the blockchain network, whereas Pearl aims to have GPUs participate in mining while running AI models and earn PRL token rewards. This mechanism is called Proof of Useful Work (PoUW). Simply put, the same set of GPU equipment has the opportunity to simultaneously earn AI computing service revenue and blockchain mining rewards. Pearl therefore positions itself as "the Bitcoin of the AI era."
Behind PRL's recent rally, in addition to the rising narrative of AI compute mining, new technological and market catalysts have also emerged. Since September, Pearl has launched the Pearl Inference service and advanced floating-point computation upgrades, planning to adapt to a broader range of AI models and high-performance GPUs through computation formats such as FP8, bringing the PoUW mechanism closer to actual AI inference demands. At the same time, Pearl has also begun extending toward verifiable AI inference, attempting to provide verifiable proofs for AI computation processes. On October 7, Pearl held its first Pearl Connect event during TOKEN2049, supported by No Limit Holdings, further increasing the project's exposure among crypto investors. These developments, combined with the low-cost inference service previously launched in collaboration with Together AI, have led the market to begin assessing Pearl's potential to evolve from GPU mining toward AI computing infrastructure. However, the technological upgrade is still underway, and real commercial demand has not yet been fully verified. At the same time, PRL's low circulating ratio may also amplify price movements. Currently, approximately 344 million PRL are in circulation, accounting for only 16.4% of the 2.1 billion maximum supply. With relatively limited circulating tokens, concentrated buying could drive rapid price increases, but the continuously increasing mining supply in the future will also bring potential selling pressure. PRL's recent strong performance reflects the market's optimistic expectations for AI compute mining, but the nearly $4 billion FDV also implies a high valuation threshold. Whether Pearl can translate technological innovation into sustained growth in real business will determine whether market expectations can be supported.
2. How Pearl Enables GPUs to Participate in Mining While Running AI
To understand Pearl, one can start with Bitcoin mining. Bitcoin miners need to invest in large amounts of computing equipment, continuously performing hash operations to compete for the right to record new blocks and earn rewards. This mechanism is called Proof of Work (PoW), which can secure the blockchain network, but the computation results are usually difficult to directly use in other commercial scenarios. Mining machines consume large amounts of electricity, and their main output is network security. Pearl aims to change this. It is a Layer 1 public blockchain adopting the PoUW mechanism, replacing the hash operations in traditional mining with matrix multiplication, which is closer to AI workloads. Matrix multiplication is one of the fundamental operations in AI training and inference. Whether it is having a large model answer questions or generate images, GPUs are required to perform large amounts of similar computations behind the scenes. Through a special algorithm, Pearl enables GPUs to simultaneously generate proofs of work that can be verified by the blockchain while performing these operations.
For GPU operators, equipment that previously could only rely on charging for AI computing services now has the opportunity to add a stream of PRL mining income. If the additional revenue exceeds the extra computational cost, operators may achieve better profit margins or offer cheaper AI services to customers. In terms of project background, Pearl is being developed by Pearl Research Labs, with a core team including Omri Weinstein and Ilan Komargodski, who have research backgrounds in algorithms, cryptography, and computer science. The project chooses to start from the underlying consensus mechanism, attempting to solve the problem of the single-purpose use of computational resources in traditional PoW. Pearl's technological innovation is to make blockchain mining and AI computation compatible. This creates the possibility of dual income for GPUs and also differentiates it from ordinary GPU rental platforms.
FLOP Network shares a similar approach to Pearl. Both aim to combine GPU compute with blockchain rewards, enabling AI computation to generate additional economic value, but their implementation paths differ. Pearl adopts Proof of Useful Work (PoUW), introducing matrix multiplication into the underlying mining mechanism. Even if the computational tasks have no real paying customers, miners may still earn PRL rewards by completing the prescribed operations. FLOP, on the other hand, adopts Proof of Useful Inference (PoUI), allocating compute around real inference requests from AI Agents. After miners complete tasks and submit verification proofs, they receive FLOP payments and block rewards. Pearl places more emphasis on making AI computation compatible with blockchain mining, while FLOP attempts to directly bind real inference demand to token payments and miner rewards. Therefore, Pearl needs to prove that mining compute can continuously serve real AI business, while FLOP needs to verify the paid demand of AI Agents, the inference result verification mechanism, and whether the compute market can operate effectively. However, the mechanism currently disclosed by FLOP is still mainly in the white paper design and development stage, and cannot yet be directly compared with Pearl's already-launched network and commercial partnerships in terms of actual results.
