Let’s start with the question that everyone is talking about:
Yes — Bitcoin miners are shutting down hardware and exiting the network.
No — this is not going to cause a “death spiral.”
Since the start of 2018, the price of bitcoin has declined from a peak of nearly $20,000 to approximately $4,000 — an 80% fall. Miners receive their income in bitcoin and thus depend on exchanges — and exchange prices — to cover their expenses and (hopefully) turn a profit.
So what does this all mean for the network?
Before we go there, let’s ‘level set’ with a brief primer on CAPEX, OPEX and ROI for non-MBAs.
For those unfamiliar with the lingo I am about to employ, when I use the term capex I am referring to capital expenditures, and when I use the term opex I am referring to operational expenditures. Briefly explained:
Capex describes all expenditures related to the acquisition of capital, such as mining gear, racks, property (if applicable) etc.
Opex describes all ongoing expenditures, like wages, electricity cost, rent (if applicable) etc.
I will also be using the term ROI which simply means return on investment. ROI is positive if an investment is profitable, and negative if it produces losses.
When establishing a mining operation you can structure it between two extremes:
On one end of the scale, you can rent all your equipment, housing, maintenance, etc. and pay-as-you-go for electricity. Under this setup, all expenditures are opex and your operation owns no capital. If nothing is contracted, you have no assets and no liabilities.
On the other end of the scale, you could decide to buy all of your gear, a plot of land, and fixed electricity and employment contracts. In this structure, all expenditures are pre-paid capex. Your operation owns capital (assets) but also has contractual liabilities.
In reality, most mining operations fall somewhere between these two extremes — a combination of both opex and capex.
For example, some may own gear, but rent space and pay as they go for electricity. Others rent gear, own land, and buy electricity on short contracts. The point is that many miners have both capex and opex components to their cost function, and our assumption is that on average, most mining operations have a bit of both. More on this later.
Mining gear, like most other productive capital has a finite lifetime and will see its productivity deteriorate over time until it is either worn out or obsolete. This is the concept of depreciation. In practice, it means that mining gear must generate more free cash-flow over its lifetime than its acquisition cost plus ongoing capital costs in order to be profitable.
The value of the gear you have is therefore understood to depreciate from its acquisition cost down to either scrap cost or zero.
If your gear is productive for a long time, it has a long depreciation horizon; conversely if it is only productive for a short time, it has a short depreciation horizon.
With longer depreciation schedules, you have more time to spread out the total purchase price.
Depreciation is booked as an ongoing cost in accounting, but does not impact actual cash-flow as it is normally “front-loaded”— or in colloquial terms, prepaid — and therefore does not impact cash-flow, or by extension, cash cost.
Free cash-flow here refers to cash-flow that exceeds all of your opex. Cash cost, or the cost of capital, refers to the return you would have received if you put your money into a “risk-free” (lol) investment like US Treasury notes instead of investing in mining.
For miners, there are two cutoff levels for bitcoin prices that matter:
The first is their all-in ROI breakeven level, above which they make a profit on their investment, and below which they make a loss on their investment.
The second is their cash-cost breakeven level, above which they are cash-flow positive but still potentially loss-making (ROI can still be negative if they never make enough cash to cover what they paid for their mining gear); and below which they are cash-flow negative and thus — depending on their industry view, risk appetite and capital levels — are likely to shut off mining gear entirely.
This is an important distinction because miners — even if they are realizing a negative ROI and unable to recover their capex — will keep mining for as long as the gear is returning any positive cash-flow in order to recover as much of their capex as possible.
As soon as opex exceeds income and mining gear becomes cash-flow negative, it no longer makes sense to even leave on, as it is now just burning cash.
This is when miners remove gear from the network — not at the point of negative ROI.
In our report we calculate a market-average all-in breakeven cost for creating one bitcoin. This is our best approximate for the bitcoin creation cost above which the market-average miner will make a positive ROI.
At the risk of repeating myself ad nauseum here, this is not the level below which miners will turn off their gear.
It is simply the level below which they will lose money on their investment, making them unlikely to remain players in the mining industry over time, unless they have bottomless pockets of investment capital available.
Now, the results… The Cost to Mine One Bitcoin: Average ‘all-in’ cost: $6,800/btc | Average cash cost: $3,400/btc.
In our June report we estimated a market-average all-in cost of creation of approximately $6,500 per bitcoin. We arrived at this number by assuming a market-average capex based on all available pricing information, electricity cost of ¢5/kWh and an 18-month depreciation schedule (for a full treatment of the methodology and all assumptions we direct readers to the appendix of our June report).
I need to stress that this is an average figure. Not all miners are operating at these assumption levels. Some have it better, some have it worse.
The figure itself means that if our assumptions are correct, the average miner, at the network conditions present in June, would run a positive ROI at bitcoin prices above $6,500. The bitcoin price at the time was approximately $8,500.
In our latest report, under the same exact assumptions, we estimate the creation cost to now be approximately $6,800, an increase of $300. The current bitcoin exchange price is about $4,000 — an entirely different situation altogether.
For reference, we estimate that the opex component of this all-in cost is exactly 50% of the total (at an 18-month depreciation schedule). That means the market-average cash cost at our assumptions is approximately $3,400. Pretty close to current bitcoin prices.
