Pearl and its token PRL became one of the projects discussed heavily by overseas attendees during TOKEN2049, according to PANews. The attention comes as some Layer 2 projects, including Blast and Abstract, were described in the source as shutting down in succession, while new proof-of-work names have started to appear.
Pearl is trying to answer a long-standing question: can AI compute produce something like a Bitcoin-style currency while doing useful work for customers at the same time?
What Pearl is building
Pearl is an independent PoW blockchain that went live on April 27, 2026. It uses Proof of Useful Work, or PoUW. Instead of asking miners to perform SHA-256 hashing in the way Bitcoin does, Pearl turns mining into matrix multiplication, a type of computation used widely by large AI models.
The idea is that if an AI company is already using GPUs to run inference for customers, that same batch of compute can also participate in Pearl mining and earn PRL rewards. In the project’s ideal setup, one GPU can generate revenue from AI customers and from mining at the same time. That could lift miner income, while AI service providers could use part of the added revenue to lower prices.
Academic pedigree is a key part of the pitch
The source describes the team as one of Pearl’s strongest points. CEO Omri Weinstein holds a PhD in computer science from Princeton University and is an associate professor at the Hebrew University. He previously taught at Columbia University, and his background includes NVIDIA and VAST Data.
Co-founder Ilan Komargodski is an associate professor of cryptography at the Hebrew University, with a doctorate from the Weizmann Institute of Science. He previously conducted research at Cornell Tech and NTT Research. A third co-founder, Idan Sugarman, comes from a business and investment background. Cornell cryptography researcher Rafael Pass is also involved in work on economic security.
PANews said the core team has spent years studying computational complexity, cryptography, and verifiable computation, and published the original PoUW paper in 2025. In the source’s framing, that makes the technical direction look less like a quick attempt to chase the AI theme.
Commercial progress exists, but core business metrics are still missing
On May 15 this year, Together AI officially announced a collaboration with Pearl Research Labs to launch Gemma-4-31B-it-Pearl inference service. According to the announcement cited in the source, mining subsidies reduced inference pricing for that model by more than 25%.
Pearl has also built its own inference platform compatible with the OpenAI API. It supports open models including GLM and DeepSeek, allows users to top up in U.S. dollars, and charges by token consumption.
Even so, the source says the project still lacks several pieces of data needed to show the business is working at scale: real paid revenue, daily inference volume, and customer retention. For now, those gaps leave the commercial case unproven.
Who is backing PRL
No Limit Holdings, or NLH, has publicly said it is actively accumulating PRL. One of its investors, Brian Breslow, published a lengthy research note in September calling Pearl a potential next-generation Bitcoin and giving a long-term target above $30.
The source also notes that Gin Chao, a central figure at NLH, previously served as Binance’s chief strategy officer and is also a BinanceUS board member. PANews drew a line here: that background suggests PRL is attracting money with meaningful industry ties, but it should not be taken as evidence that Binance invested in or endorsed Pearl.
Token distribution and market structure
PRL did not follow a traditional venture capital token round or a pre-mined allocation model, according to the source. Early supply was generated mainly through GPU mining, then circulated through third-party over-the-counter venues including Pearl OTC and Black Pearl OTC. Pearl OTC says on its website that cumulative trading volume is about $31.5 million. Spot trading has also appeared on exchanges including BigONE and SafeTrade.
The token has a maximum supply of 2.1 billion. The source says there is no conventional team pre-mine or VC unlock schedule. Issuance declines smoothly through block rewards rather than through Bitcoin-style halvings every four years.
That structure does not remove all concerns. Early mining difficulty was very low, and about 121.7 million PRL were mined in roughly the first 5.9 days, equal to around 36% of current circulating supply. The source says it is not transparent how much of that went to the team and how much went to early miners.
Using a reference price of about $1.28 on Oct. 8, the article puts circulating market capitalization at roughly $430 million and fully diluted valuation near $2.7 billion. It also says theoretical new issuance is still around 1 million PRL per day, leaving ongoing miner selling pressure as a factor that cannot be ignored.
A central technical dispute
One of the biggest technical debates is that Pearl can prove miners completed matrix computations, but it cannot prove those computations came from real paying AI customers.
The source says research published in June, based on observations of miner software and actual mining behavior, found that a large share of the computation at that time could be completed using random matrices. In other words, those workloads did not require anyone to buy inference service.
The project’s argument is that miners serving real AI customers will eventually have two revenue streams, customer payments and token rewards, while pure miners will only have token rewards. As competition rises, the cost advantage of the second group should fade. PANews says that economic logic is reasonable, but not yet backed by large-scale real commercial data.
The report also points to a Mixture of Experts, or MoE, mining vulnerability that appeared in July as another reminder that the consensus implementation is still early. An FP8 upgrade meant to support a broader range of AI workloads still needs to be proven on mainnet.
Can AI demand translate into token demand?
At present, Pearl’s inference platform mainly charges in dollars. Miners receive customer payments and PRL rewards, but the token itself does not yet have a confirmed large-scale revenue-sharing model, nor a mandatory buyback-and-burn mechanism, according to the source.
If the business grows, that could improve miner economics and strengthen network security and market confidence. But whether AI demand can turn into direct buying demand for PRL depends on what comes next: actual token-based payments, token-denominated settlement, staking use cases, or reserve demand. Without that, even if inference revenue grows, token price could still be diluted by ongoing mining issuance.
PoW plus AI as a post-L2 narrative
The PANews piece argues that the wave of L2 shutdowns shows the market does not need so many chains built around repeated ideas, while projects like Pearl are trying to find a new route for PoW. In that sense, PoW plus AI may become one of the narratives to watch next.
What decides how far PRL can go, though, is whether AI companies are willing to keep paying for the compute over the long run. Making AI compute generate a token is one technical achievement. Making that token an asset the market actually needs to hold is a separate business challenge.


