PEARL public mining pools cross 4.7 EH/s — the inflection point, May 2026
In the last 24 hours, the public mining pools on the PEARL Proof-of-Useful-Work network moved from a single-exahash story into a two-pool exahash race. Total reported hashrate across all four public pools just crossed 4.7 EH/s, combined active miners cleared 23,000, and a network that needed eighteen months to get its first pool to 1 EH/s just added the second one in less than a week. This is the inflection point we've been waiting for.
Here is what's actually on the table, why the curve is now compounding rather than incrementing, and what it means if you mine, hold, or build on PEARL.
The numbers that changed
Live snapshot from lordofpearls.xyz/pools, ordered by hashrate:
| Pool | Hashrate | Active miners | Blocks (24h) | Fee |
|---|---|---|---|---|
| 🐚 PearlHash | 2.84 EH/s | 12,843 | 167 | 3% |
| ⛏️ AlphaMine | 1.42 EH/s | 9,447 | 108 | 5% + 1% dev |
| 🛡️ Akoya Pool | 395 PH/s | 1,247 | 25 | 5% |
| 💎 MinePRL | 81 PH/s | 290 | 13 | 4.4% |
| Total | ~4.74 EH/s | 23,827 | 313 | — |
For context: forty-eight hours ago, the same table showed two pools at exahash combined — total public hashrate around 1.8 EH/s, and only PearlHash had personally crossed 1 EH/s. As of this morning, each of the top two pools individually exceeds 1 EH/s. The network just doubled.
The miner-count change is even more striking. AlphaMine added roughly 5,700 active miners in 24 hours. PearlHash added 4,600+ in the same window. Numbers like that don't come from speculation — they come from real GPUs being pointed at a real workload, with real economics that close.
What's driving the curve
Three independent factors landed in the same week, and they're stacking.
1. The Pearl Research × Together AI partnership
On May 15, Together AI — one of the top three serverless AI inference providers in the world — launched Gemma-4-31B-it-Pearl as a production inference endpoint powered by the PEARL PoUW protocol (our full deep-dive on the deal). For the first time, paying AI inference traffic now routes through PEARL-mining hardware as a byproduct. That changes the economic floor under mining — and the smart money noticed within days.
2. The PearlHash fee cut + exahash milestone
PearlHash crossed 1 EH/s for the first time on May 19, then dropped their pool fee from 10% to 3% on May 20 — making the largest pool simultaneously the cheapest on the network. We covered the move and its implications in PearlHash crosses 1 EH/s — the first PEARL pool to hit exahash scale. The combination — largest + cheapest — created an asymmetric draw that pulled hashrate onto the pool faster than anyone expected.
The follow-on effect: AlphaMine had to respond. Their fee landscape now sits at 5% + a 1% dev fee, with PPLNS and a solo port, and they've added 5,700+ miners to keep their share of the network from collapsing. Competition is now real in a way it wasn't two weeks ago.
3. The macro AI-compute environment
The AI infrastructure market is on track to spend roughly $400 billion on data centers in 2026 alone. The single most important cost line for every AI startup is GPU compute. PoUW pitches a structural answer to that: every dollar of inference compute can simultaneously produce blockchain security as a byproduct. Until last week, that was theoretical. As of this week, with Together AI live and two exahash-scale pools in production, it's a live experiment with measurable throughput.
Miners who were on the fence in April are no longer on the fence in May. The thesis no longer requires belief.
What this means if you mine PEARL
Three concrete signals, ranked by how much they should actually change what you do this week.
1. Capacity is being absorbed faster than it's being added
Block production is up to 313 blocks in 24 hours across the four public pools — close to the network's protocol-target rate. As more hashrate joins, the protocol auto-adjusts difficulty upward, so each block becomes harder to find. The window in which the current per-block reward economics are favorable is shrinking, but it isn't closed yet — and current network-difficulty levels still leave room for new entrants.
Practical implication: if you've been planning to spin up a rig, now is materially better than two weeks from now. Every additional EH/s that lands on the network erodes the per-block math.
2. Pool choice now matters more than it used to
When there was one large pool, where you mined was largely a question of preference. With two exahash pools competing on different fee structures (PearlHash 3% flat, AlphaMine 5% + 1% dev), the take-home gap between pools is now meaningful enough to matter on bigger rigs. Run the math on your specific hashrate — the profit calculator is the fastest way.
For the live comparison, /pools updates hashrate every ten minutes and on-chain-verifies the 24h block count for each pool.
3. Hardware repurposability is the floor under your downside
The chain just moved from "this might work" to "this is working." But PEARL is still a young chain, and its price will be volatile. The reason mining PEARL on an H100/H200 carries less risk than mining Bitcoin on an ASIC is that the hardware retains independent value — render farms, AI startup training, scientific computing all bid for the same chips. Hardware-choice guide. If you don't own the GPU yet, the cleanest test-run remains RunPod — per-second billing on H100/H200 with a signup credit that covers the first session.
What this means if you hold PEARL
The thesis-validation argument is getting stronger every week, but the price tape doesn't track the fundamentals tick-by-tick. What's quietly happening underneath the price:
- Real-world inference revenue is starting to flow through the protocol via Together AI. That's a floor under the protocol's economic productivity that didn't exist two months ago.
- Capital is committing — the only way to put 4.7 EH/s of hashrate on the network is to deploy hundreds of millions of dollars worth of GPU hardware. Operators don't do that on speculation; they do it when the unit economics close.
