Blogs & Articles: Noahbjectivity on Bitcoin Mining 🔗 3 years ago

A response to Noah Smith

Excess natural gas being employed to mine Bitcoin at an oil well in Alberta, CA. Used with permission by Upstream Data

The Bloomberg columnist Noah Smith has a lot of thoughts on Bitcoin. Some of them are really solid and engage with the reality of the protocol itself, which is rare for a member of the mainstream media circuit. He also discloses that he owns Bitcoin, which is impressive for an economist and a member of the establishment. So I’m pretty happy with him overall. I don’t want this piece to be interpreted as a blanket critique of Noah’s stance on Bitcoin. However, Noah’s recent column in Bloomberg, Bitcoin Miners are on a Path to Self-Destruction, makes a few claims that warrant a response.

Noah’s basic premise is that Bitcoin miners are effectively hogging the grid in the various places where they operate and risk getting banned entirely. Not only is the notion of a global coordinated ban on mining far fetched, but Noah relies on a few claims that are dubious at best. Let’s investigate.

Claim: Bitcoin is unique in that a rising price entails more energy draw

Noah starts by noting that because miners are rewarded partly through new issuance, increasing unit prices means more real-world resources that are put into mining. That part is true. But the rest of the quote isn’t quite fair:

But Bitcoin’s high price may now be leading to new problems for the cryptocurrency, because unlike other financial assets, Bitcoin uses more resources as its price goes up. […] So the more Bitcoin’s price goes up, the more resources it consumes.

Bitcoin’s most obvious real-world analogue is of course gold. Gold has this exact same property, so it doesn’t makes sense to single Bitcoin out here. When the unit price of gold goes up, gold mining increases (sometimes with a lag), and hence so does its energy consumption. This is because mines are heterogenous in their profitability thresholds, and some are only profitable at higher prices. You can see this phenomenon clearly on this chart, courtesy of Arkadiusz Sieron at Sunshine Profits:

Sieron, 2015

As the price of gold increases, it drives up production (with a lag). This is similar to Bitcoin price and hashrate dynamics. You can see the lagged relationship between hashrate and BTC price here:

This is an incidental point so we won’t dwell on it — but it was worth mentioning that Bitcoin isn’t the only commodity to feature this additional energy draw as unit prices rise.

Claim: Bitcoin ‘hogs local power resources’

But [Bitcoin] undoubtedly hogs local power resources, which makes other customers mad. […]

Bitcoin miners are trying to fix this [tendency of the network to hog local power] by making use of the excess solar and wind power produced during peak hours, but it remains to be seen how much of this extra energy is just lying around.

Well, let’s see, shall we? Best not to keep Noah in the dark.

As it turns out, there’s a tremendous amount of stranded energy floating around. (Skip to the end of this section if you just want the numbers.) This includes both on-grid energy that power grids cannot accommodate for various reasons like mismatches between production and demand, and off-grid potential energy that simply has no chance of ever making it to the grid. I call this category of energy nonrivalenergy, because its use doesn’t deprive anyone anywhere of energy nor does it drive up their costs — in fact, monetizing surplus energy could actually drive down grid costs (because it sponsors the buildout of otherwise-unprofitable energy infrastructure). Now, precisely how much nonrival energy is being employed to mine Bitcoin today is an interesting, and challenging question, to which I don’t have the answer.

Ultimately, the earth is being bombarded by sunlight which is itself a constant source of stranded or unexploited energy, as well as the wind eddies that ultimately derive from that same hot ball of gas. But wind and solar in their current form aren’t particularly amenable to mining Bitcoin, because they generally have low capacity factors. If you buy ASICs and put them to work mining with wind and solar, you will only be running them roughly half the time, because the sun doesn’t always shine and the wind doesn’t always blow.

However, to make money mining Bitcoin, you only have a finite amount of time in which to do so with a given ASIC, because chips generally get better, and so you have to depreciate ASICs over time. ASICs eventually get priced out because they simply aren’t efficient enough to keep up with the rising hashrate. So they have a finite ‘useful lifetime’, during which you want to run them as much as possible. Mining Bitcoin exclusively on solar or wind isn’t going to be competitive.

