🥛 PEARL: Belief is real, but useful work isn't (yet) ⛏️

GM. This is Milk Road, the newsletter that gets suspicious whenever a coin calls itself "the next Bitcoin”.

Here’s what we’ve got for you today:

✍️ This crypto network wants AI to mine the next Bitcoin.
🎙️ The Milk Road Show: Wall Street Has Crypto FOMO - Here's Who Could Win BIG.
🍪 SKY burns 7.37M tokens.

Prices as of 2:00 p.m. ET. Powered by CoinGecko.

THIS CRYPTO NETWORK WANTS AI TO MINE THE NEXT BITCOIN ⛏️

Pearl Network has been all over my crypto Twitter timeline for the past few days. Its native token PRL is up ~195% in the past 30 days. That’s ~7x from its July lows for a token most people have almost never heard of…so naturally I was curious to explore what the heck is going on.

Today, I’ll dive into what Pearl is, why it exists, and tackle the most important question: Is PRL…ahem…actually the next Bitcoin? (Yes, some people are saying that.)

Truth is, there are only two types of cryptocurrencies: coins that are money (like Bitcoin) or coins that make money (like Hyperliquid). PRL is building “AI native money,” meaning it fits into the first bucket alongside BTC, and that the TAM for the network is directly tied to the growth of AI, which by many accounts is in the ballpark of ~$28T-$30T.

If PRL can prove that its “Proof of useful work” algorithm can deliver on its promise, the base case over a 3-5 year time horizon puts PRL capture at at least ~1% of Bitcoin’s market cap. That’s a big “if” though, so let’s dive in.

So WTF is the Pearl Network?

Pearl is a blockchain like Bitcoin. It went live on April 27, 2026. It copied Bitcoin’s code, set a fixed max supply of 2.1B coins, and swapped out the mining puzzle. There was no pre-mine and no team or VC allocation, so every PRL comes from mining.

Bitcoin miners guess random numbers all day. That work keeps the network safe, but it's useless for anything else.

Pearl miners instead multiply huge grids of numbers (matrix multiplication), which is what AI models do.

So the pitch to users basically becomes this:

You run your AI workload on your own GPUs or rented cloud GPUs, exactly as you would anyway - but you use Pearl-modified software while doing it, allowing the work your GPU is already doing to double as a mining attempt on the Pearl blockchain.

(The end result: you run your AI work as normal → plus you get the added chance of mining some PRL.)

The trick comes from a 2025 paper by co-founders Ilan Komargodski and Omri Weinstein. Miners add random “noise” so they can’t cheat, then strip it out cheaply so the AI result is still usable.

That’s called Proof of Useful Work: the money spent securing the chain doubles as real AI work.

Bitcoin mining burns ~175 TWh of electricity a year, about as much as Sweden, and none of that work can be reused.

AI is a much bigger pool of spending. Goldman Sachs expects cumulative AI investment to hit ~$1.8T by the end of 2026.

That puts Pearl in the AI crypto sector next to Bittensor (~$3.22B market cap) and Nockchain (~$56M).

"Useful" is also where the criticism lands. In June, researcher Abhinaba Basu studied 8,012 Pearl mining machines and found that the dominant mining software had no AI code. It just multiplied random numbers.

He even mined successfully with pure junk and got 44 shares accepted by a pool, because Pearl checks that the math was done, not that it was useful.

Weinstein doesn't really dispute it. Letting miners pick their own inputs "will always allow spam mining," he says.

For most of the mining happening today, Basu is right...but Pearl is one of the rare chains with a real AI company mining it as part of an actual product, and the market has noticed. 👇

In May, Pearl partnered with Together AI, an AI cloud that runs open-source models like Llama and DeepSeek on its GPUs. Every provider serves the same open models, so they mostly compete on price.

Pearl gives Together an edge. The same GPUs that serve AI customers also mine PRL, so Together launched a Pearl-powered model priced 25% cheaper without hurting its margins.

But the catch is that PRL pays for that discount. If the price falls, the discount shrinks with it (or just disappears).

For Pearl, it’s the first proof that a real AI company will mine it as part of a product, and the best evidence so far for the “AI Bitcoin” story.

Source: Milk Road

Crypto tokens mostly get valued in one of two ways:

As money (like BTC): No cash flows. Value comes from scarcity and belief. Bitcoin has run for ~18 years, adds ~0.82% new supply a year, and is worth ~$1.7T.
As a business (like HYPE): A real product with real cash flows, plus token sinks (ways for the token to capture value) that tie business growth to token price.

So the whole play hinges on one question: will people hold PRL the way they hold Bitcoin, even while ~1M new coins hit the market every day?

Source: CoinGecko

Pearl stats at a glance:

Market cap: ~$531M.
Total mining power: ~49.2 EH/s, up from ~24 EH/s in June.
Mining overhead (how much slower AI work gets due to PRL mining): 10-15% today, 0-3% claimed after an upgrade called FP8 (no mainnet date yet).

