pull down to refresh
whether or not LLMs solve it, several hundred billion dollars of capex is being spent as though they will, and that capex is bidding for the same megawatts I am. I have to plan around the spending regardless of whether the thesis is right.
It's liquidity. Our payout flow is one directional. We only ever send, we never receive over Lightning, so channels drain in one direction and someone has to sit there rebalancing them forever. That's a standing operational job, not a feature. And our larger accounts get payouts that don't fit comfortably in a channel anyway, so for most of our volume Lightning is the wrong rail.
It trades today
The shape: a miner sells forward the output of some hashrate over a tenor, usually 30 days out to a year, a few EH at a time, and takes cash or BTC up front. For the miner that's funding. Cheaper than equity, no lien on the fleet. On the other side is a fund or a treasury company that wants hashprice exposure without buying machines, hiring an ops team, or explaining a datacenter to their board. Both sides get something they actually want, which is the only reason the trade exists.
Difficulty risk sits with the buyer by construction, because what's sold is hashrate over time and not a fixed quantity of coin. Difficulty rips, buyer gets less BTC. That's what they're paid to take and it's precisely the risk the miner wanted off the books. The price leg gets handled separately with a non deliverable forward so the miner can lock fiat.
The risk that actually keeps me up is performance. A miner who loses a site or goes bust mid contract can't deliver, and there's no exchange standing behind this. We manage it with collateral, with sizing, and with the structural advantage that the seller mines to our pool. We're the ones paying them, so delivery comes out of a stream we can see and net against.
Variance absorption gets cheaper as you get bigger, engineering cost is basically fixed, and the marginal cost of onboarding another exahash rounds to zero. The economically optimal pool is the whole network.
The only things holding it apart are customer preference for not having a single vendor, counterparty caution, and jurisdictional convenience. Those are preferences. They aren't economics, and preferences lose to cost curves over long enough periods.
Which is why shaming the big pool is a losing strategy. You cannot moralize your way out of a cost curve. If you want pool size to stop being frightening, make it stop mattering, which means separating the accounting function from the block building function. A pool that doesn't build the template is a payment processor, and nobody loses sleep over how big Stripe is.
I think your point is that since you don't control the miner phsyically, do you have agency over it's hashrate?
You can take the analogy further - if you don't own your power source, do you actually mine?
At the end of the day there is coordination that needs to occur, and you inject yourself into that stack at some layer. I don't think it's possible to be a completely sovereign miner.
Also, your S9 will look great in a museum or on your bookshelf :)
Yes
FPPS is heroin for miners. Once they are on, they never get off.
That said, FPPS naturally dies over time. As the coinbase reward goes to zero, transaction fee volatility (as a percentage) increases and makes the "F" part of "FPPS" vary significantly day to day and week to week.
Pools that want to keep "PPS" will have to switch to "PPS+" which eventually just evolves to PPLNS.
So... bitocin fixes this?
That almost nobody who points hashrate at us cares about anything this website cares about.
In eight years I can count on one hand the customers who have asked me a question about template construction, transaction selection, or censorship. It is not a purchasing criterion. It has never once come up in a competitive deal. What comes up is fee, uptime, payout timing, and whether a human answers at 3am when a site trips. That's the whole list.
I don't say that to dunk on anybody. I say it because it explains the thing people keep looking for a conspiracy to explain. Pool concentration isn't a plot. It's an absence of demand. Nobody is suppressing miner-selected templates. Miners aren't asking, so nobody builds it, so it stays hard to use, so miners don't ask. That loop has been running for years and it will keep running until somebody's revenue depends on breaking it.
The second surprise, smaller: how little of this job is cryptography and how much of it is plumbing and customer service. I thought I was getting into Bitcoin. I got into industrial ops with a payment processor attached.
Complete and total lack of onchain activity. The reasons are up for debate.
This is an existential problem for Bitcoin and there is no workaround. All the BIPs, future challenges, etc... are meaningless without transaction fees onchain.
Let me take a look - there isn't much to optimize right now so it's hard to prioritize a change like this. The reason is that orphan rate is already basically 0, and the only other optimization to make would be transaction stuffing, but since blocks are basically empty now anyway, they all fit and there isn't much knapsacking to do.
