Written by: Leo Zhang and Karthik Venkatesh, founders and data analysts of Anicca Research, a research organization for computing power and derivatives, respectively
When an event has subjective participants, the subject is no longer limited to facts, but also includes the participants’ opinions. The chain of causality is not directly caused by individual facts, but is triggered by facts and fed back from perception to facts.
Financial crocodile George Soros “Financial Alchemy”
In the previous article, we discussed the framework of computing power as an asset class. Everything is related in crypto mining. In order to fully understand the dynamics of the computing power market, we need to study the relationship between the variables that affect computing power.
In this article, we first define the mining market cycle as four basic stages, each with different price trends, hardware capacity and market sentiment. We studied the main driving factors in the market in each context and demonstrated the role of hardware reflex arcs and the reflexivity of computing power in shaping these macrocycles.
Through a series of case studies and theoretical demonstrations, we intend to introduce a guiding framework to understand the different investment environments in mining. At the end, we discussed the broader significance of transaction fees for the growing revenue of mining. Based on the new opportunities in the transaction fee market, and how the fee rate as a major variable will profoundly change the structure of the computing power market.
Four seasons of the computing power market
The dynamics of the computing power market is driven by the complex linkage relationship between external and internal factors such as price trends, reflexivity, hardware reaction time and handling fees. Although the logic connecting them seems clear, the randomness of each variable makes it very difficult to build a generalized model.
For this reason, sometimes the macroscopic phenomenon shown by the market will appear illogical, as if price and hash rate are in a completely different time frame of reference. Nevertheless, the actual profitability of miners can be traced and determined. Based on how historical mining income has evolved in different market environments, we can identify the basic patterns of prosperity and depression in the mining cycle:
Source: Bitcoin data in CoinMetrics
Taking the specification parameters of the Antminer S19 Pro as an example, the cycle stages are divided by the vector of price and network hash rate moving in different directions at different rates. The mining income will change in different environments:
The Rising Bull
Token price growth exceeds hash rate growth
When the growth rate of mining difficulty lags significantly behind the price increase, mining is the most profitable. The “rising bull market” phase usually occurs after a long period of relatively weak volatility. Prices have just begun to form momentum, and most of the market is still uncertain about the next direction. The growth rate of computing power is much lower than the rate of price growth. The increase in computing power is mainly attributable to miners who expect price increases or miners who can get very low electricity bills. For example, between January 2019 and April 2019, the BCH-BSV “Hash Power War” overlapped with the dry season, and the price of Bitcoin was suppressed. Resourceful miners bought cheap second-hand mining machines on the secondary market. Some also use synthetic mining contracts or cloud mining to establish positions at low prices.
Although Bitcoin prices are on the rise at this stage, sometimes environmental factors may even cause the hash rate to drop. It is usually related to physical conditions, such as extreme weather and floods forcing large mines to go offline. The floods in the rainy season in Sichuan in 2020 are particularly destructive. But these are temporary setbacks, which usually recover over time.
Another special situation that may cause a decrease in computing power is a hard fork initiated by a developer. After Bitmain first announced that its ASIC chip mining machine could mine Monero, Monero developers decided to switch the algorithm every six months. Each time the network changes its algorithm, part of the network hash rate will drop. The hard fork initiated by the developers is not just a phenomenon in projects that are hostile to mining machines. Sia developers are very accepting of ASIC mining, but they specifically removed Bitmain and Jianan Zhizhi ASIC mining machines from the network through a hard fork.
Source: Monero data in CoinMetrics
This special situation may temporarily prevent the increase in hash rate, but as the overall upward trend continues, the positive sentiment of participants further strengthens, and the demand for computing power also increases.
Mining Gold Rush
The price of tokens is rising rapidly, and the growth rate of hash rate is rising
Once the bull market pattern is confirmed, people will be more eager to buy mining machines. The new mining machine was sold out almost immediately. Large miners placed generous orders to mining machine manufacturers to strive for priority in supplying them. In this article, we described the correlation between the pricing of mining machines and the static days-to-breakeven based on static days-to-breakeven (chain news note: China is generally called static days-to-breakeven). The shorter the return days, the higher the price set by the seller for the mining machine. The price of tokens rebounded rapidly, followed by the demand for new mining machines, but the growth rate of the whole network’s computing power has not yet accelerated. This period is a window for mining machine manufacturers to obtain astronomical profits. The secondary market for mining machines and the cloud mining market are also trading at a premium.