What Bitcoin can do, however, is absorb structural surpluses of energy. The best data we have comes from the Cambridge Center for Alternative Finance which ingested data from Bitcoin mining pools to geolocate mining and create an energy consumption index (CBECI). According to the CBECI, roughly 71% of mining took place in China from 4Q19 to 2Q20 (they don’t offer more recent data). Within China, there’s four standout provinces where the vast majority of mining occurs: Xinjiang, Sichuan, Inner Mongolia, and Yunnan. Collectively, they produced 63% of the global Bitcoin hashrate during the period in question, according to the CBECI.

Those provinces can be grouped into two categories: coal rich (with significant contribution from renewables like wind and solar), or hydro rich. Xinjiang and Inner Mongolia (IM) are heavily coal powered, but their grids are also 35 and 30 percent wind/solar, respectively, according to Bloomberg (further sources on Xinjiang and IM). By contrast, the U.S. grid was only 20% renewable (inclusive of hydro) in 2020 when measured by generation, according to the Energy Information Administration. So if Bitcoin was mined exclusively in Xinjiang and IM, it would still be more renewable in origin than if it was mined with the average energy mix of the U.S. Interesting.

What all of these regions have in common is a relatively low population density (collectively, they only host 12.7% of the population of China), but they have abundant energy resources. It’s not a coincidence that Chinese Bitcoin mining mostly happens in these four provinces. They have a massive overabundance of energy and a general inability to direct it to population centers. Look at the below map courtesy of Bloomberg and see if you can spot the pattern:

Zhou and Lu, 2017

If it wasn’t clear enough, here’s another chart showing how renewable or hydro-rich the various provinces in China are, compared to the actual population centers. I’ve put stars on the four provinces in question. They’re nowhere near the actual load centers.

Zhou and Lu, 2017

It’s very clear that Bitcoin miners didn’t just choose those locations at random. Xinjiang and Inner Mongolia have tons of available capacity — including from wind and solar — and little grid demand to mop it up. This chart from Liutong Zhang at the Lantau group makes the energy abundance of these two provinces very clear:

Zhang, 2017

For a visual guide to the excess of wind and solar resources in China, see below:

Zhou and Lu, 2017

Sichuan and Yunnan have oodles of hydro, and are similarly distant from population centers. This map from Shen et al (2019) grades the provinces by hydro capacity:

During the wet season, there is simply too much water flowing through these regions into dams for the grid to consume, and it has to be let out. You can explicitly see the seasonal curtailment on this chart of Yunnan’s hydro resources on the chart below.

Liu and Davidson, 2021

Note that the curtailment rates declined in 2017 and 2018 because Bitcoin mining began to ramp up in Yunnan with the rally. So you can start to see its effect on the energy excess.

While it’s hard to get exact numbers, it’s fair to assume that Bitcoin is being mined in large part with nonrival energy in these key provinces. It’s not depriving anyone of energy, because energy is wildly overabundant. And when the authorities get sick of Bitcoin, as with Inner Mongolia, it’s a net good for the Bitcoin network, since that means less Bitcoin is being mined with coal. IM is slightly worse than Xinjiang as far as the energy mix is concerned. (The ban also has the side benefit of reducing China’s leverage as it pertains to Bitcoin. If they were planning a sneak attack on Bitcoin, banning mining is the opposite of what you’d expect from the CCP.)

But the Bitcoin mining story is increasingly being written outside of China. As Mustafa Yilham of the industrial miner Bixin noted on a recent podcast appearance, most recent buyers of mining hardware are ex-China. There’s a huge market for mining machines in the U.S. as Galaxy Digital and Digital Currency Group have established large mining units alongside a vibrant set of publicly traded miners which have raised significant capital in recent months. Mustafa’s comments are worth reading carefully:

Based on our conversation with different mining manufacturers, on average around 60%, of the mining machines sold (sic) in the past two quarters or so have been outside of China, and mostly actually in North America. And I think that there’s few factors that plays into why North America could be huge in mining in future. One of them is price. I think you’ll be surprised to hear that. In fact, the price right now in North America notice are cheaper than the average price in China. And in the past, we used to think, US high labor costs, less mining ecosystem, but we can clearly feel that things are changing rapidly in the past 1–2 years. And I think during the next cycle, you will see a much higher involvement from the US and other countries into mining. And I think the most important factor also for US miners, is that they have access to cheap capital cost, meaning, you know, they have the ability to go out there and borrow at a much cheaper rate than Chinese miners are able to.

