This article is the original research of SAI team and does not constitute investment advice. For reprints, please contact: service@saiheat.com

Many people think that Bitcoin mining is a special industry, in fact, macro view of mining it is not much different from the traditional computing industry: the cost of the computing industry mainly comes from the purchase of servers and electricity expenses, the main cost of mining is also the purchase of mining machines and electricity expenses, both are the arithmetic power or its products sold to obtain returns. Previously, I pointed out in ” computing and energy – The Fourth Technology Revolution Driven by Computing” that the computing industry is a production manufacturing industry with electricity and data as the means of production, and energy and chips are the core of the computing industry.

I. Calculation of the Cost and Return of Mining

Many people take the purchase of a mining machine as the main cost when mining and ignore the importance of energy cost; some people only focus on the cost of single-T computing power when buying a mining machine and do not pay attention to the power consumption ratio of the chip; some people may spend a higher price to buy the machine with the lowest power consumption ratio, but finally find that the return on investment is not as good as other machines. This is the problem is due to the decision is not to do a precise quantitative analysis of the target results (ROI), only based on the feeling to do qualitative analysis generated results.

1. What is the purpose of mining?

This question seems simple, but in fact many people have not thought clearly. First of all, miners choose to mine, must be to obtain BTC excess return, which is based on the long-term (1-3 years) price of BTC bearish case to make the choice. At present, there are only two ways to get BTC: buying it directly in the secondary market such as exchange and OTC, or acquiring it by mining. So we have a major premise that whether you choose to buy BTC directly or mining, it is when the price of BTC is expected to be in an upward channel. After the bull market in 2017, the bottom of the bear market at the end of 2018, and the shock reversal in 2019, the whole market is still generally believed that a new round of bull market will come after the halving in 2020 and the halving, so various funds start to prepare to enter this market in different ways. It is also in this context that we are discussing whether to mine, how to mine and which is better to mine or buy coins.

Many ordinary miners will feel that mining is a lower cost way to obtain BTC. This idea comes from a simple calculation: as shown above, taking the latest generation of miners as an example, with the current coin price of 7300USD and the electricity cost of $0.36/kWh, the ratio of the daily electricity bill payable to the revenue mined (i.e. the electricity cost as a percentage) is about 40%, thus showing that mining can be done at 40% of the market price lower cost to obtain Bitcoins.

This is obviously not true, because compared to taking a sum of money and buying Bitcoins directly, mining involves using the money to buy a miner first and being in a state of unpaid miner for a long period of time, while buying Bitcoins directly at this point is a direct net gain if the price of the coin goes up. If you think about it more carefully, you will think that the miner itself is also an asset, and selling it again will have a certain value, and if the timing is right, you can sell it at a higher price than the original one, which is equivalent to mining coins for a period of time. Not necessarily, this needs to be calculated precisely.

2. Mining ROI calculation

The operation of miners in the actual mining process is a complex process, just say sell coins to pay the electricity bill, there are many ways to operate: dig out every day and sell the electricity bill, choose a suitable price every month to sell coins to pay the electricity bill, do not sell coins to pay the electricity bill with extra fiat money, etc. Which is a better way in the end? This is to return to the premise we discussed before, that is, the BTC price is in an upward channel, eventually there should be more BTC in hand to get higher income, the early mined coins because of the low price of the coin at that time need to sell more coins to pay the electricity bill is not cost-effective, because the electricity bill is in fiat currency is a definite number, mining the annual electricity bill can be calculated, then choose to sell part of the coins at 7500-9500USD to pay the electricity bill. The best way is not to sell the coins and use the extra money to pay the electricity bill, so the number of BTC left in the end is the most.

Let’s use a simplified model to verify, with 10 million funds mining for two years, each difficulty cycle difficulty up 2% (the actual will be higher), with an average annual electricity cost of 0.36 RMB / degree, according to the S17PRO 50T 10000 RMB / unit model to calculate the final number of BTC can be obtained.

Option 1: 10 million RMB all bought into the machine, buy coins every month to pay the electricity bill. In order to simplify the calculation, the first year coin price is 7500USD, the second year is 15000USD, through the calculator, we can calculate the final cumulative number of remaining BTC is about 112.7.

Option 2: 4.46 million RMB to buy the machine, 5.53 million RMB to pay the electricity bill for two years, the final cumulative number of remaining BTC is 127.5 (as shown in the figure).

