PEARL Network Hits 24 EH/s ATH — 2.5% of Bitcoin's Hashrate, 100% of the Useful Work
PEARL's network hashrate just hit 24 EH/s — a new all-time high, less than 12 months after mainnet launch. That's roughly 2.5% of Bitcoin's hashrate, except every joule spent on PEARL produces something the network can actually use: real LLM inference. This article breaks down what 24 EH/s means, how PEARL got here this fast, and why most of the market hasn't priced this in yet.
The headline number, in context
As of late May 2026 the PEARL network's measured hashrate sits at ~24 EH/s, aggregated across the four major public pools and indexed solo miners. For comparison:
- Bitcoin network hashrate (May 2026): ~900 EH/s
- PEARL as % of BTC hashrate: ~2.5%
- PEARL network's annualized energy footprint (estimate): ~3.9 TWh/year
- Bitcoin's annualized energy footprint: ~155 TWh/year (Cambridge BECI)
Three quarters of one big number. But the crucial detail isn't the size — it's what the energy actually does. Bitcoin's 155 TWh/year produce SHA-256 hashes that secure the chain and nothing else. PEARL's 3.9 TWh/year produce a custom proof-of-work that is itself the matrix multiplication done during LLM inference. Same energy, two outputs.
How PEARL got to 24 EH/s in 12 months
The timeline matters because it tells you something about the demand curve. PEARL went from genesis to 24 EH/s in roughly one year — faster than Bitcoin's first decade, faster than any sm90-class compute network in history.
| Milestone | Approx. date | Network hashrate |
|---|---|---|
| Mainnet genesis | April 2026 | ~0 |
| First public pool live (PearlHash) | April 2026 | ~50 PH/s |
| MinePRL launches (H100/H200 optimized) | May 2026 | ~2 EH/s |
| PearlHash crosses 1 EH/s | May 2026 | ~5 EH/s |
| AlphaPool + Akoya launch with RTX support | May 2026 | ~12 EH/s |
| Current ATH | May 30, 2026 | ~24 EH/s |
For live pool-level data, see the always-updating PEARL mining pools comparison. For the breakdown of which GPUs are doing the work today, see the best GPU for PEARL mining.
Why 2.5% of Bitcoin is a bigger deal than it sounds
Two and a half percent sounds small. It isn't, for three reasons:
1. It happened in 12 months, not 12 years
It took Bitcoin nearly a decade to reach 24 EH/s of network hashrate (Bitcoin crossed that mark sometime in 2018). PEARL replicated that level of compute commitment in ~12 months. The growth slope, not the absolute number, is the story.
2. PEARL hardware is far harder to acquire than ASICs
Bitcoin's hashrate is dominated by purpose-built SHA-256 ASICs. They are cheap per TH, mass-produced, and have one job. PEARL's compute is done on NVIDIA Hopper- and Blackwell-class GPUs (H100, H200, RTX 5090, RTX 4090) — the same chips ChatGPT, Claude and every major LLM are trained on. The same hardware that mines PEARL is what the entire AI industry is fighting over. 24 EH/s of competing demand isn't just a hashrate stat; it's a quiet allocation signal from people who already paid GPU prices.
3. The work isn't disposable
A Bitcoin hash is computed, found or discarded, and adds nothing to the world other than chain security. A PEARL block, by construction, is the by-product of LLM inference done on the same kernel. The same matrix multiplications that secure the chain are simultaneously serving model inference requests. The energy double-counts as useful output.
What 24 EH/s of useful compute looks like
To make this concrete: imagine 4 TWh/year of pure GPU compute, dedicated entirely to LLM inference, with cryptographic security as a side effect. That is roughly equivalent to:
- The annual electricity consumption of Iceland — entirely producing AI inference, paid for by miners
- ~500,000 H100-equivalent GPUs running continuously (mixed with consumer Blackwell), or several large hyperscaler clusters worth of compute
- Enough inference capacity to serve roughly the same volume of LLM tokens as a top-5 AI provider
The economic implication is the unusual part: every block reward is a payment for compute that also produced security. PEARL's emission is, in effect, a subsidy mechanism that pays the AI inference market in PRL.
The market hasn't caught up yet
PRL is currently trading around $1.35 on the only meaningful OTC venues:
Open Lord Of Pearls OTC → · flat 1.8% per trade (0.9% per side) · instant settlement · 0% slippage · 15% lifetime referral
At ~$1.35, with 24 EH/s of GPU compute committed and a daily emission rate of ~1.2M PRL, the network's "fully-diluted compute-backed market cap" reads small relative to the AI compute it represents. Whether the market re-rates that is a separate question — but the underlying compute is no longer hypothetical.
See the full neutral side-by-side of where to buy and sell PRL today on the PEARL exchanges comparison, or read the detailed OTC desk comparison guide.
What this means going forward
For miners
PEARL is now the densest payout-per-watt PoUW chain in production. With 24 EH/s of competing hashrate, individual contribution shrinks — but the pool ecosystem (Akoya, MinePRL, PearlHash, AlphaPool) is mature enough that even modest rigs (RTX 4090 / 5090) can run profitably. See the best GPUs for PEARL mining and the live pools comparison.
For holders
The thesis is no longer "what if this works." It's "this works, and it costs less energy than 3% of one country to do it." PRL is the asset that captures the residual value of the AI compute being paid to secure the chain.
For the rest of crypto
PEARL is the first network where the proof-of-work IS the useful work — not adjacent to it, not bundled with it, but the same operation. That distinction is doing the heavy lifting in the energy comparison below, and it's why the next 24 months of useful-work crypto debate will almost certainly center on this chain.
Frequently asked questions
Is PEARL's 24 EH/s comparable to Bitcoin's 900 EH/s?
No — they're entirely different kinds of hashes. Bitcoin uses SHA-256; PEARL uses NoisyGEMM (a noisy matrix-multiplication primitive that is also the kernel of LLM inference). The number is comparable only as a measure of total GPU compute committed to the chain.
How much energy does PEARL actually use?
Estimated at ~3.9 TWh/year at the current hashrate, scaling roughly linearly with the network. This is approximately 2.5% of Bitcoin's energy use.
Is the AI inference "real" or is it just a story?
Real. The proof-of-work submitted by miners is exactly the matrix-multiplication done during NVIDIA-tensor-core LLM inference. The pool architecture (PearlHash, MinePRL, etc.) explicitly forwards inference jobs to miners as the work they're already paid to do. See the explainer on Proof-of-Useful-Work.
Where do I buy PRL?
On OTC desks. Lord Of Pearls OTC at otc.lordofpearls.xyz publishes the lowest fee on the market (flat 1.8% per trade). For the full neutral comparison see /exchanges.
Is PRL on a CEX yet?
Not yet (May 2026). The current OTC volume is the entire price-discovery surface. See the Bybit alternatives article for how to be set up the moment a CEX listing happens.
Will the hashrate keep growing?
Indicators suggest yes — RTX 5090 supply is broadening, RunPod and TensorDock are seeing PEARL-driven demand, and the pool fees are still in their high-ROI window. The next obvious milestone is 50 EH/s, which would put PEARL at ~5% of Bitcoin's hashrate.
Bottom line
24 EH/s is the moment PEARL stopped being a thought experiment. It is now 2.5% of Bitcoin's hashrate, 100% of the useful work, and the only meaningful way to access the asset is through OTC. The market's pricing of that fact is — by definition — what will move from here.