So what will these ex-China miners mine with? Some will use grid power, but others will pursue renewable strategies. Many of the announcements you will see regarding Bitcoin mining will follow Square’s lead with their Bitcoin Clean Energy Investment Initiative, or Aker’s Seetee initiative which has vowed to mine with renewables.

Increasingly prevalent is the notion of ‘Pipe-to-Crypto.’ This process entails the off-grid mining of Bitcoin with methane, a natural byproduct of oil extraction. Depending on the regulatory framework, natural gas is either vented or flared at rig locations. Because these oil wells are frequently remote, off-grid, with no pipeline infrastructure, and unviable economics for capturing (due to low prices of natural gas), oil rig operators often combust waste gas on site.

However, flaring tends to be very inefficient and on windy days large fractions of the methane just bubble out. Methane is a worse greenhouse gas than C02, the output of combusted methane, so flaring is a net positive. When this gas is put into a generator and used to mine Bitcoin, operators can ensure a full burn (eliminating vented methane from inefficient flaring) and can additionally mitigate the emissions produced. Given the baseline situation of venting/flaring, a clean, supervised burn in a generator is a net positive from a carbon perspective. A number of firms are now pursuing this opportunity, some of them partnering with publicly traded energy companies. It goes without saying that completely off-grid natural gas is entirely nonrival with household or commercial energy consumption. It was never going to be monetized, captured, consumed, or delivered to households. Its fate was simply to be combusted or vented.

The scale of flared gas is massive. In the U.S. in 2019, according to the Energy Information Administration, 538 billion cubic feet of natural gas were vented and flared. That’s 1.2% of the total gas withdrawn from the earth in the U.S. in 2019. And according to energy analyst Brannin McBee whom I consulted for this article, that number is likely a vast underestimate. In McBee’s words:

This is the official number reported from the Energy Information Administration. This value is collected by self-reporting and state mandates across the U.S. However, it is well known in the industry that the true gas venting/flaring value is substantially higher. Pipeline leakage, initial production grace periods, legacy wells with poor infrastructure all contribute to estimates that reach nearly 10x higher than what is reported to the EIA.

So with these case studies in mind, let’s consider some numbers on curtailment. First let’s contextualize. Digiconomist pegs Bitcoin’s current annualized energy consumption at 89 TWh/year, while Cambridge estimates it at 138 TWh, so let’s assume the answer is somewhere in between. The numbers on curtailment are absolutely massive, and it’s easy to come up with figures that exceed the consumption of the Bitcoin network. This is because the world produces far more energy than it consumes, and it has the capacity to produce far more, through unexploited sources that go to waste (like flared methane).

Here are some figures to give you a sense of the scale of curtailment that goes on. I can’t get comprehensive data on global energy curtailment or stranded energy resources, but the numbers below should provide you ample assurances that Bitcoin can run exclusively on nonrival energy.

• In 2016, China curtailed 40.7 TWh worth of wind and 11.5 TWh of solar power alone (Zhou and Lu 2017 & Luo et al, 2018)
• In 2016, Yunnan alone curtailed 31.4 TWh worth of hydro power (Liu et al, 2018)
• In 2016 and 2017, China curtailed 100 TWh on average worth of hydro, solar, and wind energy, collectively (Dong and Qi 2018)
• The (very conservative estimate of) 558B CF flared/vented natural gas in the U.S., if put to use in 7 Heat Rate (7m BTU/MWh) combined-cycle plants, would have generated 76.9 TWh in 2019 (EIA, McBee calculations)

And we’re just considering two sources of nonrival energy: clean energy curtailment in China, and vented/flared gas in the U.S. There are many other sources of stranded and nonrival energy sources globally. Suffice to say, there’s enough nonrival energy out there to run Bitcoin many times over. It’s just a matter of deploying hashrate in the right locations, which miners are doing — aggressively.

