The sad story of SHA-256 and why we need a new PoW algorithm
Back in the early days of Bitcoin, anyone could be a miner, and the Bitcoin software came bundled with tools to help normal users generate coins. Anyone with a decent CPU or GPU could mine, and many did. The line between being a user running a full node, and being a miner, was very blurred. It was extremely rare to be an enthusiast about Bitcoin in those early days and not be mining in some way. Back then, the Bitcoin network largely dodged attention and attack by staying under the radar. The hashing power was small, and it was feasible for someone to rent enough hardware, at a really cheap cost, to launch a 51% attack.
No, don’t “bring it on”.
The project needs to grow gradually so the software can be strengthened along the way.
— Satoshi Nakamoto
During this stage of Bitcoin’s development, people would solo mine. The hashing power was spread out between a few different individuals with what would today (in hardware terms) be really small scale mining operations. People like Laszlo, ArtForz and tcatm built the first GPU mining operations, and so began the industrialisation of Bitcoin mining. It became clear at this stage that the way proof-of-work was set up made it possible for existing hardware to be quickly made irrelevant, with the introduction of more efficient ways of mining. The transition from CPU to GPU mining favoured a few hardcore users who had the technical knowledge to mine with GPUs, but eventually, regular users got access to hardware also, and began their own operations.
[13/08/2010 15:22:53] <ArtForz> If I sold the BTC I have sitting here right now at $0.04, I’d have made back the initial investment for GPUs, power costs and a few $100 for my time
Those clever enough to mine with GPUs and FPGA’s before anybody else were rewarded with thousands of Bitcoins. As it should be. But nobody really seemed concerned, that at times, the hashing power was concentrated in the hands of very few individuals. Eventually with the release of more software to optimise GPU mining, and the invention of pooled mining, it became easier for regular users to mine again. The hashing power eventually became more distributed.
A similar story repeated itself with ASICs. They replaced GPUs, and the first people to use ASICs made a lot of money, and controlled a very large amount of hashing power, but eventually, actual users got access to ASICs for themselves and could create their own mining operations. And because of pooled mining, they could make more consistent income. Gradually the power of huge solo mining operations declined. It was no longer possible for one person to directly control a huge amount of the hashing power. The pools were a point of centralization, but they were still actual pools, and composed of thousands of independent users.
In those days, miners were not political. They simply mined, and nothing more. They didn’t attempt to impose their will on the users, because mostly, they were the users. A miner with a lot of hashing power had a very small window to make as much money as he could, before the rest of the network caught up to him. He couldn’t afford to arbitrarily flaunt his hashing power, and use it in some way to gain power or influence for himself.
Eventually ASIC manufactures figured out that they could exploit the way of proof-of-work works to make even more money and compete against their own customers. Rather than create an ASIC, and selling it right away to customers, or using it themselves exclusively, they could instead do both, by pre-selling the hardware, secretly mining with it, driving up the difficulty for everyone else, and then send this used and mostly obsolete hardware to their customers at a much later time.
This behavior made it extremely hard for regular users to compete with the companies they were buying ASICs from. The manufactures sustained a permanent and total advantage over their own customers. Those entering the mining business already were on their way to becoming losers, at best they could ROI and make slightly more than their investment, at worst they would lose significant money. Many “cloud mining” operations also popped up, which was just another way for regular miners to get scammed by either ASIC manufactures, or people closely connected to ASIC manufacturing.
Most of these ASIC companies died because of bankruptcy by being out competed by other ASIC companies, or by being too greedy and fraudulent, to the point that they were shut down by governments. The Bitcoin community quickly learned to distrust mining companies, and for a while there wasn’t really any reliable ASIC company. The market was ready for disruption and a new player.
