Is Pearl, Dreaming of Bitcoin in the AI Era, an Innovation or Just Simple Hype?

Recently, if one were to pick the hottest project in the digital asset market, it would undoubtedly be Pearl Network (PRL). At the Korea Blockchain Week (KBW) and TOKEN2049, it was evident that discussions about Pearl were prevalent among industry insiders. The same sentiment was echoed on platforms like X (formerly Twitter) and Telegram. Many people are evaluating Pearl as 'the future of AI and Bitcoin,' expressing high expectations. In fact, the price of $PRL has surged nearly sevenfold in less than two months.

Price trend of $PRL. Source: CoinMarketCap

Why is Pearl attracting so much attention? Is it merely a project that combines two popular narratives: AI and Bitcoin, or does it have the potential to carve out a new market like Hyperliquid? In this article, we will explore the core structure and operational mechanism of Pearl Network, as well as the opportunities and validation challenges that lie ahead.

1. The Growth of AI Shakes the Economics of Bitcoin Mining

The growth of the AI industry is putting pressure on the economic foundation of Bitcoin mining. Amidst halving and price declines that have worsened mining profitability, AI infrastructure businesses have emerged as new sources of revenue, prompting some mining companies to begin reallocating power and capital to the AI sector.

Bitcoin is a proof-of-work (PoW) based blockchain where miners invest power and computing resources to secure the network, receiving Bitcoin as a reward. The structure is such that mining profits must exceed costs to maintain the economic incentive for continued network participation.

In fact, the hash rate of the Bitcoin network has undergone adjustments since peaking in Q4 2025. According to the data presented, it decreased by about 20% from approximately 1,134 EH/s in October 2025 to about 908 EH/s in June 2026.

The hash rate is an indicator of the speed at which miners perform cryptographic operations per second and the total computational power of the blockchain network.

Bitcoin hash rate trend. Source: Bitcoin.com

In contrast, AI businesses among mining companies are rapidly growing. According to Castle Labs, the share of AI and high-performance computing (HPC) in the total revenue of 11 publicly listed Bitcoin mining companies rose from 1.6% in Q1 2025 to 22.3% in Q2 2026. During the same period, revenue increased from $17 million to $283 million, a rise of about 16.6 times.

The share of AI and high-performance computing (HPC) in mining companies' revenue. Source: Castle Labs

In light of these changes, digital asset analyst Ran Neuner warned that AI and Bitcoin are competing for power, and if AI gains an advantage in securing power based on higher profitability, the security of the Bitcoin network could weaken.

So, could Bitcoin mining and AI computation create two values from one computation instead of competing against each other? Pearl Network starts from this very question.

2. Pearl Connects AI Computation and Blockchain Mining

Launched in April this year, Pearl Network is a proof-of-work (PoW) based layer-1 blockchain developed by forking Bitcoin. It focuses on the limitation that the vast computing resources used in traditional Bitcoin mining are not utilized for anything other than network security. Pearl aims to combine AI computation and blockchain mining to perform both tasks simultaneously using one computing resource. This is intended to process the computations required for AI services while also securing the blockchain network, transforming the previously competitive relationship between AI and mining into a complementary one.

The key difference with Pearl lies in its mining method. While Bitcoin generates blocks through SHA-256 hash operations, Pearl utilizes matrix multiplication (MatMul), the core operation of AI models, for mining. Matrix multiplication is a fundamental mathematical operation that AI models repeatedly perform to process data and generate results. The important aspect is not just performing operations of the same kind as AI but actually utilizing computations needed for real AI services in blockchain mining. This is referred to as 'Useful Proof of Work (PoUW).'

The operational mechanism is as follows. When an AI company processes customer requests with GPUs, the AI model performs numerous matrix multiplications. Pearl combines this with additional operations and verification procedures for mining. Specifically, it adds random noise to the matrix for computation and then removes it to restore the results needed for the original AI service. Simultaneously, it hashes the intermediate results of the matrix multiplication to check if the mining conditions are met. If the conditions are satisfied, it generates a cryptographic proof, and the miner who creates a valid block after network verification receives PRL tokens as a reward.