This is true for ASIC chips and GPU graphics mining. From 2016 to the end of 2017, major graphics card manufacturers AMD and Nvidia benefited greatly from the rapid development of Ethereum. Miners are willing to pay the highest bid for each GPU available on the market. The GPU supply shortage was very serious at one time, and Nvidia even considered requiring retailers to limit purchases of less than 2 pieces per buyer. In the current market, the DeFi boom has once again triggered people’s strong interest in Ethereum.
Bubble hype can easily create a “panic FOMO” among miners. The positive price bias continues to strengthen itself, and expectations rise faster. Alternative currency projects that have experienced this stage for the first time with a relatively short birth time may attract the attention of ASIC mining machine manufacturers, such as Handshake and Filecoin, which have become popular recently.
In early 2019, rumors about Grin’s investment of hundreds of millions of dollars spread like wildfire. Venture capital, VC, scrambled to provide funds for companies specializing in Grin mining to purchase and operate GPU miners. Soon after the project was launched on the mainnet, the mining difficulty rose sharply. Mining machine manufacturers such as Xindong and Obelisk competed to build the first Grin mining dedicated ASIC mining machine. History has proved that the project has never matched the expectations of the hype, and the corresponding ASIC mining machine has never received enough orders to put it into production.
Inventory-flush
Token prices are falling, and the hash rate growth rate remains strong
As Howard Marks, the founder of Oaktree Capital, said: “Anything that generates abnormal profitability will attract incremental capital until the capital becomes overcrowded.”
After the coin bull market and mining machine manufacturers overproduced machines, it is not uncommon for mining machine inventory dumping. In 2017, mining machine manufacturers such as Bitmain misjudged the market development direction and produced too many mining machines during 2018. They have to gradually reduce the price of mining machines to dump inventory. In order to clear the stock of excessive chips, Bitmain even launched products with extremely low market demand, such as home Wifi routers that can be mined. As a result, despite the drop in Bitcoin prices, computing power continued to climb for several months until profit margins were fully squeezed out.
During the same period, due to the exponential increase in hash rate competition, many GPU mining farms became unprofitable. Altcoin ASIC mining machines (Monero, Zcash, Sia, etc.) were released into the market, while altcoins The price has fallen all the way. The bear market hit the hardware supply chain so quickly that they had little time to react. Nvidia released a disappointing financial report. Its founder Jen-Hsun Huang’s argument changed from “cryptocurrency will become an important driving force for our business” at the historical high of Bitcoin in 2017 to “I don’t want anyone to buy it. Cryptocurrency, okay? Stop. Enough. Don’t buy Bitcoin, don’t buy Ethereum.”
Inventory dumping due to overproduction has occurred in many markets with high response delays. For example, about 10 years ago, due to the large sums of overseas buyers, there was an epic bull market in luxury apartments in New York City. Developers are eager to start new projects. In recent years, due to various reasons (such as capital control), purchasing power has gradually dried up, but the newly built luxury apartments have only just been put on the market. As a result, the empty real estate fell into the hands of the developer.
The Shakeout
The price of tokens has fallen and the hash rate has dropped
Sometimes, mining revenue will drop below a threshold, and it has been unprofitable for miners who persist. Chinese miners call it the “shutdown price.” In the traditional market, when there is a price correction, the negative bias will start to snowball, bringing the price into a downturn. However, since the computing power rate is self-referenced, the more computing power that is eliminated from the market, the “richer” the remaining computing power.
In Bitcoin, this excessive price correction tends to be short-lived. Miners’ expectations of future mining revenues buffer the decline in computing power. They believe that the chance of recovery is very high, so they are willing to persist in mining at a loss, and even buy new mining machines even when the market is shuffled out. On the other hand, if the network is flooded with speculative miners, such surrendering out will happen frequently.
Things that underperform for a long time will eventually appear cheap. As the American economist John Kenneth Galbraith said that “financial memory is extremely short,” cycle changes will repeat itself again and again.
Plato’s prediction of market fundamentals
Why does the computing power market exhibit these cycles? Intuitively, the increase in computing power is positively correlated with price trends. But why does the change in token price not bring about a commensurate adjustment of computing power? In other words, why is the computing power market inefficient?
Conceptually, the market is an information aggregation device that can distill the views of participants into price information. The faster the price absorbs new information, the more efficient the market. When the theoretical balance is reached, the difficulty of network mining should converge to the level where most miners are close to the break-even operation.