Claim: Bitcoin is hogging space at chip foundries, interfering with global manufacturing

Noah asserts the following:

Meanwhile, Bitcoin’s demand for computer chips has hogged the production lines at Taiwan Semiconductor Manufacturing Co. and Samsung Electronics Co., contributing to a global chip shortage that is costing automakers tens of billions of dollars and threatening the phone industry as well.

This is actually the easiest claim to dispute, because it relies on sources which aren’t talking about Bitcoin at all. If you trace the sources Noah relies on, he links a FastCompany article which says the following:

But the straw that finally broke the proverbial camel’s back was the sharp rise in bitcoin prices in early 2021. This increased the demand for the graphics processing units that are traditionally used in mining the digital currency, exacerbating the semiconductor supply issues further.

Immediate red flag with ‘graphics processing units’ — everyone knows that Bitcoin isn’t mined with GPUs. If you trace that article to its source, you land at an SCMP article which discusses a GPU shortage worsened by high cryptocurrency prices. As anyone with passing familiarity of Bitcoin should know, you haven’t been able to GPU mine Bitcoin since 2013. Bitcoin mining relies on specialized hardware using ASICs. It’s mainly Ethereum’s price resurgence which is driving the GPU shortage. Because Ethereum miners mostly use high end NVIDIAs, an increase in Ethereum usage (and hence fees, and the price of ETH) absolutely increases demand for GPUs and prices out gamers. Gamers, and I suppose some opinion columnists too, sometimes erroneously deride Bitcoin for their expensive gaming rigs. But it’s Ethereum that should be the target of their ire.

Interestingly, NVIDIA is aware of this problem, and has built a crypto-specific GPU, while building in anti-mining mechanisms in their mainstream GPUs. The general purpose GPUs will throttle usage if they detect crypto mining activity. This is a very smart way to segregate their product lines across different customer segments and solves the problem of miners pricing out gamers and other GPU consumers.

I don’t exactly blame Noah here, because it’s Fastcompany which is making a mistake by attributing the Bitcoin price rise to a GPU shortage.

So let’s give Noah the benefit of the doubt, and tackle the more general claim, that demand for Bitcoin chips (and not GPUs) is interfering with crucial points in the smartphone and vehicle supply chains.

This claim is also wide of the mark — but in a much more interesting and revealing way. I don’t personally know much about the semiconductor market, but I know a number of analysts that cover semis at large hedge funds. So I dialed one of my friends who is an expert on semis, and he agreed to speak with me on background. I’ll refer to him as ‘Big Al’, his chosen pseudonym.

As it turns out, the way foundries like TSMC (the foundry Bitmain relies on) operate, is that they tier their customers. They don’t optimize just for revenue, but consistent demand. They treat their Tier I customers much better, and give them privileged allocations to foundry space. Bitmain, the largest ASIC manufacturer, is not a Tier I customer. They have to settle for scraps. According to Big Al, the chip shortages are mainly due to a skyrocketing demand for consumer electronics and cloud during Covid (as everyone was stuck at home with nothing to do). Additionally, Qualcomm’s issues (the major smartphone chip manufacturer) were compounded by a shutdown at Samsung’s Austin plant during the recent power cuts there where Qualcomm sources RF chips.

The way it works is that reliable Tier I customers are prioritized for allocations, while more cyclical and less predictable buyers like Bitmain have to wait. In Q3 and Q4 2020, TSMC was absolutely jammed with orders from their Tier I clients: Apple, Qualcomm, NVIDIA and Broadcom, to name a few. This was an active time for smartphone manufacturers due to Christmas and Lunar New Year shopping, the launch and rollout of 5G in phones, and Oppo, Vivo, and Xiaomi making a market share play (due to the political pressures hurting Huawei). Crypto miners would not have had much allocation.