As a result, Pearl is designed to allow AI services and blockchain mining to be performed simultaneously on a single GPU. AI companies can generate revenue by providing inference services while also earning additional income through PRL mining. Previously, computing resources had to be allocated separately for AI computations and mining, but Pearl aims to enhance resource utilization efficiency by combining the two tasks. This structure can also improve the economics of AI services, as companies can use the additional income from mining rewards to lower service prices, thereby attracting more customers at a cheaper rate.

3. Is it Innovation or Just Simple Hype?

The structure of Pearl examined so far is certainly intriguing. However, whether it is as efficient in practice as expected remains a separate question.

Are there actually AI companies and customers utilizing Pearl? Will there be performance degradation or additional costs while performing AI computations and mining simultaneously? Is the economic structure based on mining rewards sustainable in the long term? Furthermore, can the growth of the Pearl ecosystem lead to an increase in the value of PRL tokens?

In this section, we will examine these questions one by one to evaluate whether Pearl can create real value beyond a simple mining narrative.

3-1. Are there AI companies and customers actually utilizing Pearl?

Currently, the most publicly confirmed representative commercialization case of Pearl is Together AI. Together AI is a company that provides AI model inference and GPU cloud services, recognized with a corporate valuation of $8.3 billion in July 2026, receiving an investment of $800 million. At that time, it was also reported that annual recurring revenue exceeded $1.15 billion.

Together AI entered into an exclusive partnership with Pearl in May 2026 and launched the inference service 'Gemma-4-31B-it-pearl' based on Gemma 4 31B. They provided the service at more than 25% lower prices by offsetting some inference costs with PRL mining rewards. This is significant as it goes beyond a simple partnership announcement to the actual launch of a usable service. However, currently, Together AI does not support the model and is unable to utilize Pearl-based inference services.

Ultimately, Pearl has demonstrated the potential for technology commercialization through the actual launch of services, but it has not proven market competitiveness or sustainable demand. It remains to be seen how many paying customers and inference usage can be secured in the future, and whether mining rewards can continuously translate into cost competitiveness for AI services.

3-2. Is there no performance degradation?

Pearl combines additional calculations for mining with matrix operations of AI models. This process incurs additional operational costs, and if the processing speed slows down, the number of requests that can be handled by the same GPU decreases, potentially affecting the profitability of AI companies.

However, theoretically, the additional operational burden is not significant. The 2025 paper that underpins Pearl (Proofs of Useful Work from Arbitrary Matrix Multiplication) explains that the additional computational load required for mining proofs is negligible compared to standard matrix multiplication. According to Pearl, in actual mainnet operations, an additional computational burden of about 10-15% generally occurs. While this is a meaningful burden for large-scale AI services where processing speed is crucial, it may have a relatively limited impact on services with low GPU utilization or less stringent speed requirements.

The main cause of performance degradation is the difference in computational methods. The existing Pearl protocol only supports integer operations (INT8), while most AI models use floating-point operations, leading to performance loss during the conversion process. However, some models that the Pearl team has directly converted and optimized have succeeded in reducing performance degradation. The self-benchmark Llama 70B recorded about a 5% performance drop, while DeepSeek V3.2 recorded about a 4% drop.

Additionally, Pearl is pushing for an upgrade called Pearl-GEMM (Pearl-GEMM) that supports floating-point operations. According to Pearl, performance degradation has been reduced to between 0-3% in the testnet, and they plan to apply this to the mainnet in the future.

However, there are still aspects to be verified. Pearl-GEMM is described as outputting approximations instead of exact matrix multiplications, and it remains to be seen how this affects the quality of responses from actual AI models. Furthermore, it is uncertain whether the performance improvements observed in the testnet can be replicated across various AI models and in real commercial environments.

3-3. Is all mining occurring in Pearl useful for AI operations?

Pearl claims to be 'AI Bitcoin', but not all mining occurring in Pearl is utilized for actual AI services.

Pearl verifies whether matrix multiplication is performed in a predetermined manner, but it does not confirm whether the operations stem from actual AI customer requests. In other words, while the validity of the operations is confirmed, the economic utility is not assessed. Therefore, arbitrary matrix operations unrelated to AI services can also participate in mining and receive PRL rewards.

Real cases have also been confirmed. A study by Basu published in June 2026 pointed out that there was no actual AI inference code in the major mining software at the time, and mining tasks utilizing random matrices could also pass verification. However, since this study was based on specific points in time and samples, it is difficult to generalize this to the entire network situation.

