Some claim that Bitcoin mining is doomed to be concentrated in the hands of a few large mining corporations, and that the advent of ASIC mining is the culprit.
Well, I don’t know for a fact that this will not happen. However there are several factors in play that could prevent this scenario – a scenario which is undesirable, because the more centralized mining is, the more likely it is that a majority of hashrate would collude in an attack against the Bitcoin network.
The cited reason for centralization is that large companies enjoy economies of scale in mining. These companies, combined, will scale up their operation until the difficulty has risen so much that mining is only marginally profitable. Since hobbyist miners are presumably less efficient, at that point mining will be a loss to hobbyist miners, so they will be forced out of the market.
An analogy is sometime given with gold mining. Unlike the old days of the California gold rush, it is no longer possible for individuals to mine gold. Gold is only abundant in specific locations, and requires a complex mining operation to extract. The barrier of entry to obtaining suitable land and setting up a mine is just too high.
However, the situation with Bitcoin is much different, because of its extremely parallel nature. A gold mine will no longer work if you build a miniature version of it, so a person cannot run a small gold mine at home. In contrast, a Bitcoin mining farm is basically a huge number of copies of a single unit doing a very simple computation. Even a small chip measuring an inch across would be composed of many such units. So mining can definitely be scaled down to a level where everyone can run a Bitcoin miner at home. My first contention is that, since a large farm is basically multiple instances of a basic unit, the economies of scale that can be obtained are fairly limited.
This refers to the operation of the devices. There is still a big barrier of entry to actually manufacturing the devices; however, I do not see this as much cause for concern. Because Bitcoin mining requires a single, simple computation, designing ASIC to carry it out will require relatively little R&D costs, when compared with more complex circuits such as CPUs and GPUs. Because of this we should expect many manufacturers of such chips, enough to ensure one would cater to the hobbyist market – if they are willing to pay for it.
Assuming an equilibrium where the potential mining revenue is mostly known, what anyone – whether a professional entity or a profit-seeking hobbyist – would pay for a device, is the profit he expects to receive from it in its lifetime. And since the revenue is essentially the same for all, what differs is the cost of operation. For there to be a significant hobbyist market, there just need to be enough hobbyist with cost of operation lower than that of the large corporations.
One thing going for the hobbyists is that they may have existing, underutilized infrastructure that can be leveraged into supporting a mining operation. A large-scale company would need to set up an infrastructure specifically for its operation, while a hobbyist may have unused physical space in his residence, a power grid connection with spare capacity, or a computing device he could use to assign work to the mining devices. For example, if he is using a desktop computer (unlike some, I don’t foresee the death of those anytime soon), he may have a spare PCI express slot to which he could plug a card, feeding off the spare capacity of the PSU (it may be also possible to design it to run only when the other power-hungry components are idle, which is most of the time).
But much more importantly, It can be expected that in the long term, the major cost of mining will not be capital expenditure, but power. Therefore, anyone who can get cheaper power will have a huge advantage. A hobbyist might have an arrangement of getting free electricity from his host, effectively “leeching” power from someone else – this practice may be frowned upon, but it will happen, and only at small scale. But the crux is hobbyists living in colder countries, who would otherwise use a resistive space heater to warm up (less so, those who would use heat pumps or furnaces). Any power spent on a mining device is exactly deducted from what they would have to spend on their space heater, so for them the power really is free.
Many other factors are in play – mining is inherently risky, and hobbyists might be less averse to this risk than professionals, or maybe some of them will have a relatively higher estimation of expected profit. Hobbyists might have reasons to mine other than direct profit, which could also tip the scale in their favor.
Even if there are not enough hobbyists with favorable conditions, we should also consider professional companies large enough to enjoy some economy of scale, yet small enough to be more lean and efficient than the large companies. And again, if these companies are in the market for mining devices, some manufacturer will take their money. Many such companies can fit in the global Bitcoin mining market, and while this is not the same as the ideal decentralized vision, it is still a good approximation.
In light of all of this, I do not worry that the economics of obtaining and operating mining equipment will be a cause for harmful centralization of mining.
I disagree with one of your main premises which is that there are few opportunities for economy of scale in SHA2 mining. The cost of electricity alone is 20-40% cheaper  for industrial vs residential consumers throughout the US. Also, I suspect (but I don’t have figures for this) that the cost of a bulk order of mining devices is much lower than the cost to any retail consumer. I agree with your point that there would be an *even steeper* benefit to the economy of scale if the mining puzzle were some long sequential operation requiring custom hardware, but even with a perfectly parallel operation (sha2 works fine) there are still plenty of opportunities to save by buying bulk.