Figure 1. Comparison of PoW and Pearl PoUW Mechanisms. Source: CoinW Research Institute
3. Partnering with Together AI, Pearl Uses Mining Revenue to Lower AI Inference Costs
Whether a new mining mechanism can truly create value ultimately depends on whether anyone is willing to pay for computing services. Pearl's collaboration with AI cloud service provider Together AI provides an early commercialization case. In May 2026, Together AI launched a Gemma model service combined with Pearl's mining mechanism. When users call the AI model, the backend GPUs process inference requests while participating in Pearl mining, with the opportunity to earn PRL block rewards. Together AI stated that by using expected mining revenue to subsidize computing costs, the price discount for the relevant model API exceeds 25%. The appeal of this model is straightforward. Traditional AI cloud platforms need to rely on fees paid by customers to cover GPU procurement, electricity, and operations and maintenance costs, while after integrating Pearl, the same set of GPUs has the opportunity to add a stream of token revenue. As long as mining revenue exceeds the additional cost, service providers can lower inference prices to attract customers or increase their own profits. It should be noted that the 25% price discount comes from token revenue subsidies and does not mean that the actual GPU computing cost has decreased by 25%.Pearl hopes to create a growth loop from this. Lower inference prices attract more customers, real demand growth brings in more GPUs, and an expanded network of computing power helps improve blockchain security. If ecosystem development further drives demand for PRL, miner revenue may also increase, attracting more hashpower to join. But whether this loop can be sustained still depends on the scale of real business and mining revenue. When the PRL price rises, service providers may gain more room for subsidies; if the coin price falls and mining difficulty increases, the reward obtainable per unit of hashpower may shrink, and the price advantage will weaken accordingly. The partnership with Together AI shows that Pearl has found a path to embed mining rewards into real AI services, but it is still far from forming a mature commercial closed loop. In the future, more paying customers, stable inference requests, and sustainable service revenue will be needed to prove that this model has long-term competitiveness.
Figure 2. Pearl's dual-revenue business flywheel. Source: CoinW Research Institute
4. The PoUW mechanism sparks controversy; AI computation may just be another form of mining
Pearl's most attractive aspect is its hope of making mining hashpower truly serve the AI industry. But this also gives rise to the project's biggest controversy. That miners complete matrix operations commonly used in AI does not mean these computations actually come from AI customers, still less that anyone is willing to pay for them. If the matrix computations performed by GPUs come from real AI model requests, they may generate service revenue; if they are merely tasks generated by miners themselves to obtain PRL rewards, then in essence it is still mining for the sake of mining.
An independent paper studying Pearl's cuPOW protocol pointed out this problem. The researchers found that miners can randomly generate matrices and submit valid proof of work without running a complete AI model. Some mining software does not even include AI inference code, yet can still participate in mining and receive rewards. This means that what Pearl currently mainly verifies is whether miners have completed the prescribed computations, not whether those computations have been used by real customers.
Figure 3. The difference between real AI computation and synthetic mining. Source: CoinW Research Institute
Of course, this does not mean Pearl's technological innovation is without value. For PoW public chains, computations that meet security requirements and can be efficiently verified can be used to maintain network security. Pearl's breakthrough lies in making mining computation compatible with AI workloads, creating the conditions for real inference tasks to be connected to mining. But being able to connect to real business and having all network hashpower create commercial value are still two different things. If a large number of miners simply generate matrix tasks on their own to earn PRL, Pearl's economic model will still be close to traditional mining. The network relies on token issuance to incentivize miners, and miners may sell rewards to pay for equipment and electricity costs. Only when more computation comes from paid requests by enterprises and developers can the dual-revenue model truly leverage its advantages. Therefore, Pearl cannot be measured only by coin price, network hashpower, and the number of miners, but more importantly by how much hashpower actually generates revenue. The project needs to further disclose the proportion of real inference tasks in all mining work, as well as data such as paying customers, model call volume, and service revenue. Otherwise, it is difficult for outsiders to judge whether network growth comes from AI demand or from mining hashpower rushing in after the coin price rises.