So what does that mean? Essentially one of two things:
Either our assumptions are silly, or…
Many miners are currently feeling the squeeze, with inefficient mining gear and high-cost electricity miners likely to be forced off the network.
What does the data say?
Interestingly, the data suggests that the truth lies somewhere between the two — cliché, I know. But let’s unpack that a little.
After peaking at a 2016-block average of almost 55 Exahash per second (EH/s) in late September, the hashrate has since fallen to approximately 40 EH/s, triggering the largest difficulty decrease in more than five years at the last adjustment.
At the time of writing, Bitcoinwisdom projects the next downwards adjustment to be even larger. The combined back-to-back decrease would be one of the largest in Bitcoin history, certainly the largest since the advent of large-scale professional mining around 2014.
Clearly, some miners are struggling. They have gear that is running below cash cost which means this gear is now being shut off.
It is also possible our market average assumed electricity cost of ¢5/kWh could be too high. Or perhaps our cooling cost assumptions are too high? While we can speculate about this, that is also ultimately all we can do.
Bitcoin is structured such that the hashrate follows price, slightly modulated by increases in gear efficiency. When the price increases, the hashrate increases, and when the price decreases the hashrate decreases.
Mining cost will always tend towards the price of bitcoin minus a narrow competitive margin. However, these dynamics are not instant, and there is an asymmetrical delay in the trailing effects.
Like any other capital-driven industry, the delay in the upwards drag results from the time difference between making an investment decision and when the gear is actually switched on.
For most players in bitcoin mining, this is on the order of months and depends on their proximity and relationships with the producers of mining gear.
But even for the producers themselves, there is significant delay. Chips must be ordered from the foundry; units must be assembled, shipped and installed. Only at the end of that process does the hashrate actually increase. In the meantime, the price could have increased even more, and much more rapidly than new gear could have possibly been employed.
The same dynamic does apply in the opposite direction. When cash cost falls below breakeven, there are no barriers preventing miners from immediately pulling the plug on their gear, meaning that mining gear can be shut off immediately in response to falling prices. (Side bar: the notable exception to this are mining operations that bought fixed supply electricity contracts, thus forcing them to mine until they are insolvent).
Hashrate will therefore lag price increases on the order of months, but respond much quicker to decreases in price.
“But I thought price follows hashrate?”
It doesn’t. And moreover, how could that possibly be the case?
Miners are compensated in bitcoin, but incur costs in their local currency. Under a steady hashrate marketshare, the bitcoin exchange price is directly proportional to their payout.
At average competitive conditions, increasing your hashrate in a falling market will only make you lose money, as mining costs will increase in line with the difficulty increases caused by increasing hashrates.
Only when prices increase can the hashrate increase — in excess of efficiency gains — to compensate for increasing difficulty and costs.
In fact, I can only think of one scenario where the hashrate could act as some sort of ‘floor’ for prices. This is in the unlikely event where miners are so well capitalised that they could refuse to sell their bitcoins at market prices, and cover their costs with liquid capital from their balance sheet while continuing to mine at a loss. You be the judge of that likelihood.
So is the mining industry collapsing now?
No. There is nothing dramatic about what is currently happening. The net effect is that the highest marginal cost producers are booted off the market while the most efficient miners remain.
Difficulty resets to a lower level and the all-in cost of mining falls to a level where it is again right below the price of bitcoin.
The remaining miners then restart a new competitive cycle — both against each other, and against a new wave of prospective outsiders who believe they can mine profitably under new conditions. Through this process, mining migrates ever closer to the cheapest underlying conditions. It’s an incredible spectacle of pure free-market dynamics.
Meanwhile, bitcoin issuance effectively remains the same. No fewer bitcoins are created, just as no more bitcoins were created during the period of price growth (with the exception of single-digit perturbations caused by the relative over- or- underperformance caused by the growth or shrinkage happening between difficulty resets).
This is a defining characteristic of bitcoin mining and sets it apart from all other commodities — it is also a fundamental driver of volatility.
Unlike other commodities where the output is modulated by price, the bitcoin issuance cannot change.
If the price of gold increases, production will increase until the marginal production cost again equals the market price (minus transport costs, which for bitcoin, are negligible compared to physical commodities). If the price falls, production will decrease until the same condition holds. This dampens volatility by increasing supply in rising price markets and reducing supply in falling markets.
No such effect in Bitcoin. Issuance is predefined and no market dynamics can significantly influence it.
So what happens next?
As price does its thing, hashrate will follow and settle into whatever new market conditions are in stall. Old, inefficient gear and high-cost producers are out; and until price increases again, the hashrate can only increase by miners lowering their opex. They can do this by sourcing cheaper electricity, installing more efficient mining equipment or generally cutting costs.
There is no such risk of mining collapse or any other click-bait nonsense you might read out there.
The dynamic difficulty adjustment takes care of that by regularly resetting the difficulty so mining costs fall in tune with bitcoin’s price.
For such a collapse to occur, the bitcoin price would need to immediately plunge to near zero; thus triggering virtually the entire mining network to shut down; and therefore preventing the requisite blocks to reach the next difficulty reset from being mined for months or even years.
Perhaps possible, but not a likely scenario in our view.
Bitcoin isn’t dead. Not this time nor the 326 times before.
