The Bitcoin network at risk

In this post, I sketch out a description of a possible existential risk for Bitcoin, that has arisen from the exponential increase in price seen in 2017 and the abrupt fall that followed.

An AntMiner S9 from Bitmain

Network hashrate and mining profitability

For Bitcoin miners, Bitcoin is a system that turns electricity into money. This magic trick is performed by operating mining hardware, specialized chips (ASICs) that compute the Bitcoin hash function (SHA-256) in a fast and energy-efficient fashion. A popular device is the AntMiner S9, manufactured by Bitmain, pictured above.

Miners contribute hashing power to the Bitcoin network, increasing its hashrate, and probabilistically receive rewards in exchange (the exact mechanism is described in Satoshi Nakamoto’s whitepaper, a short and simple read that is an absolute prerequisite for any discussion of Bitcoin). Bitcoin mining is a market: all miners are competing for the same pool of rewards (the transaction fees paid by Bitcoin users, plus a fixed amount of Bitcoin unlocked when a miner is the first to verify a block in the blockchain, a group of transactions). The higher the fraction of the total hashrate you contribute, the higher the fraction of that pool of rewards your receive.

Economics 101 states that in a perfect market, the long-term profitability of any competitive activity is zero. If the Bitcoin mining market were sufficiently competitive, then Bitcoin mining would have near-zero profitability. In this equilibrium situation, if the price of Bitcoin were to rise by a bit, this would at first increase profitability for existing miners, but it would also increase the number of market actors with access to sufficiently cheap electricity to profitably mine Bitcoin, and these new actors would immediately start exploiting this opportunity, which would result in an increase in the difficulty of mining (the amount of hashing power needed to unlock a fixed amount of rewards), thus erasing the initial increase in profitability. Likewise, if the price were to fall by a bit, the hashrate would dynamically decrease, since the amount of electricity that can be profitably consumed by miners would be less, and profitability would keep hovering a tad above zero.

Thus, at equilibrium, the total hashing power expended over a given period of time would correspond to the amount of electricity you can buy for the dollar value of the Bitcoin mining rewards over that period. In this configuration, the miners would be the people with access to the cheapest electricity, and the total hashrate would be fixed by the current Bitcoin price.

A side-effect of this dynamic is that there is no ceiling to the amount of electricity that the Bitcoin network may consume during a price bubble: if the price of Bitcoin is sufficiently high, the Bitcoin network will start consuming a ridiculous amount of power. Throughout 2018, Bitcoin mining has been accounting for about 0.5% of the world’s electricity consumption, a rather odd figure given that Bitcoin is powering a number of transactions per day that’s about 0.05% of the number of transactions processed by credit card companies (and that other 99.95% isn’t consuming much power at all compared to Bitcoin). At the hashrate peak, a single Bitcoin transaction was consuming as much power as an average household during an entire month.

The great mining race of 2018

In the past year, Bitcoin mining has been highly profitable, because mining was not a very competitive market at all. The barrier to entry was the acquisition or manufacturing of mining hardware. The rise of the Bitcoin price in late 2017 was meteoric, and the production of mining hardware took a while to catch up. This has led to the momentary rise of mining companies, often operating out of China (due to access to cheap electricity and proximity to manufacturers). Bitmain, one of the world’s foremost manufacturer and operator of mining hardware, turned a net profit of $742M in the first half of 2018.

Here’s a snapshot of what has happened to the Bitcoin hashrate and to mining profitability over 2017 and 2018: mining profitability started by spiking wildly along with the price, then gradually decreased as more mining hardware went online. As a result, the total hashrate was steadily increasing, and so was the electricity consumption of Bitcoin.

Data from bitinfocharts.com

Late 2018 and 2019: a new equilibrium

In recent days, the price of Bitcoin has been falling abruptly, to a level where a lot of the mining hardware that was produced and put online in 2018 is no longer profitable to operate. The Bitcoin hashrate has nearly halved over this period (from a high point of 62 million TH/s to a low of 34 million TH/s a couple of days ago), meaning that almost half of the total mining hardware that was operating in October has been brought offline. Because mining hardware is single-purpose (it can only be used to mine Bitcoin, you can’t even crack hashed passwords with it), that hardware is now almost completely worthless. In particular, anything less energy-efficient than a S9 won’t find any buyers other than electronics recycling shops. Rumors are that you can now buy mining hardware by the pound in China, where most of the miners were operating.