We'll stay in touch and if things start to develop we can take a look
You've got the mechanics right and the premise wrong. The 1 sat/vB floor was never
a price floor. It was a relay default in Core, minrelaytxfee, and a floor only
earns you money if there's demand you're turning away. Blocks haven't been
reliably full. The space under that floor was going out empty. A 0.1 sat/vB
transaction isn't taking a slot from a 1 sat/vB transaction, it's taking a slot
from nothing.
Second, you couldn't defend it even if it were worth defending. It's a cartel with
no enforcement mechanism. Any single pool that relaxes its floor out earns every
pool holding the line, submission paths around the public mempool exist, and there
are enough of us that somebody defects on day one. Fee floors need unanimity among
competitors who don't much like each other.
Your bidding ladder point is the good part of the question, and it's correct.
Fractional bumping does compress the increments. It only costs miners money during
congestion, though, and during congestion the clearing rate is 20 sat/vB and what
happens down at 1 is irrelevant. Sub 1 sat/vB is a symptom of a slack market, not
a cause of one.
How I felt about it: fine, we reconfigured. It cuts our costs too, since our own
payout transaction is our single largest expense when fees run hot.
The thing worth being annoyed about isn't the floor. It's that we're arguing over
tenths of a sat while fee revenue is a rounding error in the security budget. No
relay policy manufactures demand for block space.
Ah the Drivechain fork.
If there is a market with economic activity, there's a world where miners will split some hashrate.
The most recent example of this was Fractal Bitcoin, where miners were arbing it against the motherchain
We'll stay on whichever tip is paying the most, which will almost assuredly not be LukeChain
Some of it, none of the technical parts. I was on Einstein at Salesforce before
this. Models from that era are museum pieces and none of that code would survive
contact with 2026.
Two things stuck. First, the model is never the bottleneck. It's the data
pipeline, the labeling, and getting the thing running in production without a human
babysitting it. Still true, and still most of why enterprise AI projects quietly
die.
Second, and I didn't appreciate this until much later, compute demand is a physical
business. Someone has to buy the machines, find the power, get the interconnect,
and move the heat. Mining learned that the hard way about a decade before the AI
people had to. It's most of the reason mining companies ended up holding the power
and not the other way around.
I left AI to go mine Bitcoin and AI came and found me anyway.
Different answers, which is the interesting part of the question.
Business: firmware. Best margin thing we do and the hardest to copy. Plenty of
people have taken a run at serious third party mining firmware and there are two
left standing. It's also the control layer, so if I keep firmware I can rebuild
most of the rest of the stack on top of it later. Keep the pool and I can't
rebuild firmware.
Personal, money no object: the pool. Which is the worst business of the bunch.
Fees only move one direction, customers switch over a basis point, and what we add
over the next pool is close to nothing on a spreadsheet. I'd still keep it.
Running a pool means being in the chain instead of selling picks to people who
are. I don't want to be a vendor to Bitcoin. The pool is also where I learn
things. Every problem in mining shows up in pool logs before it shows up anywhere
else.
Ask me this in 2019 and I'd have given the same two answers and been wrong about
the business half, so grade accordingly.
Really tough question - Something has to give. We've seen just so much investment in that space that there has to be a cool-off at some point
Based on current trajectory, I'd be forced to say at least 50% of business is AI-related.
I am a bit of an AI doomer myself, so take this with a grain of salt. But the question I ask myself is "is there infinite demand for intelligence? And I think the answer is yes. So, with that mindset, it naturally leads me to think that all economic activity will become AI-dominated.
It's tenor dependent. Inside ninety days there are enough counterparties that you can usually get a price in a day or two. Past six months it's bespoke, one buyer at a time, and the price is a negotiation rather than a quote.
Spreads are wide. Single digit percentages wide, not basis points. Clip size is low single digit exahash, so you can't move a large fleet's production in one ticket. There's no order book, no clearing house, no screen. It's a chat market with a broker in the middle, and a lot of the time the broker is us.
So the useful way to think about it: liquid enough that a miner who needs funding can reliably get funded. Not liquid enough to treat as a hedge you can exit quickly. Put a position on expecting to hold it to maturity, because unwinding early means calling the same small number of people and telling them you need out, which is not a negotiating position.