If the residual value of the mining machine is 25% after two years, the currency price is 15000 USD (the price of the new mining machine should be the net income per T of the current single day * computing power * 200 days, taking into account the halving and the increase in computing power. After two years, the computing power is likely to be double, so the daily income BTC is about 1/4 of the current number. The doubling of the currency price and the impact of machine depreciation make the remaining residual value to 25%, which is an ideal value). In the end, Option 1: Net assets of approximately $14.33 million with a 143% return on investment. Option 2: Net assets of 14.5 million, ROI 145%. However, if you buy the coins directly at the beginning, the net assets will be 20 million after two years, with a 200% ROI. If the residual value of the miner is 0 after two years, the ROI of option 2 is 15% more than option 1.

If we look at it from the perspective of a mining machine, the time stretches longer then the final residual value will be close to 0. Several transactions of the mining machine in the process are also calculated according to the price and difficulty of the coin at that time and the depreciation of the machine, so we connect the process, which is equivalent to buying a brand new mining machine until it is scrapped. According to the previous calculation we choose the model of not selling crypto coins to pay the electricity bill in fiat currency, at this time we can derive the calculation formula：

P is the coin price (yuan), mining d days, daily single-T computing power gain is α coins, single-T computing power purchase cost is A yuan, g is the chip power consumption ratio (kw/T),e is the single-day electricity cost per kw (yuan).

The trend in the number of BTCs gained per T of computing power per day α is shown in the following figure.

II. Computing and Energy Parameter SAI and Computing and Energy Theorem SAI Equation

1. Computing and energy parameter SAI

P is the coin price (Yuan), mining d days, daily single-T arithmetic revenue is α coins, single-T arithmetic purchase cost is A yuan, g is the chip power consumption ratio (kw/T),e is the single-day electricity cost per kw (Yuan).

In the above formula, the coin price and daily return α can have preliminary predictions based on the secondary market and miner production in the case of time d determination, which belong to objective unchangeable parameters; while A+d*g*e become the various parameters among them that determine the overall mining return on investment are determined by miners according to the conditions when they choose miners and mining sites. So we define the sum of arithmetic cost and energy cost per T of arithmetic power mining in d length of time as the arithmetic energy parameter SAI, or S for short, then：

With 360 days as a cycle to compare the mainstream model’s energy calculation parameter S, electricity costs at 0.37 yuan/kWh for the whole year, according to the mainstream miners first-hand spot price on December 27, the following table. From the perspective of return on investment, T17 has the highest rate of return for the smallest energy calculation parameter, but considering the residual value of the miner after one year, S17+ is definitely higher than T17, and the two energy calculation parameters are very close to each other, so S17+ is the best choice at this point in time.

2. Computing and energy theorem SAI equation

Let’s look at the BTC market from a macro perspective, if we stretch the timeline, then the demanders need to get BTC either by direct purchase or by mining to get it, and the cost of the two ways should be close. If there is a cost difference, then the market is the most real and will definitely be adjusted by the hand of free regulation. If the ROI of mining for BTC for a period of time is higher than buying coins directly, then more funds will start mining, thus squeezing the profit of mining to a level lower than that of buying coins directly, at which time the funds find it more cost-effective to buy coins directly and will start buying coins directly, thus leading to the ROI of the two ways to The return on investment fluctuates within a relatively stable range. In the real market, we will find that the ROI of mining revenue plus the residual value of the miner over a longer period of time will be close to the ROI of buying coins directly in the secondary market.

We use R1 to denote the ROI of mining, R2 to denote the ROI of buying coins directly, the computing and energy parameter is S=A+d*g*e, the initial coin price is P1, and the final coin price is P2, then according to R1=R2, it can be derived that：

After the above formula is simplified and deformed, we can get the computing and energy theorem (SAI Equation): under ideal conditions, the same time and the same amount of money chosen to mine and the same amount of BTC expected to be obtained by choosing to buy coins directly, the same return on investment is obtained, i.e.

Through the computing and energy theorem, we can determine whether it is more rewarding to choose to buy coins directly or to mine at the current time based on the current coin price and the expected change in difficulty over time.