Notably, this means that Bitcoin mining has suddenly become a perfectly competitive market: since hardware is no longer scarce, anyone can start mining. When the price of Bitcoin stabilizes, we will see a new equilibrium where the network hashrate is correspondingly much lower, and mining profitability will be near zero.

Feasibility and profitability of a 51% attack

The mining market (hashrate, difficulty, profitability) will just adapt dynamically to the new price levels, so all is well for the Bitcoin network, right? Not quite. As the price of Bitcoin keeps falling, the fraction of the total outstanding hardware that is offline and up for grabs at ridiculously low prices will increase. Now, it’s probably already over 40%, but it might be 90% in the near future (if the price falls to $1000). This is a problem for Bitcoin, because it means that a “51% attack” on the Bitcoin network becomes feasible, and potentially profitable.

A 51% attack happens when a single actor accounts for more than 50% of the total hashrate of the network. This actor would then have a significant amount of control of the Bitcoin blockchain, the ledger that records past Bitcoin transactions. Such an attacker could, for instance, revert their own transactions, allowing them to spend their Bitcoins arbitrarily many times. By doing so, the attacker would considerably damage Bitcoin, as this would destroy the trust that users put in the network. This could potentially make Bitcoin worthless altogether.

If 90% of the total available mining hardware is offline and for sale at a heavy discount, you just need to acquire ~11% of that hardware pile to be in control of an amount of hashing power equivalent to the other 10% of mining hardware that is actually online. Thus, by buying up one out of ten of the discarded S9 miners and plugging them in, you could start controlling the Bitcoin network. Note that the figure of “90% of mining hardware offline and for sale” would correspond to a Bitcoin price of about $1000. Consider that the last time Bitcoin had a price of $4000 (current price), the hashrate was 10 times lower than at the peak, and last time Bitcoin had a price of $1000, the hashrate was 20 times lower. That’s a staggering differential in available mining capacity.

Is such a 51% attack a plausible scenario? I believe it may soon be logistically feasible (as soon as more than half of all available hardware is offline), and it will certainly become highly realistic if we reach $1000 per Bitcoin. However, just because it is feasible does not mean that someone is going to do it. In particular, a 51% attack may not be profitable at all, because it would likely be detected quickly, and once detected, the price of Bitcoin and the utility of the Bitcoin network would quickly drop to near-zero (since its value derives from the fact that it implements a trusted ledger of transactions: by losing this trust, you lose everything). For instance, for most altcoins, 51% attacks have long been perfectly feasible and even relatively cheap, but few have ever been attempted. This suggests that no one would attempt such an attack, however feasible, as a means to make a quick buck.

However, someone with a bit of capital available could well decide to go for it purely for the sake of disrupting Bitcoin. A state actor, for instance. I personally don’t think that’s very likely, though, because it would still not be an easy operation, and the benefits would be unclear. The perception of subversiveness of Bitcoin has been dropping after the 2017 bubble, along with its relevance to illegal payments and dark finance. It is unclear who would bother trying to destroy it.

All in all, I don’t think the scenario described here is particularly probable, but its clear feasibility may weigh on Bitcoin like a sword of Damocles.

Conclusions

  • Bitcoin mining is a market: hashrate, difficulty, and profitability dynamically adjust to each other, with profitability converging to zero as competition increases.
  • At equilibrium, the hashrate of the network corresponds to how much electricity can be bought for the price of the mining rewards: the hashrate is set by the price of Bitcoin, the price of the cheapest available electricity, and the efficiency with which mining hardware converts electricity to hash power.
  • Throughout 2018, as the price stayed above $6000, we saw a huge build-up in mining capacity (20x compared to the start of the bull run). As the price has now fallen below the profitability line for many miners, a lot of this capacity is being turned off and is virtually worthless.
  • As a consequence, hardware is no longer scarce and mining has suddenly become a highly competitive market with virtually no barrier to entry. This guarantees that profitability will hover right above zero for the foreseeable future.
  • As the price keeps falling, the fraction of outstanding capacity that is offline may reach up to 90% or 95%, which means that a bargain hunter could relatively easily assemble a mining fleet that would be sufficient to perform a 51% attack.
  • Even if such an attack never materializes (due to the lack of benefits relative to the logistical difficulty), its feasibility will weigh on the future prospects of Bitcoin as a payment network.