Satoshi Nakamoto wrote in a BitcoinTalk post many years ago: “The price of any commodity tends to fluctuate around the cost of production. If the price is lower than the cost, then the production will slow down. If the price is higher than the cost, then It can be profitable by producing and selling more products. At the same time, the increase in output will increase the difficulty and push up the production cost and move closer to the price.”
In the current market, the price of Bitcoin is far from a passive reflection of production costs. In reality, we rarely see this kind of balance envisioned by Satoshi Nakamoto.
For most physical commodities, supply is mainly determined by production and consumption demand, but speculation causes cryptocurrency investors to make decisions based on expectations of future prices rather than the current supply and demand curve. Therefore, a simple calculation of mining costs can hardly provide insight into the market.
Market participants always carry their own biases when dealing with new information. This is similar to guessing the shape of a high-dimensional object through its projection in a low-dimensionality. This is a Plato’s fable of market fundamentals.
Error-prone cognition brings reflexivity. Reflexivity is an iterative process: the market, as a melting pot of prejudice, is always flawed in reflecting reality. When investors bet on the market, price changes begin to affect market fundamentals (for example, the company’s capitalization amount goes up or down), which in turn affects prices, thus forming a reflexive feedback loop.
Instead of focusing on hypothetical results, it is better to study the process of change. After years of development, the theory of reflexivity has gained mainstream favor. Researchers have conducted extensive reflexive observations in the stock, currency, cryptocurrency, and even mining markets.
Reflexivity of computing power
How does reflexivity behave in the computing power market?
It is well known that the demand for hash rate is driven by the value of the tokens it generates. The buying and selling decision is based on the participant’s own biased expectations of future mining revenue. Equity investors set expectations for future prices through macro, industry and company analysis. Hash rate investors set expectations for future mining revenue by evaluating the trend of token prices, handling fees, and network hash rate growth.
Everyone has their own (usually flawed) judgment of price trends. It is much more difficult to build a forecasting model for computing power growth. One of the reasons is that it is dynamically recursive: the more the hash rate floods into the market, the higher the dilution of the hash rate a unit miner has. Changes will lead to adjustments in expectations, and therefore will recursively affect current mining revenue. Every participant in the computing power market is constantly changing the rest of the market.
This means that the most scientific way to predict the increase in hash rate is to collect mining machine sales data from mining machine manufacturers, large miners, service providers and distributors. However, information asymmetry is a major feature of the mining machine manufacturing industry. It takes a lot of effort to obtain accurate and updated data. Since it is difficult to reliably set expectations for hashrate growth, and transaction fees are not particularly important in the proportion of mining revenue, expectations for the future price of tokens have naturally become the main variables in the development of the mining industry. After all, if people are not optimistic about future prices, why would they spend so much capital and energy involved in mining?
Collecting mining data is an arduous task, but is it possible to establish a quantitative model between the trend of token prices and the growth of computing power?
As demonstrated by the four phases of the market cycle, we often see the divergence of computing power and token price trends. Information in the capital market spreads rapidly. Hardware manufacturing and mining machine shipments are very slow. The hash rate market and the idealized efficient market hypothesis show the opposite trend. This makes pure correlation analysis useless. We need to review the data on different time scales.
Recently, the digital asset financial service company BitOoda published a comprehensive research report (), they found the lag time of the change in the hash rate relative to the increase in the price of the token on the change in the price of the token and the change in the hash rate in the past year Approximately 4-6 months.
Please note that this lag time is not fixed. Depending on the production capacity and supply availability of the secondary market for mining machines, the lag time varies with each market. The corresponding time of different blockchain networks is also different.
Taking Litecoin as an example, the reaction of its mining hash rate to price changes from January to May 2018 has a long lag time. The Litecoin price and hash rate changes became very “synchronized” after July 2018.
Source: Litecoin data in CoinMetrics
Extending this analysis to the corresponding lag study of Bitcoin, ETH, and Litecoin over a longer period of 2017-2020, we found that the average response time was 60-120 days, 30-60 days, and 15 days, respectively.
Analysis process:
1. 15-day data summary
- The columns in the data set are date (15-day end date), 15-day average price, and 15-day average hash rate
2. Calculated for each 15-day period:
- 15 days, 30 days, 45 days… Percent change in price after 180 days
- 15 days, 30 days, 45 days… Hash rate percentage change after 180 days
3. Calculate the correlation between price changes and hash rate changes in different time periods
4. The matrix is read as: the correlation between the price change in “y” days and the hash rate change in “y” days (after the price change in x days)
Response time is essentially not good or bad. It is a function of the availability of mining machines in the market for a given period of time. Generally speaking, the response time of smaller blockchain networks based on general mining hardware is much shorter. Since the loyalty of miners on these networks is low, their computing power can respond to price increases and decreases faster. Compared with ASIC miners, they can easily switch to other network mining when it is profitable. Some mining hashrate pools provide automatic switching services that constantly jump around between multiple different networks to maximize profits (called “profit switchers” or “machine gun pools”).