- Concentration risk is real — PearlHash alone holds about 60% of public-pool hashrate. That deserves attention. The protocol's decentralization will need at least one or two more pools at exahash scale for healthy long-term distribution, and the current curve suggests that's where it's heading.
The bear case has moved from "PoUW won't work" to "this might be priced in already." Those are very different bear cases — the first has been killed; the second is a normal market debate.
💹 Liquidity is opening too. If you're holding PRL and want a place to trade or cash out without leaving the ecosystem, we're launching a Telegram-native OTC desk for instant PRL ↔ USDC swaps. 1.8% fee, 2FA, no KYC, 15% lifetime referrals. Get the launch DM the moment the order book goes live — sign up at lordofpearls.xyz/otc-soon.
What this means if you build on PEARL
If you're an AI startup or infrastructure team:
- The
Gemma-4-31B-it-Pearlendpoint on Together AI is now backed by real production-scale hashrate. Inference latency, throughput, and reliability are no longer constrained by mining capacity. - The 2-for-1 economics pitch (useful inference now + PEARL coins for future discount) now has visible production data behind it. Worth a real evaluation against your existing inference vendors.
- Other inference platforms will follow Together AI within the next ninety days — that's the historical pattern. If PEARL's PoUW endpoint becomes a category, you want to be on the early integration curve, not the catch-up curve.
The honest qualifiers
Two things worth being clear-eyed about.
One pool dominates. PearlHash at 60% of public-pool hashrate is healthy operationally but unhealthy for long-term protocol decentralization. The next ninety days will tell us whether AlphaMine can hold its current ~30% share, whether new pools enter, or whether the curve flattens into a one-dominant-pool equilibrium. The first of those is best; the third is the risk.
Difficulty will catch up. The 313 blocks/24h pace is above protocol target — the next difficulty adjustment will bring it back. Per-block reward math for new entrants tightens accordingly. The economics still close for now; they'll close less easily three months from now. That's the nature of healthy hashrate growth.
Neither of these is a reason not to mine. They're the reason to plan with your eyes open.
Where to start, if you're new
Three steps, in order:
- Set up a PEARL wallet — walkthrough, takes about five minutes.
- Get an H100 or H200 — if you don't own one, RunPod is the fastest practical path to a working pod. Full step-by-step in how to mine PEARL on RunPod.
- Choose a pool — most newcomers will benefit from joining the largest pool with the lowest fee. As of today that's PearlHash (2.84 EH/s, 12,843 miners, 3% fee, 1 PRL minimum, auto-payout every epoch — read the deep-dive). Verify the current top of the table on /pools before committing.
From wallet to first share submitted is realistically under an hour for a first-time setup. First payout, with PearlHash's per-epoch auto-payout at 1 PRL minimum, typically lands the same session.
The bigger picture
For sixteen years, crypto's pitch for a slice of the AI economy has been "blockchain can mediate it." That pitch mostly failed because the value flowed in one direction — out of crypto, into AI infrastructure, with very little of substance flowing back. PoUW inverts the relationship: every dollar of compute now produces both AI value AND blockchain security, on the same hardware in the same operation. The two industries become co-located on the GPU instead of competing for it.
The Together AI partnership was the first time that inversion had a paying customer. The exahash-scale pools are the first time that inversion has measurable throughput. A week from now, this inflection moment will look small compared to where the trajectory is heading. The work itself is now visible.
If you've been waiting for a moment when the PoUW thesis stopped being a slide deck and started being a network you could measure — this is that moment.
FAQ
Is the 4.7 EH/s figure verified?
Yes — every pool's hashrate is polled from its own public API every ten minutes, and 24h block counts are verified on-chain by attributing coinbase outputs to each pool's wallet. The live, continuously-updated comparison is on /pools; nothing is estimated.
Which pool should I join?
For most new miners today, the math favors PearlHash — largest hashrate (predictable block discovery), lowest fee (3%), auto-payout per epoch, 1 PRL minimum. For miners specifically prioritizing zero minimum payout (relevant for very short rented sessions), MinePRL's no-minimum policy remains useful. All four pools are listed on /pools with live numbers.
Is it too late to start mining PEARL?
No — but the window where current network-difficulty levels favor new entrants is narrower than it was a week ago. Difficulty auto-adjusts upward as hashrate grows, so per-block reward math tightens over time. Run the calculator on your specific hardware before committing.
What hardware do I need?
NVIDIA H100 or H200 with sm90 compute capability. Full breakdown: best GPU for PEARL mining 2026. Don't own one? RunPod rents H100/H200 per-second with a signup credit.
How does mining PEARL produce useful work?
The mining workload IS LLM inference — the same matrix multiplication that ChatGPT, Claude, or Gemini perform when answering a prompt. The PoUW protocol uses that inference output as the proof-of-work claim. With the Together AI partnership live, real customer inference now drives mining as a byproduct. Full background: what is Proof-of-Useful-Work?
How do I convert mined PRL into USDC?
We're launching a Telegram-native OTC desk for instant PRL ↔ USDC swaps — 1.8% fee, no KYC, 2FA, 15% lifetime referrals. While it's still in pre-launch, sign up for the day-one DM at /otc-soon — first thousand subscribers will be notified the moment trading opens.
Where can I follow what happens next?
This site (lordofpearls.xyz) for live network data, independent pool comparisons, and analysis. @prlnet for Pearl Research announcements. @Lord_Kuba for editorial updates and chain-watching.