There is a reason why Pearl does not directly filter out malicious mining. If the usefulness of AI outputs is evaluated by verifiers, it would require confirming whether external customer requests are made or introducing a separate trust system, making it difficult to maintain an objective and permissionless verification structure like Bitcoin. The possibility of meaningless mining is a limitation of this design.

Instead, Pearl is designed to allow actual AI service providers to gain a competitive economic advantage in mining. AI businesses can provide inference services to customers while also earning mining rewards, which can offset some of their operating costs. In contrast, miners unrelated to AI services must cover their GPU operating costs solely with mining rewards. Ultimately, the logic is that actual AI businesses gain an economic edge, while meaningless miners are naturally pushed out of the competition. To achieve this, Pearl is pursuing upgrades to reduce the additional operational burden from the current 10-15% to 0-3% and to secure actual AI customers.

However, for this structure to work, two issues must be resolved.

The first is the cost issue. Miners using general graphics cards who have already recouped their equipment purchase costs can continue mining with low operating costs, potentially putting them at an advantage over actual AI businesses. In contrast, AI businesses may incur additional opportunity costs, such as reduced inference throughput, during the process of combining mining.

The second and more fundamental issue is the lack of actual AI demand. AI businesses can utilize the amount of computation requested by customers for mining, but miners performing meaningless computations can add more equipment as long as profitability is maintained, regardless of customer demand. Therefore, if there is not enough actual AI demand, mining rewards that are not filled with useful computations can be claimed by pure miners.

Ultimately, reducing the proportion of meaningless mining is a core challenge that Pearl must address. In the future, it will be important to observe how much the proportion of useful computations can be expanded in overall mining through technological upgrades that enhance the mining competitiveness of actual AI businesses while securing sufficient AI customers.

3-4. Does the increase in Pearl usage also raise the value of $PRL?

The growth of Pearl's AI services does not necessarily mean that the value of the PRL token will also increase. Currently, AI service providers charge customers in dollars and receive PRL mining rewards separately. The revenue from services does not directly translate into the purchase or burning of PRL, and instead, miners can sell the PRL they receive as rewards to realize profits.

In other words, AI companies can enhance their profitability by utilizing Pearl, but this does not automatically imply an increase in demand for PRL. Customers do not need to purchase or hold PRL to use AI services.

Additionally, there is a burden associated with the new token supply. The total supply limit of PRL is 2.1 billion, and as of October 2026, approximately 346 million have been mined. This represents about 16% of the total supply, meaning that 84% will be issued in the future. If demand corresponding to the new supply does not arise, this could exert significant pressure on the token price.

Ultimately, Pearl's core challenge is to connect the growth of AI services to the actual demand for PRL. In the future, it will be important to observe how the introduction of a computing marketplace and payment/burning structures translates into the purchasing, holding, and burning demand for tokens as AI usage increases.

4. Capturing Actual AI Demand and Token Value is Key

So far, we have examined the structure and key challenges of Pearl. Pearl is a project that aims to create two economic values from one computation by combining traditional blockchain mining with AI computations. It is a meaningful attempt to support the growth of both the AI industry and blockchain instead of competing for resources.

However, there are still several challenges that need to be addressed for sustainable growth. Not only is there the issue of performance degradation, but actual AI demand is still limited, and the structure connecting the growth of AI services to the value of PRL tokens has not been sufficiently established. While short-term price increases may occur due to market expectations, the actual demand and economic foundation needed to support this in the long term have yet to be validated.

The positive aspect is that Pearl is pursuing technological upgrades, expanding collaborations with AI companies, and building a computing marketplace to address these issues. In the future, the key will be how many actual AI customers and usage can be secured and how this can be connected to sustainable demand for PRL.

In conclusion, it is still difficult to determine whether Pearl is a project that will create new innovations through the combination of AI and blockchain or simply a project that is receiving attention due to market expectations. Personally, I believe that the ecosystem remains overly in its early stages compared to the current high interest and expectations. While the technological potential is intriguing, there is still much to prove in terms of actual AI demand, business performance, and the structure for capturing token value. Ultimately, for Pearl to justify market expectations, it must demonstrate actual demand and economic value beyond technological potential. Until then, I think it is necessary to view Pearl's long-term growth potential somewhat conservatively.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.