Crypto miners, after being stonewalled in 2020H2, got their allocations in Q1 and Q2 2021 to manufacture their new 5nm chips. Given that Bitcoin ASICs don’t have to perform to the tolerances that smartphones do (ASICs are more disposable and are expected to depreciate more quickly), foundries are happy to give lower quality chips to miners, and the miners happily pay for them. In fact, to get foundry space, miners pay a premium, which has the effect of subsidizing the buildout of new infrastructure at the 5nm level. So the Bitcoin hardware manufacturers are very much second class citizens in this incredibly competitive game of bidding for foundry allocation — and that’s likely to remain the case for a long while, because the industry is so cyclical. The foundries will likely continue to mistrust Bitcoin hardware manufacturers. When Q3/4 2021 roll around, Bitcoin ASIC manufacturers will likely find themselves out of luck once again, with foundry capacity extremely scarce. According to Big Al, there’s no price at which a Bitcoin miner can outbid a Tier I client for allocation — the Tier I clients are absolutely privileged.

Ultimately, Bitcoin miners represent a small fraction of TSMC revenue — around 1% according to Bernstein. The notion of a marginal, Tier II industry being responsible for chip shortages is fanciful. The more immediate cause is the supply inelasticity of foundry space (due to gargantuan fixed costs) and the massive surge of demand for electronics due to a global lockdown and new technologies coming online.

Claim: Proof of Stake is a viable alternative

To sum up his piece, Noah says:

To avert [bans on Bitcoin mining], the developers who control Bitcoin’s algorithm need to think about switching to a cheaper technology. One alternative is a proof-of-stake system, where mining can only be done by people who already own a lot of the cryptocurrency; this cuts down massively on resource use by limiting competition.

This is a cornerstone of the anti-Bitcoin energy argument: the notion that you can have something for nothing with Proof of Stake. No energy consumption, yet still a functioning decentralized consensus. If this logic reminds you of perpetual motion machines, it’s because that’s exactly what is being proposed here: a completely free lunch where you get precisely the same assurances as Bitcoin with no costs whatsoever.

Of course, this is fantastical. ‘Proof of Stake’ is just a fancy phrase meaning “those who have the most wealth wield political control.” That sounds a lot like our current system, which Bitcoin is specifically designed to solve. Bitcoin explicitly rejects politics, and doesn’t grant any special privileges based on coins held. If holding more coins gave you more control, the attempted takeover of Bitcoin through the 2X movement (backed by the largest custodians and exchanges in the industry) would have succeeded.

Additionally, as Paul Sztorc pointed out way back in 2015, Proof of Stake is often simply a veiled (and obfuscated) form of Proof of Work. If the protocol imposed a limit of $100 worth of a coin per staking node, for instance (in a bid to foment ‘decentralization’), you’d see industrial staking farms emerge with tens of thousands of nodes. It would just be a roundabout PoW.

Capital has a cost, and a PoS system at equilibrium would consume capital, in just the same way Bitcoin does. The way this would actually be instrumentalized would be through a carry trade of sorts — borrowing in dollars to take advantage of the high ‘interest rates’ offered in, say, ETH. Capital that would be allocated to PoS systems could be used to build nuclear power plants, windmills, or solar farms. Capital is just our abstraction for energy. If a PoS system consumed $1T of society’s capital resources, that entails a lot of potential carbon sequestering farms gone unbuilt.

So if PoS is just veiled PoW, it provides no marginal benefit. Not only that, it is very unclear if PoS actually grants equivalent assurances to PoW. In my view, it is unambiguously worse from a decentralization perspective. The cost of capital is profoundly inegalitarian, and gives would-be oligarchs the upper hand in consensus. Problematically, large custodians (where coins inevitably end up settling) can easily be employed to take control of Proof of Stake systems.

Coin Metrics benchmarks the number of exchange-held ETH at 14.8m units; Viewbase has it at 22.8m. That latter figure is worth $37b, equivalent to 29.2% of the trailing 12-month active supply on ETH. Those custodians — the largest of which alone holds north of 8m ETH — could interfere with consensus in a PoS system, as happened with Steemit, where exchange votes were used to change consensus rules and ultimately confiscate coins from certain users. Additionally, in PoS, we’ve seen vote buying and validator cartels.

