The Future of Bitcoin Mining: POW vs POS — what’s the natural way to mine?
Do you even Proof-of-Work bro? | POW Mining: A tragedy-of-commons, sort of | Proof-of-Stake to the rescue | Proof-of-Systemic-Risk | Ok, so what?
Do you even Proof-of-Work bro?
Bitcoin mining is a high-stakes game with asymmetric returns built into it. The network protocol pays rewards to participants (aka miners) in exchange for the computer power they contribute to it. Every day, roughly 1,800 Bitcoins are issued to miners as rewards for their work. A miner’s rewards can be ball-parked by dividing the hashrate it contributes to the network by the total network’s hashrate. All mining computers are not created equal — some are faster, and more profitable than others. It’s an arms race between miners to develop more powerful and efficient equipment, and to operate in the lowest cost structures globally.
As the price of Bitcoin goes up, so does the dollar value of the rewards. Juicier rewards motivate more participants to join and existing ones to grow. Since miners cannot be built and installed as quickly as Bitcoin price rises, the installed miners will experience asymmetric gains during sharp rallies. Conversely, if more miners pour in at a faster rate and the price of Bitcoin collapses, many will experience losses. The network is regulated by market incentives that allow for boom-and-bust cycles.
The recent run-up in Bitcoin prices from November 2017 to January 2018 motivated an onslaught of new mining investments totalling over a thousand megawatts. Some mining companies even went public (Hive, HashChain, CryptoGlobal), raising millions to buy more computers and set-up sites.
The power demand from miners is seen by some as ludicrous. This has led to people debating other ideas of reaching network consensus. Prominent developers are already working on one potential solution that gets rid of power-hungry miners — a new consensus algorithm called Proof-of-Stake.
Proof-of-Work Mining: A tragedy-of-commons, sort of
The POW algorithm in Bitcoin provides economic incentives for anyone who contributes computing power to verify and post transactions to the network. Literally anyone with a computer can participate. When new transactions are broadcasted to the network, computers compile them and compete to write them into a new block. The computer that first hashes the correct answer will earn the right to write the block and be rewarded 12.5 Bitcoin.
Each proposed block solution must include: all of the transactions in the previous block, new transactions in the last 10 minutes, and an acceptable “nonce”. This “nonce” is an artificial complication built into the protocol. It is intended to make the hash solutions harder as more miners join. By regulating the difficulty of the nonce, the network ensures that blocks will be written every 10 minutes and pay the exact same fees, regardless of how many miners are participating. The difficulty of the nonce gets adjusted every 2,016 blocks or roughly every 14 days.
As mentioned above, mining returns are a function of equipment cost and operation costs.
Miners that bought equipment and were profitably mining Bitcoin at $5k will be much more profitable if Bitcoin goes to $10k in 2 weeks. But miners that set up profitable operations with Bitcoin at $10k may not be profitable if Bitcoin goes to $5k. Let’s use examples to illustrate:
On July 1st, 2017 Bitcoin was at $2,514 USD. Bitmain Antminer S9s were priced at $1,415 USD — and colocation rates for miners in North America were around $0.11/kWh. If you started mining on delivery date (September 1st, 2017) — and depreciated the S9 over a year, you’d get approximately 1.44 BTC over the course of 1 year*, at an all-in cost of $2,721. This equals a proxy value of $1,947 per mined BTC. A 22% discount to outright purchasing the coin on the same July 1st, 2017.
Let’s fast-forward to December 1st, 2017, Bitcoin was trading at $15,006. The same S9 was priced at $2,730 (yes, exact same one, in the exact same website). Colocation rates were up to $0.15/kWh, if you could find any space. Assuming that you started mining February 1st, 2018 and depreciated the S9 over a year, you’d get 0.41 BTC* at an all-in cost of $4,474 — or $10,859.11. That price was a great discount to the price in December — but not so much right now when you can buy Bitcoin outright for $7,800 at the time of writing.
This arrangement almost guarantees asymmetric gains to miners in periods where Bitcoin price and adoption outpace the growth in the hash rate. Many are drawn to these expansion gains — unaware of the pain that can come during market corrections.
Mining can be a great business, if you have access to the best equipment (at cost), and a facility in the lowest cost hydro jurisdiction in the world. In the absence of either of these 2 conditions, outsized returns (or returns overall), will be temporary.
POW is not a perfect system but it works. And it works very well because the “wins” and “losses” are contained to a few market participants. Yes, it lends itself to centralization. Yes, it favours those with ASIC building capabilities and abundant natural resources asymmetrically. But that is how nature works — and containing the wins and losses keeps individual participants honest, providing temporary outsized incentives to innovate and accelerating the death of the inefficient ones in downturns. It makes POW a very antifragile and robust complex system.