Sometime in 2013, a new ASIC company emerged out of China, founded by Jihan Wu and Micree Zhan. They perfected the Bitcoin ASIC company routine down to a science, and with their access to cheap Chinese electricity and connections to cheaper fabrication, they pushed out all competition. BITMAIN earned a reputation for extremely reliable delivery times, their hardware was robust and performed well relative to other mining hardware at the time. While everyone was suspicious of American and European ASIC companies, BITMAIN was busy building trust and connections with their customers.
They figured out what nobody else did; that it was possible to out-compete your own customers, but make them happy with the hardware and delivery times at the same time. You just had to be a little less greedy, a little less scammy, and you could make significant amounts of money with almost no risk. Other ASIC companies rushed to make as much money as possible, and burned out quickly with backlash, but BITMAIN adopted a more measured, scientific and clinical approach to the ASIC business.
Patience is power; with time and patience, the mulberry leaf becomes a silk gown.
— Chinese proverb
As a miner, every time you buy hardware from BITMAIN, you are losing. You are funding your own competition, and eventually, they will put you out of business. Maybe in 6 months, or 1 year, or 2 years. But your time will come. Rather than securing the Bitcoin network, you are doing the opposite by making it more likely that BITMAIN continues to dominate long into the future.
BITMAIN disguises their massive mining farms behind “pools”. But these are most likely not actual pools, they are just fronts for massive operations, with only a small minority of independent hashing power. As the NYA signalling showed, the hash rate is remarkably consistent, and does not behave independently. Miners did not move away from BITMAIN pools, despite the community’s complete rejection of BITMAIN’s support of NYA. Then almost overnight, when Jihan permitted Segwit to be activated, 80%+ of the hashing power suddenly supported BIP 91.
As with NYA, BITMAIN has no qualms with using their control of an overwhelming majority of the hashing power in order to achieve political objectives. These objectives do not align with those of the Bitcoin users, such as their disregard for preserving decentralization. The Antbleed scandal exposed that they go so far as to backdoor their equipment, to spy on their customers. They significantly delayed the Segwit soft fork and will likely delay other technical upgrades that goes against their agenda.
They have used dangerous optimizations such as ASICBOOST which have degraded the performance of the Bitcoin network. Indeed, just the existence of ASICBOOST alone is enough to justify switching to a different algorithm. They have prompted other companies to form patent alliances to try to overcome their dominance. These patent alliances are even more harmful than BITMAIN’s activities, and the patented version-rolling version of ASICBOOST (an attempt to “one-up” BITMAIN) makes it even harder for new competitors to enter the space. By trying to “decentralize mining”, companies like Halong Mining and pools like Slush have made the problem, through the introduction of patent alliances and intellectual property wars, even worse.
What can we do?
Based on conservative estimates of gross margin of 75 percent and operating margin of 65 percent, Bernstein analysts calculate that Beijing-based Bitmain made $3 billion to $4 billion in operating profits in 2017.
I believe BITMAIN is slowly directing Bitcoin mining into a future in which they are in total control of 100% of the hashing power. They will, slowly but surely drive out other miners, their own customers. They may develop more sophisticated ways to hide their hashing power, but as they IPO, we can hope it becomes more clear the extent to which they dominate Bitcoin mining. They have demonstrated complete contempt for consensus, and being located in China makes them at risk of being forced to engage in Chinese state censorship.
There has never been, and will never be, actual competition against BITMAIN. Due to economies of scale, it’s not possible to compete on cost per J/THs for long, without BITMAIN slashing prices in response. Mining is a ruthless business where the overwhelming majority of customers will go with the cheapest (and reliable, as BITMAIN has proven to be) hardware that gives them the best performance. Let’s look at some of their so called “competition”:
Let’s first look at what you get from BITMAIN right now. Using their latest batch due to be shipped in July. They are offering an Antminer S9i for $673, which gives me a hashing power of 14TH/s with power consumption of 1340W at the wall. That comes to an efficiency of about 96 J/TH. So that’s about $7 per J/TH. Let’s see if we can get cheaper than that from the “competition”.