Note this has nothing to do with ASICs in particular.
I also disagree to a lesser extent about a) that electricity will eventually exceed the cost of equipment, and b) that otherwise unutilized power and equipment gives casual users much of an edge, but I’m not sure how to go about actually investigating either of these.
Thank you for your feedback.
1. I never said there were no economies of scale, only that they are fairly mild. Even a 40% discount on electricity (which I was not aware of, admittedly) is not too much when taking into account that some hobbyists may enjoy free electricity; and all the extra costs of doing this professionally.
2. There are and will be volume discounts on buying mining hardware, but long-term I suspect these to be smaller than in some other markets. If you take the market of desktop CPUs, for example, you’ll notice two things:
a. There are only two vendors (a result of the high R&D costs that go into making a modern CPU).
b. The customers need CPUs. If the prices of all CPUs were to double, the customer would still buy a CPU (perhaps an inferior one).
In these conditions of low competition and low sensitivity of total demand to price, vendors can maximize their profits by putting a pricetag far above the manufacturing cost; and therefore there is room for volume discounts.
In comparison, mining ASIC chips are expected to have many manufacturers because of their relative simplicity; and the customers don’t need them, they only buy them if they think the revenue is higher than the cost. If the price is increased to the point of no longer being profitable, customers will not buy and demand will drop sharply. Because of this, I do not believe there will be much room for profit margins and volume discounts.
3. Despite the post title, I did not go in depth into what ASIC has to do with this. However, the facts are that:
a. When Bitcoin mining was dominated by CPUs and GPUs, it was not centralized.
b. The advent of ASIC has caused mining to currently be more centralized.
c. People claim this trend will continue and that ASIC will cause centralization.
d. A mining market dominated by ASIC is different from one dominated by general purpose devices in several key ways: There are more manufacturers; manufacturers are less specialized and therefore prone to vertical integration; the mining market influences the pricing of devices. My goal was to analyze an ASIC market, not to compare it with a GP market.
4. If you take for example a 7850 graphics card, it costs $160 and consumes 130W. This means that with cheap $0.05 / KWh power, it takes less than 3 years for the power costs to exceed the hardware cost, which is well within the effective life of a device. There are some added costs on both the power and hardware sides so let’s assume they cancel out. However, as I claim above the price tag of ASIC mining devices should converge to be closer to the manufacture cost, and since the mining market will be (hopefully) smaller than the GPU market, it is expected to use an older process with a lower manufacture/power ratio. Together this should make for power costs being significantly higher than the hardware.
It is also possible what we will see is that devices are first bought by professionals, and when they outlive their usefulness sold to hobbyists with lower power costs.
5. Some other costs professionals should have that hobbyists do not: Insurance and guards for the equipment; lawyers and accountants; maintenance personnel (this one assumes the hobbyist enjoys tinkering with the hardware and volunteers his time to maintain it).
6. My initial goal was to give a complete quantitative analysis that “proves” that hobbyists will dominate. I then realized I don’t have the numbers to back that up, so I aimed for the lower target of highlighting some moving parts that work in the hobbyists’ favor. This may or may not be sufficient, but I do believe that even if not, all that is still required is a nudge in the right direction; for example, someone with a vested interest in the health of the Bitcoin network could donate subsidies to hobbyist miners for buying equipment.
7. In the forum I mentioned I do not believe it is necessary to assume “irrationality” for ending up with decentralized mining. However, my main goal in this post is to demonstrate decentralization is possible, so if factors other than classical profit seeking tip the scale, that is fine with me.
8. All in all I hope you were not too disappointed by this after my “promise” in the forum.
I agree that pooled mining does not offer much in the way of economies of scale. But I think that misses the point. The reason why people tend to prefer pooled mining is because of risk aversion. Solo mining offers a low probability of a high reward. Pooled mining offers a high probability of a low reward. The latter is more attractive to most people for the same reason that people generally prefer to invest the bulk of their cash in low-risk low-reward investments rather than lottery tickets. Indeed this is the case even if the net discounted return from pooled mining is actually lower than from solo mining. Of course none of this has anything to do with ASICS (yet another reason why Litecoin is a scam!). It is simply a reflection of the low probability of any given solo miner mining a block. The solution is simply P2P pooled mining which avoids the centralised control problem.
I’m assuming hobbyist miners use pools with a modern reward method such as Multi-PPS. Then variance is not an issue and it all boils down to expected gains.
Pingback: Multi-PPS: Dissolving mining pool centralization | Fiery Spinning Sword