If there is any value to Bitcoin, someone somewhere should find it profitable to mine the next block.
Proof-of-Stake to the rescue
The most anticipated Proof-of-Stake upgrade is Ethereum’s. It is a network with a market capitalization of $44 Billion USD that is scheduled to switch from POW to POS as early as this year. Its creator, Vitalik Buterin, described his POS design philosophy back in December 2016 and I have been trying to understand how it would play out since.
As I understand Vitalik’s concept (feel free to correct me), the idea sounds …simple? Participants would deposit their own funds into (presumably) a staking account. Staked funds would be inaccessible during a fixed period of time. Stakers would then be rewarded for good behavior or penalized for malicious conduct. Details have not been released but, presumably the network rewards and transaction fees would be evenly distributed amongst the stakers. Again, details are not released but let’s assume that stakers would be rewarded proportionately based the amount staked and the length of the stake. Kind of sounds like something I’ve heard before. Let’s dig deeper.
For simplicity’s sake, let’s assume that there are 3 stakers in the network during month 1. All of these participants are behaving well, in order to get their rewards. This would be a good time to mention that under Ethereum’s POS, the penalties would be asymmetrically larger than the rewards to prevent attacks. Ok — so 3 participants all staking different amounts over the same period of time. Assume the staked amounts below:
Month 1 Funds at Stake: 1,011
Participant A: 1,000
Participant B: 10
Participant C: 1
Now, assume that over month 1, a total of 3 Ethers were collected in fees and rewards. The bounty would then get divided as follows:
Participant A: 98.9% of the bounty (2.97 ETH)
Participant B: 1% of the bounty (0.3 ETH)
Participant C: 0.1% of the bounty (0.03 ETH)
That works out to the same interest rate for everybody (3/3,031 = 0.29%) for the month. Equivalent to a 3.5% annualized return (0.29% x 12 months = 3.49%).
This sounds great so far — let’s see what happens if we randomize one of the inputs like, say, the network rewards or the value of the amounts staked:
Assume that the same participants remain staking for month 2, but that due to a fortuitous event, many people decided to buy and transact in Ether during that period. A total of 100 Ethers were collected during that time.
Month 2 Funds at Stake: 1,031
Month 2 Network Rewards: 100
The effective annualized return for the stakers in month 2 would be 116%! This would, in turn, cause many more people to stake. As staking could be done immediately, people would rush to buy Ether to stake.
Now, let’s assume that for month 3, ten times as many people staked and the staked amount went up to 10,310 Ethers. Later in the same month, unfortunately, people found it too expensive to transact in Ether and rewards took a huge 10x dip.
Funds at Stake: 10,310 ETH
Network Rewards: 10 ETH
Effective annualized interest rate: 1.16%
People that bought Ether looking for 116% staking returns will not be happy with 1.16%. They will likely sell their holdings and cascade the price down.
This is an extreme example, just to see what would happen under volatile environments. This type of staking, unaltered, welcomes wild swings in Ether demand and price as the interest rates go up and down (much like FIAT). Some would argue that these the peaks and troughs can be “smoothened out” artificially. What would that look like?
Options include, creating a central authority to regulate on a “normal” interest rate. To ensure the rate was kept, it would collect excess rewards in up times, and issue new Ethers to ensure enough rewards are paid in down times. I have many issues with this solution. For one, its central banking all over — and second, artificial tranquility comes at the expense of large asymmetric shocks in the long run (think of it as a grand “bust” cycle made up of all the contained “little busts”, compounded) — idea borrowed from @nntaleb’s Black Swan.
Ok, so what?
My central argument is that boom and bust cycles are part of nature and therefore necessary. Said cycles tend to distribute rewards asymmetrically — and that bothers some, but it is the most efficient way we know of to keep a complex system thriving. Think of nature, the smartest and most complex system of all.
Trying to reduce asymmetries of returns between participants will simply transfer the shock of an asymmetric risk to the system itself. Making the it much more vulnerable and risky.
I am all for innovation, and especially for the conservation of nature. I would love to see consensus algorithms get faster and more efficient. However, think Proof-of-Work mining itself is a huge breakthrough and gets a bad rep. I think it is going to be around for a while — until we figure out something better.
Bitcoin. Becker et al.  estimated the typical cost structures in Bitcoin network and discussed the general viability of proof-of-work approach.
- Bitcoin Network hash rate & Price data from www.blockchain.info
- Assuming that difficulty of September 1st, 2017 remained constant at 888,171,856,257
- Assuming that difficulty of February 1st, 2018 remained constant at 3,007,383,866,429