Please note that blockchain networks dominated by ASIC mining machines are not necessarily value-driven (such as Litecoin after 2019); GPU-dominated mining networks are not entirely speculative (such as Ethereum).
In most cases, the change in token price is earlier than the change in computing power. Sometimes we can observe the opposite in the altcoin market, usually altcoins that are about to undergo a halving. The halving is one of the recurring and self-fulfilling prophecies in cryptocurrencies. The anticipation of the token rise after the halving has prompted miners to deploy new mining machines in advance. Sometimes, coin speculation organizations that work together will deploy computing power to accumulate enough tokens, and then operate to push up the price of tokens in order to obtain final benefits.
This model is also common in GPU miners speculating on new projects. After the issuance of the new tokens, most of the tokens are exclusively traded on the OTC market, with extremely poor liquidity. Miners do not have a good exit channel for positions. They continue to operate at a loss until project development gains momentum. With the development of the community, the tokens of the project are listed and traded on larger exchanges, giving miners a chance to earn some income.
The increase in the hash rate does not guarantee a future increase in the token price. This is a high-risk bet, and there are countless examples of failures. Many factors are needed to coordinate to ensure that this process goes smoothly. In each step of the following process, problems may occur:
Various GPU mining projects were very popular between 2017 and early 2019. Some analysts said that the simple agreement of future token SAFT issuance by the proof-of-work project to venture capital is fairer than ICO. How to issue coins more fairly is a very broad issue. Even in the DeFi field, which has nothing to do with PoW, it is a controversial topic. The coin issuing mechanism and the obtained computing power cannot guarantee future price increases. In essence, this is a form of ICO with a high barrier to entry, which is the same as a casino game that throws darts in the dark.
The response time of mining hardware (regardless of length) is a source of endogenity. This means that when modeling the impact of price on hash rate growth (and vice versa), the impact may be underestimated or overestimated. Therefore, using model-based inference decisions as one of the input values of investment decisions may bring catastrophic consequences.
The main point of this article is that this connection between computing power and token price does not mean that there is a causal relationship between the two. One will not automatically induce the other. The expectation of future mining revenue and the expectation of hash rate increase mutually strengthen each other.
The influence of transactions continues to increase
To further refine the macro model at the beginning of this article, the change trend of handling fees should also be the main variable. As mentioned above, current expectations for mining revenue are mainly driven by token price trends. In August of this year, Ethereum miners made a total of $113 million in profits. The previous record (US$64 million) was in January 2018. The continuous increase in the on-chain traffic of the DeFi project can explain the huge increase in the income obtained by the miners from the commission.
Taking transaction fees as a key variable opens up possibilities for new profit models. For example, the arbitrage opportunity in the decentralized exchange in Ethereum encourages competitive automatic market makers to continuously increase their bids in the Gas priority auction bidding. Miners who control the ordering of transaction orders can profit from these auctions by optimizing the ordering of gas fees. This is part of the MEV that can be extracted by miners. MEV is the value that miners can obtain directly from smart contracts. There will be more services and infrastructure projects (such as the Taichi network of the Spark Pool that can improve transaction broadcasts), which can solve all aspects of the new rate market.
As the proportion of transaction revenue in mining revenue continues to increase, the calculation of mining revenue adds a new dimension. Both price expectations and fee expectations will affect the expected future mining revenue:
The theory of reflexivity is an effective way to understand the ebb and flow of computing power. However, this model cannot replace the understanding of the basic loopholes in the computing power market. As the mining industry becomes more industrialized, capital expenditures will inevitably increase. At the same time, the percentage of commissions in mining revenue will increase, and the four basic stages of the mining cycle will expand to even more complex scenarios. This comprehensive impact will bring more challenges and uncertainties to the cash flow management of miners.
After 10 years of development, the computing power capital market is still plagued by the lack of standard contract terms and pricing ideas. The industry needs appropriate risk management practices and mature market mechanisms to ensure continuous long-term investment in computing power.
In the next article, we will discuss in depth the risk management framework, innovative financing and hedging strategies, and the long-term impact of financialization on the mining industry.