Brian Breslow of No Limit Holdings (which disclosed a major PRL position) has written the loudest bull case.

His pitch goes something like this: Bitcoin's security is paid for by machines that can't do anything else, and they switch off when rewards stop covering the power bill.

A GPU serving AI customers already has most of its bill covered, so mining Pearl on top is close to free. Junk miners pay full price and should get squeezed out over time.

Source: Brezshares and Goldman Sachs Research

His shorthand: "AI CapEx > Bitcoin CapEx." His target is $30+, which assumes PRL reaches 1% of Bitcoin's value.

PRL has no fee burn and no revenue share, and AI customers pay in U.S. dollars. The team says that's on purpose: Komargodski says Pearl is optimizing for store of value, "unapologetically."

The AI story makes Pearl sound like a business. But as money, it still has to pass Bitcoin's tests…

Before we get to the price tag, here's the first test any money has to pass: scarcity. 👇

Source: Castle Labs, Blockchaininfo, Pearltrack

Only 343.3M PRL exists today, or 16.35% of the 2.1B max. There are no halvings. The reward shrinks a little every block instead.

That's ~1.0M new PRL a day, and by my own math from Pearl's emission formula, about +303.9M more over the next 12 months.

That's ~89% new supply in a year, vs. ~0.82% for Bitcoin. PRL may get scarce in the long run, but it's going through heavy dilution right now.

Distribution is fair on paper, but mining difficulty started very low, so ~121.7M PRL came out in the first six days.

Pearl's whitepaper says the low start was partly so the core team "had resources for research and development." Translation: the team likely mined early, but how much it holds isn't disclosed.

Security is where it gets a little wobbly. 👇

Source: PearlTrack

A mining pool is a group of miners that combine mining power and split rewards.

In the last 24 hours, Kryptex found 56.4% of blocks and PearlHash 22.1%, so two pools control 78.5%. One pool above 50% could, in theory, rewrite recent blocks.

Liquidity is thin too. Spot trading happens on just SafeTrade and BigONE with ~$7.4M of daily volume.

As one trader put it, PRL is "a race between PRL emissions and an actual perps/spot exchange listing."

At ~$1.55, PRL has a fully diluted value (what it'd be worth if every coin that will ever exist were trading today) of ~$3.245B, about the same as Bittensor's entire market cap.

The people behind it are public and academic: CEO Omri Weinstein and CTO Ilan Komargodski are computer science professors, and Cornell's Rafael Pass wrote Pearl's economics paper. Pearl is in NVIDIA's Inception program, which they claim is not an investment or partnership.

The miners are mostly regular folks: in Basu's June sample, ~96% looked like consumer graphics cards. Together AI is the only confirmed commercial "useful" miner.

So does PRL capture value generated by AI useful work?

So far the answer is no. There's no burn, no revenue share, no staking, and no voting rights. Miners also sell PRL to fund the AI discount, which adds persistent sell pressure on the PRL token.

The team is considering pricing AI services in PRL or burning some PRL per purchase, but nothing is confirmed. Even Breslow admits AI companies "could love mining Pearl and still have absolutely no interest in holding Pearl."

I believe the money framing is the right approach, but it puts all the weight on belief.

Plus there are some obvious holes in the narrative:

1. Useful work isn't proven: If junk mining stays cheaper than AI mining, "AI Bitcoin" is just Bitcoin on GPUs. FP8 is the fix, but it has no release date yet.

2. One pool runs the show: Kryptex alone finds more than half the blocks, and the team is still shipping rule changes, including one that let miners fake their mining power before it was switched off.

3. Supply nearly doubles in a year: ~1M new PRL a day has to find buyers, and most of it goes to miners who need to cover power bills.

4. You can barely trade it: Two small exchanges and ~$22K of bids near the price means big holders can't get out quietly.

5. We can't see the team's bags: Early mining and team holdings aren't disclosed, and even supply data disagrees: CoinDesk listed ~269.2M PRL circulating on Oct 2 vs. 343.3M on CoinGecko and the explorer.

So here’s my actual take

Pearl has the most interesting money story I’ve seen in AI crypto, and a real usage partner in Together AI. Over three to five years, I think PRL could reach ~1% of Bitcoin’s market cap if useful work gets proven. Over the next few months, the bigger risk is dilution: ~1M new PRL a day has to find buyers on two small exchanges.

None of this is financial advice. PRL is young, thinly traded, and high-risk.

What I’m watching next:

An FP8 mainnet date, plus outside tests that verify the ~0-3% overhead claim.
The top two pools dropping below 78% of blocks, and a second AI partner beyond Together.
A major exchange listing, and whether demand keeps up with ~1M new PRL a day.

If FP8 ships and another AI cloud plugs in, useful miners become the cheapest miners, and today’s price could look early.

If belief cools before the big listings arrive, that’s a lot of new supply for two small exchanges to absorb, and I expect a big chunk of this run to give itself back.