Caution：

a. We cannot speculate that the price of coins will increase based on the increase in mining difficulty and hence the decrease in return rate in the energy calculation theorem. This is a typical inversion of cause and effect, the price of coins is the most realistic response to the supply and demand relationship and expectations of the whole complex market, it will only increase because of the increase in the price of coins so more people go mining leading to the increase in difficulty, and the opposite is not true.

b. The computing and energy theorem reflects the overall macro dynamic stability of the current market and is the average level of the whole market, so the value of a specific parameter derived from the control variables is only a judgment basis, for example, there is free electricity mining, then the actual computing and energy parameters will be much lower than the current theoretical average computing and energy parameters, you can decisively choose mining.

c. In actuality, not all miners dig the miner to scrap the residual value of 0, so R1 will increase the value of the residual value of the miner. And the arithmetic energy theorem can be understood as leasing a period of cloud computing power, that is, there is no ownership of the residual value of the mining machine.

III. The Computing and Energy Theorem and the Practical Application of the Computing and Energy Parameters

1. Current mining market cost analysis

Today (December 29, 2019) the network-wide difficulty is about 12.95T, and the gain per T α= 0.00001942. According to the orders of mainstream miners in Q4 2019 and TSMC and Samsung 7nm and 8nm chip capacity arrangements, it can be expected that the network-wide computing power will reach 120-140E by May next year before halving. Therefore, after halving α may be around 0.00000600, the current coin price is 7350USD, so according to the computing and energy theorem, the current computing and energy parameter S=120. Corresponding to the current average electricity cost of 0.38RMB/kWh during the dry period and the average power consumption ratio of 80w/T across the network (simplified as 50E computing power is 100w/T for 16nm, and the other 50E is 60w/T level for new miners) , it can be calculated that the cost of arithmetic power A = -10, which is obviously not in line with the actual situation, then it proves that the final return on investment for miners who currently choose to buy mainstream new miners will be much lower than buying coins directly.

Through the above analysis of the actual situation, we will find that the computing and energy parameters in the market are higher than the current theoretical measurement, how should this be explained?

About the single T computing power cost A: a handful of new miners A is limited by the cost of production, mining machine manufacturers can not sell miners at a loss, the main cost of new miners is the cost of the chip, mining machine manufacturers invest heavily in research and development of new nano process is also the reason for reducing the energy parameters. Chip cost is mainly wafer cost, is calculated by area, smaller nano process not only lower power consumption than g, and smaller chip area to reduce A, according to S = A + d * g * e can greatly reduce the computing and energy parameter S to improve the return on investment in mining. In the secondary market, the miner is relatively free flow, plus the machine itself loss will be a single T computing power cost A is lower than the first hand with the same miner, A is also closer to the average cost calculated by the energy formula, but the miner is, after all, a physical object, it is impossible to circulate in the global trading market 24 hours a day at very low cost like BTC, so it will still deviate from the theoretical price. In the non-bull market phase, especially in the bear market, the arithmetic energy parameters of the second-hand market machines will be closer to the theoretical value of the computing and energy theorem.

About the cost of electricity e: currently in the dry period, the vast majority of new miners choose dry electricity and signed a full year, the average cost in about 0.37 yuan / degree. However, when it comes to the abundant water period, the price of bare hydroelectricity may be around 0.2 yuan / degree, when the actual calculation of energy parameters will be reduced.

About the gain per T α and power consumption ratio g: currently about to usher in the halving, halving is not only the new output of BTC production reduction, but also will eliminate a number of old computing power, the current old computing power mainstay is S9, the current network of about 40E computing power is s9, if all eliminated, then the difficulty and power consumption ratio will have a considerable impact. But the actual process should not be as drastic as imagined the whole network instantly drop computing power, because mining is originally the same competition, some people brush the frequency reduction firmware, some people change the control board to let the s9 two in one to further reduce power consumption, there are people with lower electricity costs, plus the coin price will also increase revenue. The real elimination of the s9 should be around the peak of the new bull market, in the high coin prices to promote the production of new mining machines in large numbers into operation, will bring further “metabolism” of the entire network of computing power.

2. shutdown coin price and miner elimination

Regarding the shutdown coin price, the shutdown coin price can be easily derived by using the computing and energy theorem formula.

For specific miners specific models can calculate the shutdown coin price, but the ROI has nothing to do with the shutdown coin price, and the actual operation may be a model that will not sell coins to pay the electricity bill, so the shutdown coin price has little reference significance for this type of miners. The miners who need to sell coins every month to pay the electricity bill are the ones who are more concerned about the price, but we have proven before that the ROI of selling coins every month to pay the electricity bill for mining in an upside expectation is not ideal.