They don’t sell any hardware anymore, so at present, they aren’t competing with BITMAIN. Something very similar to what they sold is now being sold by another company, so let’s look at that…
They sell a mining unit called the Terminator T2, which does 17.2TH/s at 1430W at the wall. That comes to an efficiency of 83 J/TH. The unit costs $1080. That’s about $13 per J/TH. Almost costs doubly more than what BITMAIN is offering, so it’s not really cost-effective to buy from this company.
They haven’t shipped their miners yet, but they are claiming they operate at 24 TH/s and consume 1950W at the wall. That comes to an efficiency of 81 J/TH. The unit will cost $2000, so that’s a massive $24 per J/TH. This won’t compete at all.
They are currently selling their Ebit Miner E9.3. The unit does 16TH/s at 1760W. That comes to an efficiency of 110 J/TH. The miner costs $950, so $8.6 per J/TH. Still more expensive than BITMAIN, even though it’s newer generation hardware.
The AvalonMiner 841 does 13TH/s at 1290W power consumption. That comes to an efficiency of 99 J/TH. The unit costs $710. That’s a cost of $7.1 per J/TH. This is closer to BITMAIN, but still not quite there, and this is a product line that came out just a few months ago, BITMAIN’s S9 series is very old, and should be getting more convincingly out-competed by now. At present though, this hardware can be considered competition, but that’s largely thanks to BITMAIN’s delay in releasing their S11 as response.
The problem is that many manufacturers already exist, ready to take on BITMAIN, but they can’t overcome the monopoly
Many of these competing manufacturers won’t be competing with BITMAIN in a few years, BITMAIN will just release new hardware soon, putting all these companies in an even more difficult situation. These companies could join the graveyard of failed ASIC companies (KnCMiner, Spondoolies-Tech, etc). The difficulty will be driven up even further, and mining will continue to get further consolidated by BITMAIN. If any real competition does enter the space, BITMAIN just uses their dominance and ability to withstand pain to brutally cut prices, hurting themselves short term, but in the long term driving away more competition.
PoW change is one way we can restore the state of mining to be more competitive. The ecosystem already exists for ASICs, we just need a fresh start, without the need for new players to catch up to huge monopolies that have spent years consolidating and centralizing SHA-256 mining. Additionally we can look into schemes to adjust and tweak the economics of proof-of-work, to make the swings less erratic and dramatic. Making it harder for ASIC manufacturers to push their customers and competitors out of business.
Won’t a PoW change benefit BITMAIN the most?
This is total nonsense. BITMAIN are the biggest miners, and they would be hurt most by a PoW change. It’s likely their biggest source of income, by far. If the community were to support a PoW change, it’s unclear if BITMAIN could survive long enough to sell/build ASICs for the new algorithm at significant scale. And even if they did, they would be in a much weaker position relative to the competition, giving others a better chance to compete against them.
Assuming the PoW change doesn’t help, we would be right back to where we are right now anyway. So it doesn’t harm us to at least try to fix the problem. It’s become clear that SHA-256 has failed its job of being a distributed PoW. If the hashing power is consistently controlled by one company, it puts the security of the entire Bitcoin network at risk.
It’s true that a PoW change will hurt small miners too, but they are already being hurt by the current situation, and most may never ROI anyway. BITMAIN has shown that they don’t mind short term pain, and smaller miners should think in the same way. Eventually, with a new PoW, smaller miners that got hurt by the switch to the new algorithm may be able to better compete thanks to the disappearance of an abusive monopoly that has been killing them for years.
At present BITMAIN has such a sustained and dominant advantage that its almost impossible for any new players to reasonably compete with them. They have completely broken proof-of-work. They have become a trusted third party, with many in the community being seemingly OK with their direct control of >50% of the hashing power, so long as they behave. Real competition has not arrived and given the economics of the situation, it may never come. A PoW change is the only tool we have that could fix the present situation. We should at least consider it over the coming months if the problem continues to get worse.