In addition, even if the real business can grow, Pearl still needs to compete with traditional cloud platforms in inference speed, service stability, and long-term price advantages. If it loses competitiveness after reducing PRL subsidies, the sustainability of its business model will be challenged. Pearl has already demonstrated a technical path for AI computation to participate in blockchain mining, but it still needs to prove that these computations can continuously meet real customer demand. Without a stable paid business, even a massive hashpower network may simply be a traditional mining farm using a different form of computation.
5. PRL adds about 1 million tokens per day, and continuous mining sell pressure tests a $4 billion valuation
Pearl's commercial prospects ultimately still need to be tested by the token market. Even if the real AI business can grow, investors still need to understand how PRL is issued, how much new supply there will be in the future, and whether business value can truly be transmitted to the token. PRL has a maximum supply of 2.1 billion tokens, with no team pre-mine or VC reserved allocation common in traditional crypto projects. Tokens are mainly issued gradually through mining. Unlike Bitcoin's halving about every four years, PRL adopts a smoothly declining issuance mechanism, with block rewards gradually decreasing over time. According to the official design, about half of the token supply will be released after about four years. This arrangement reduces the risk of concentrated unlocking by the team and institutions, but the absence of unlocking pressure does not mean there is no selling pressure. Miners need to pay for GPUs, electricity, and operations and maintenance costs, so they may continuously sell PRL to sustain operations. According to estimates by third-party platforms, about 1 million new PRL tokens are currently issued each day. At $1.8, that is equivalent to about $1.8 million in newly issued token value per day. Although new issuance does not mean all of it is sold immediately, if market buying demand is insufficient, the continuously released supply may still weigh on the coin price.
Figure 4. PRL supply growth and block reward decline curve. Source: CoinW Research Institute
As of October 14, PRL's circulating supply was about 344 million tokens, accounting for 16.4% of the maximum supply, with a circulating market cap of about $650 million and an FDV close to $4 billion. The gap between the two reflects the difference between the current circulating supply and the future maximum supply scale. A relatively low circulating ratio may amplify short-term gains, but as more tokens enter the market, the existing price will also face continuous tests. In addition, the absence of a pre-mine does not mean the chips are naturally dispersed. In the early days of the Pearl mainnet launch, mining difficulty was relatively low, and some early miners may have accumulated large amounts of PRL at low cost. Therefore, early miner holdings, exchange inflows, and mining pool hashpower concentration are all important indicators for judging potential selling pressure.
More noteworthy is PRL's value capture mechanism. At present, PRL mainly serves the functions of block rewards and network incentives, and is not a computing power voucher that can be directly redeemed for GPU usage time. Inference fees paid by AI customers will not automatically translate into PRL buying demand. Even if Pearl serves more AI customers in the future, the token still needs to establish a clearer source of demand in order to continuously share in the economic value brought by ecosystem growth. Therefore, an FDV close to $4 billion actually contains two layers of expectations. On the one hand, the market expects Pearl to expand its real AI computing business; on the other hand, it also expects these businesses to ultimately support long-term demand for PRL. If business growth cannot be translated into token value while mining supply continues to increase, the current valuation may face repricing.
6. Pearl's commercialization potential and valuation risk
Pearl already has noteworthy technological differentiation and preliminary commercialization cases, but current market pricing is clearly ahead of verifiable business results. PRL's short-term strong rise reflects investors' optimistic expectations for AI computing power mining, while an FDV close to $4 billion also means the project needs to deliver growth faster. In the future, two directions should be closely watched. On the commercial side, verify whether the proportion of real paid inference, customer growth, and service revenue can continue to increase; on the token side, observe new mining supply, changes in miner holdings, and whether PRL can establish a clearer value capture mechanism. If real business growth can gradually reduce dependence on token subsidies and form stable token demand, Pearl's long-term valuation may gain stronger support. From an investment perspective, PRL combines the growth potential of an emerging sector with the volatility risk brought by low circulation and high valuation. Especially after a sharp short-term rise, market expectations being front-loaded, continuous miner selling, and insufficient liquidity may all trigger price adjustments. Pearl has already proven that it can attract market attention. What it needs to prove in the next stage is whether commercial revenue and token value can truly keep up with market expectations.