And if you expect a Proof of Stake token to go truly mainstream, the system would end up privileging the entities that have access to the cheapest capital: large financial institutions that have access to effectively unlimited liquidity from central banks. If you think Proof of Stake empowers the individual, compare the cost of capital for regular folks (an obvious proxy would be credit card APRs) to the cost of capital for hedge funds like Citadel.

Gigantic, too-big-to-fail institutions get favorable access to liquidity, even when their trades go awry, as happened in 2020. Last year, the Fed also bought corporate debt — vastly empowering large firms. It’s no secret that the larger you are, and the more proximate to the central bank, the cheaper your capital is. So Proof of Stake crucially lacks the ‘hardness’ that an energy cost provides, and has the negative feature of empowering large firms at the expense of smaller ones.

In contrast to fiat currency and credit, energy is much more globally distributed, and cannot be conjured from thin air by a central bank. So far there have been no coalitions or cartels built around mining Bitcoin that have actually imposed systematic censorship. And hashrate only continues to get more distributed. Expect far more US-based and non-China hashrate in the next 12–24 months — that’s where all the newly-minted ASICs are going. China kicking miners out of Inner Mongolia is extremely positive by that rubric.

Claim: Bitcoin mining bans are bad for miners, investors, and developers

Noah goes on to say that “[bans on Bitcoin mining] will be bad for Bitcoin miners, as well as crypto investors and software developers.”

This is questionable. Yes, a globally coordinated ban on Bitcoin mining would force it underground. I don’t expect a global ban. Already, certain jurisdictions encourage Bitcoin mining, because they have stranded energy resources that can be monetized through mining. In other words, they can export energy, the way Iceland has historically done with aluminum smelting. The world is a big place, and policymakers have diverse reactions to Bitcoin. We have seen Kentucky encourage Bitcoin mining through new tax breaks. It’s an open secret that the government of Georgia (the country in the south Caucasus) subsidizes Bitcoin mining. The Pakistani province of Khyber Pakhtunkhwa is piloting a state-sponsored mining initiative. Some jurisdictions will object to Bitcoin mining in their borders. Others will embrace it.

But a ban isn’t bad for your average miner. First of all, hashrate is extremely mobile. Miners will just migrate their ASICs, as they do seasonally between Sichuan/Yunnan and Inner Mongolia/Xinjiang already. Bitcoin eats energy subsidies for breakfast, and ultimately punishes governments (like Iran and Venezuela) that uneconomically subsidize energy to maintain regime legitimacy. If you dislike the environmental impact of nonrenewable subsidies and dislike the authoritarian states that employ them, Bitcoin’s subsidy-toxicity is a very good thing.

Second of all, since miners are in constant competition with all other miners worldwide, crackdowns on specific miners mining at competitive rates raise the per-KWh price threshold at which mining is profitable. If a miner is mining with subsidized energy deriving from coal (giving the miner a below-market rate of, say, 2c/KWh), and the local authorities take exception to this practice, they can end the subsidy or kick out the miner. That raises the profitability for all other miners worldwide.

Investors and developers don’t really care about where Bitcoin is mined, as long as Bitcoin hashrate is sufficiently decentralized and its assurances hold up. So far, it has. The important thing is to continue to develop technologies like Stratum v2 and others that continue to empower individual miners at the expense of pools and enhance the genuine decentralization of block formation.

At the end of the day, Bitcoin doesn’t care where it is mined. And Bitcoin doesn’t have a defined threshold of hashrate at which it is considered secure. It has been secure at many levels, historically. In my opinion, it substantially overpays for security. If states like China take umbrage at Bitcoin’s consumption of grid resources, hashrate will flow elsewhere. In the long term, I believe Bitcoin will be mined almost exclusively with nonrival energy, because all on-grid mining is always exposed to the exact political caprice that Inner Mongolia miners are dealing with. As we know, there’s ample stranded assets just waiting to be monetized. That couldn’t be more positive for Bitcoin.

Thanks to Brannin McBee, Takens Theorem, Alex Gladstein, Leo Zhang, and Big Al for their generous assistance with this article.

Update (04/03/21): Noah wrote a rebuttal to this piece. Find it on his Substack.