A user, Bob, on Bitcoin Talk sent me some question in response to my article:
1) why does one assume transaction fees will substitute decreasing block rewards. are users not equivalent in choosing the operating software in the Bitcoin network why not just charge a mandatory 0.01% charge on all coins that are younger than X blocks and limit free transactions to say 30%. When the thermodynamic limit is achieved economic friction in Bitcoin will be just 0.01% of all energy consumed in the the economy.
2)the thermodynamic limit is not a limit in that there is entropy, the waste heat just becomes an input for a new system, eg. water can be desalinated, cooling as we know it today is a wasteful activity.
3) I also don’t seem to understand why when ASIC chips reach the thermodynamic limit they’ve wouldn’t start decentralizing in location (not ownership). There will always be centralized mining where energy is cheep but there will always be a need for heating at the very least everywhere and that is potentially a free energy input.
The free market is the perfect motivation for innovations not explored in your analysis. Still you leave me in awe as to how alts are going to evolve as the energy equilibrium evolves.
My responses to Bob are the following:
1) In practice, users of the Bitcoin network do not like including transaction fees and there are endless threads on Bitcoin Talk of people complaining about fees. In fact, one of the purported — wildly incorrect — selling points with Bitcoin is that it is somehow “free.” Obviously this is incorrect, utilizing scarce resources is not free. Someone has to pay. The people who pay in this case are all bitcoin holders as roughly every 8-10 minutes new bitcoins are minted, diluting the shares of everyone through inflation.
But let us assume that we fast forward 100 years into the future when there are no longer block rewards, that miners will continue providing labor solely for transaction fees. If these fees are floated and chosen by miners, it is impossible to say a priori what the actual market clearing fee will be. Will it be 0.1%, 1%, 10%? Or something in between?
One issue in Bob’s scenario is the “30% free transactions” — this is completely arbitrary. Miners still have to bear a real cost to transacting and securing the network and only do so today because of block rewards. If there are no block rewards and they want to continue providing “free” transactions, then they will be doing so out of charity which is not a sustainable business model.
2) The thermodynamic limit is something Andrew Poelstra has written about. For the purposes of Bob’s specific point, Poelstra’s document can be ignored for the moment because it is still necessary to actually describe what is happening today. If a bitcoin is worth $1,000, then an economically rational miner will only spend (capital costs + operating costs + taxes, etc.) no more than $1,000 to extract rents on that token.
In practice this is not the case as there are numerous examples of people and companies operating at losses for a variety of reasons, primarily because of price expectations: they believe that the token will eventually appreciate in value and the market value of the token will eventually cover their operating losses. Thus today, the network is being “oversecured.”
As far as “waste,” that is how Proof-of-work works. Someone, somewhere has to “burn” (or dissipate) something in order to secure the network. Irrespective of what part of the supply chain or logistical operations it takes place, market participants are provided signals by token value (e.g. a $1,000 bitcoin) to turn off or on their hashing systems. It doesn’t matter what the energy source is or how efficient ASICs are, market participants will simply use a calculator to find out if their inputs (capital costs + operating costs, taxes, etc.) allow them to profitably provide their labor. The same goes for a $1 million bitcoin.
3) Let’s assume that tomorrow several chip manufacturers announced that they were now shipping chips with fabrication node spacing that reaches the Planck limit (see this interesting paper). That essentially, irrespective of who you bought from, their hardware design was the most maximum efficient chip possible. We will call this the Alice design. What would happen in this case is that whoever was able to get a hold of Alice first would profit from it disproportionally at first (as other competing farms were using older less efficient designs). But over the months, the distribution of Alice became widespread and you could go to a store and buy Alice off the shelf from a neighborhood retailer.
What would happen then is, since everyone is competing with the same hardware, the only variables to profitability would be land costs (plus taxes and compliance costs in your jurisdiction) and operating costs (electricity). As a consequence, there would be global arbitrage, a dance in which miners would gravitate towards the cheapest region of the globe with favorable tax policies and cheapest electricity prices.
We already observe this happening today, which are discussed in that article.
The benefits that heating may play could be a factor, but if history of cloud computing is any guide, it is relatively unimportant — Google does not put employee housing within the middle of its data center to warm them with a Carnot engine of some kind (yet). But again, this is unimportant. All mining facilities, just like any data center, will have a profitability calculator. Irrespective of how they displace or use the energy they have a bottom line of whether or not they can continue providing the service (via their computational labor) at a profitable rate.
Also, in practice, data centers typically receive subsidies from a variety of sources (like local tax breaks). Even if you removed all of the subsidies and all geographical regions were “pure free markets” — there are still areas on the planet with better infrastructure, cheaper energy resources, better property rights, etc. Those are the same locations capital moves to on a yearly basis.