With regard to miner obsolescence, with the computing and energy theorem and computing and energy parameters, we will have a whole new definition of miner obsolescence. Previously, miners would be considered by the market as needing to be obsolete because of the shutdown coin price, but it is actually the computing and energy parameters that can really decide whether or not to obsolete a miner. We take the current market 6200 a T17 42T and has been eliminated 200 yuan a 841 to illustrate, the same mining 180 days.

If you don’t consider the residual value of the miner, then the ROI of choosing 841 will be higher than T17 for 180 days from today’s mining. of course the residual value of T17 will be high after six months, but if 841 can be used for less than 0.15 RMB electricity, then the ROI of 841 will be twice as high as T17, and even close to the ROI of buying coins directly under the current price of the coin.

3. Leveraging operation of mining machine mortgage lending

Now there are richer types of financial products in the market, assuming that the initial funds are invested in mining at single T cost A1 (no coins are sold to reserve fiat currency to pay for electricity), then you can first use these funds to buy BTC at the current price P1, then pledge 60% of USDT at the current price P1 in the pledge platform, then use USDT to buy a single T cost A1 mining machine with 60% down payment, assuming that mining 180 days, after 180 days the coin price is P2, the residual value of mining machine per T computing power is A2, and the coin loan and mining machine installment payment can accept the final capital repayment and interest of 1% per month, according to the formula of energy calculation the number of BTC per T computing power can be obtained as follows：

The number of BTC that can be purchased by buying Bitcoins directly is

According to the current actual coin price and difficulty, if you use 1 million capital to buy T17, assuming that the coin price doubles after 180 days, then you can calculate that the ROI of this operation is close to that of buying coins directly. In fact, it is equivalent to doubling the leverage of mining, and the initial funds were first taken to buy coins at the current price, but because the mortgage bought the miner, and the current new miner’s computing and energy parameters are much higher than the theoretical value of the computing and energy theorem, so only the doubling of the coin price plus the double leverage formed by the mortgage can make the ROI of the funds catch up with the direct coin purchase. But this is undoubtedly a very high-risk operation, on the one hand, the trend of the coin price is not 100% certain, on the other hand, the actual loan repayment method is not likely to be the final principal and interest repayment, so the final conversion may not be cost-effective, the key still depends on the computing and energy parameters.

4. Mining returns on investment are higher than buying coins

Back to the computing and energy parameters, according to the computing and energy theorem, the ROI of mining and buying coins directly will fluctuate repeatedly within a certain range, and the market will constantly correct the difficulty, computing power price, coin price, electricity cost, chip power consumption ratio and other parameters. In most cases, since miners are not chip producers and chips are limited by production costs, the actual computing and energy parameters will be higher than the calculated value of the computing and energy theorem, but will be compensated by the multiple effects of the residual value of the miner and the coin price. Since the coin price and computing power gain are objectively uncontrollable in the arithmetic energy theorem, and it is also objectively fair to treat all participants equally, the cases where the mining yield R1 is greater than the direct coin buying yield R2 can be divided into two categories.

a. Due to the lagging nature of computing power, when the coin price rises rapidly and sharply and the computing power does not rise in time, it may lead to R1>R2 depending on the real computing power parameters of the market in a short period of time. but the probability of this is very low and the window is very short. The more important thing is that if the price of the coin has risen sharply and quickly, it probably means that the bull market has come. Historically, every bull market in Bitcoin has had a very short duration of rapid and sharp rise in the price of the coin. In addition, the actual process of mining “friction” is very large, mining machine transportation and deployment of power needs time, mining back to the capital also needs time, a high probability will miss. So it is pure luck that this situation is encountered.

b. Around the computing and energy parameters, choose the right chip (decide A and g), find lower cost energy and optimize the overall solution to reduce energy consumption (reduce e) is the overall reduction in computing and energy parameters, to achieve a higher ROI on mining than buying coins directly from the market. According to the current level of technology, although TSMC has started the trial production of the 5nm process and even started the preparation of 3nm research and development, but subject to the limitations of the technical conditions of the new generation of process than 7nm in power consumption than the improvement may only be 20%-30%, and the new process yield is not high plus to recover the cost of flow and production line, so the chip single T computing power price A will be very high, at this time the computing and energy parameters will also be very high. On the other hand, as long as the energy cost is reduced by half, it is equivalent to the next nanometer process to improve two generations, and the corresponding computing power cost A because of the maturity of the second-hand market or even face elimination can become very low, at this time the computing and energy parameters are also very low, is a high probability to achieve R1>R2.

5. Mining: the question of return capital

The reason for putting this issue so far behind is that if the miner is still considering the issue of return on capital, then he may not be suitable for mining, but still a simple analysis based on the arithmetic energy theorem, based on the computing and energy theorem, the return on capital for mining can be expressed as follows:

(where P2 is the desired coin price and A2 is the residual value of the miner)

Miners can project what conditions they need to achieve in terms of coin price and difficulty to get back their capital based on their purchased mining machine situation and hosting mine situation, and then judge based on the coin price trend and computing power growth forecast.

6. The significance of energy and mining for Bitcoin

According to the computing and energy theorem

Under ideal conditions, the total amount of BTC expected to be obtained from mining and buying coins directly is the same for the same time and the same amount of money, and the return on investment obtained is the same.

From the perspective of supply and demand, the left side of the equation can be regarded as the number of BTC available in the primary market, and the right side of the equation can be regarded as the number of BTC available in the secondary market. The validity of the equation under ideal conditions and the fact that the two fluctuate within a certain range under realistic circumstances can prove that mining is a means to dynamically regulate the supply and demand in the primary and secondary markets, and this regulation is achieved under a completely free choice.

From the entropy perspective, the act of mining has two purposes, one is to complete the bookkeeping task to obtain the bookkeeping reward of new blocks i.e. to mine new coins, and the other is to complete the bookkeeping act to obtain the fee to ensure the security and stability of the whole BTC trading network. Both tasks can be considered as entropy-decreasing behaviors, and therefore require constant energy input to maintain relative stability. The essence of mining is computing, and computing is an entropy-decreasing behavior with electrical energy as input, which perfectly fits the needs of the BTC network. Therefore, in the future, even if all the blocks are mined, the fees will become the entire source of income for miners and can still support the stable operation of the whole system.

From the perspective of liquidity, intra-exchange trading and inter-exchange moving arbitrage bring BTC extremely strong liquidity in the secondary market, while mining can regard BTC as the product of using electricity as the raw material for production, and mining confers global liquidity to electricity owners. The idle electricity resources (nest electricity, abandoned water and electricity, clean energy) around the world, which were limited by incomplete infrastructure and unable to create more value, can now be converted into BTC to obtain global liquidity and the corresponding liquidity premium, thus getting higher returns. At the same time, the cost of energy varies around the world. BTC regulates supply and demand through the price of coins in the secondary market and the arithmetic power in the primary market, and mining is equivalent to energy arbitrage for miners to motivate them to continuously search for quality energy with lower cost. The high cost supply is eliminated and the low cost supply receives more revenue to reach a new equilibrium.

IV. SAI Products & Computing and Energy parameters

1. SAI Company Introduction

SAI is the world’s leading chip energy technology service provider, with a three-dimensional technology matrix of chip-level liquid cooling, chip waste heat utilization, and clean energy generation to reduce the energy cost of supercomputing and computing power cost, and realize the vision of the era of energy-driven computing and computing-driven world.

We provide supercomputing equipment hosting and mobile cabinet sales services, and our core product SAIHUB computing and energy center has three outstanding performance of liquid cooling, waste heat utilization and mobile assembly, which can effectively reduce energy costs by more than 50%.

2. SAI products introduction

The core product, SAIHUB, has multiple product models that can adapt to a variety of scenarios and can significantly optimize the computing and energy parameters for mining, shorten the payback period, extend the life of the chip, and improve the return on investment.

l SAICAB, replaces the fixed infrastructure of a traditional mine and the heat dissipation structure of a miner with a lower cost, while having mobility and ease of transportation, and a high reusability rate. Enables miners to use lower cost bare electricity, reducing electricity costs e.

l SAICAB provides liquid cooling heat dissipation that is 20 times stronger than traditional air cooling, providing better working environment conditions for the chip, effectively improving arithmetic power, reducing power consumption ratiog and unit arithmetic cost A.

l SAICAB can collect the heat emitted by the miner, and in the heating season, it can heat the chip with waste heat to obtain additional revenue to flatten the electricity bill; in the period of abundant water, it greatly saves the power consumption used for heat dissipation, further reducing the electricity cost e.

l SAICAB is compatible with a variety of mainstream mining machines, you can choose the mining machine with higher ROI mining according to the calculation energy parameter SAI and combined with the market situation, to get the real lower cost BTC with a higher ROI than directly buying coins.