Tuesday, October 18, 2022

The Power Of Ethereum's Merge

The laudable goal of Ethereum's "Merge", the long-awaited transition from Proof-of-Work to Proof-of-Stake, was to eliminate more than 99% of the massive environmental damage caused by Ethereum's consuming about half as much power as Bitcoin. There are many reasons to criticize Proof-of-Stake, but the Merge definitely achieved this goal. We can no longer point the finger at Ethereum and its users and claim they are wrecking the climate half as much as Bitcoin.

However, as usual when cryptocurrency advocates tout claims like "more than 99%" it is necessary to apply skepticism. From the planet's point of view the issue is not whether the Merge reduced Ethereum's carbon emissions, but whether the Merge reduced the carbon emissions of cryptocurrencies as a whole. The answer is "not so much". Below the fold I discuss the details and estimate the real reduction to be ~35%, because most of the power has been diverted to mining Bitcoin.

Before the Merge the Ethereum (ETH) blockchain was maintained by a vast network of computers, most using high-end GPUs with a few using the E9 ETH mining ASIC that Bitmain launched in July 2022. After the Merge, this massive investment in hardware could no longer generate returns by mining ETH. Its owners had seven options:
  1. Use it to mine Ethereum Classic (ETC).
  2. Use it to mine on the pre-Merge blockchain, now renamed EthereumPOW (ETHW).
  3. Use it to mine other ASIC-resistant cryptocurrencies, such as Monero.
  4. Turn it off, trash it, then:
    1. Stop mining, start staking.
    2. Buy new BTC mining ASICs and mine BTC.
  5. Turn it off and sell it on the used hardware market. Then:
    1. Stop mining, start staking.
    2. Use the power and rack space freed up to mine BTC using rigs previously out of service because they were uneconomic.
    3. Use the power and rack space freed up to mine BTC using state-of-the-art rigs previously on order.
Note that option 4 increases cryptocurrencies' serious e-waste problem, and only option 4a actually reduces their carbon emissions. Option 5a provides brief dip in emissions until the buyer turns the hardware back on. Options 5b and 5c actually increase the emissions. Lets try to estimate which of the options were taken.

Source
Immediately before the Merge, the ETH hash rate was around 1PH/s and the ETC hash rate was around 60TH/s. Immediately after the Merge, the ETC hash rate spiked to about 225TH/s, so about 16% of ETH mining switched to ETC. But they found it unrewarding; the ETC hash rate has declined to around 135TH/s, so only about 7.5% of ETH mining is still mining ETC.

Source
Immediately after the Merge, the ETHW hash rate was around 45TH/s, and is now around 40TH/s. So only about 4% of ETH mining is still mining ETHW; the fork fizzled.

There are a large number of alternative ASIC-resistant cryptocurrencies for obsolete ETH rigs to mine. For example, Monero's hash rate was about 2.8GH/s before the Merge, spiked to around 3.3GH/s, and is now around 2.4GH/s. So Monero was not a significant alternative.

Source
There are two interesting things about the Monero hash rate:
  • It is volatile, probably because a significant part is mining by malware.
  • It is remarkably small, about 3*10-6 of the ETH hash rate. Thus even a minute fraction of the GPU hash power made redundant by the Merge could mount a much more than 51% attack on Monero.
Source
Redditor RxBrad recently posted the latest of his detailed tracking of post-Merge mining in Where have ETH miners gone - 4 weeks after the Merge... and concluded:
23.29% of former Ethereum miners are still mining, one month after the Merge. This rate is down 1.59% from 24.88% last week. About 232k RTX 3070 equivalents were shut down in the last seven days, at a rate of about 33k per day.

15.68% (down 1.7%) of ETH miners are mining an alternate Ethash coin, and 7.61% (up 0.11%) are mining non-Ethash coins.
RxBrad's explanation is:
Network hashrates still have not fallen enough to curtail the continued drop in mining profits. At 10 cents per kWh, there are no coins that can be mined profitably (as has been the case since very shortly after the Merge). At 5 cents per kWh, only 13.82% of current mining activity is on coins that can be profitably mined (only counting energy costs for the GPU itself).
Of course, as I discussed in Generally Accepted Accounting Principles, simply accounting for power consumed by the GPU gives a wildly optimistic estimate of "profitability"; the major omitted cost is depreciation.

Source
David Pan's Bitcoin Mining Surges to a Record as Merge Opens Up Capacity provides an interesting account of what happened post-Merge:
The amount of computing power dedicated to Bitcoin mining surged to a record as more companies made use of the energy and data center space freed up after the upgrade of the Ethereum network, likely further compressing profit margins.

Mining difficulty, a measure of Bitcoin miners’ computing power for the blockchain, has jumped by 13.6% in the two-weeks ended Monday. That was also the largest bi-weekly adjustment since last May. The increase is in part thanks to the decline of Ether mining, analysts said.
...
“Rack space for Bitcoin miners was limited, freeing up the space paves the way for machines previously unplugged to get plugged in,” said Ethan Vera, chief operations officer at crypto-mining firm Luxor Technologies, which has provided services to both Ether and Bitcoin mining companies. Vera estimated around 4% of the current computing power for Bitcoin mining has been flowing from Ether mining to Bitcoin miners in the past two weeks.
It isn't just rack space, it is the power to go with it. Note that the spike in BTC hash rate was about 20%, starting around 12 days after the Merge. Presumably this spike represents miners re-using rigs they previously turned off as uneconomic rather than a flood of newly purchased rigs. If Pan is right that:
Ether mining consumed about half of the energy used for Bitcoin mining before the Merge.
This shows the massive difference in energy consumed per hash, because half the BTC hash rate before the Merge was about 110EH/s, or 110,000 times the ETH hash rate. Of course, the hash algorithms aren't the same, so the hash rates are not directly comparable.

The ~75% of ETH mining that has stopped represents ~37.5% of pre-Merge BTC mining power usage. BTC mining increased by ~20% post-Merge, so the actual reduction was ~17.5% of pre-Merge BTC power, or ~35% of pre-Merge ETH power. Nice to have, but a whole lot less that the "more than 99%" claim.

Additional competition for the fixed supply of mining rewards will harm miners' margins, as Pan reports:
A higher level of computing power will lower mining revenue for Bitcoin miners who are already battered by low Bitcoin prices and soaring energy costs. The more mining power there is, the less Bitcoin each miner will receive as the network only gives a limited amount of the token reward after successfully processing a certain amount of data.
...
“What’s critically important for Bitcoin mining is access to cheap electricity,” said Matthew Kimmell, digital asset analyst at crypto asset manager CoinShares. “If those Ethereum operations were built with cheap power sources in mine, then I think well-capitalized Bitcoin miners could have seen that as an expansion opportunity to buy those assets and deploy machines.”

More resources from Ether miners are only one of the reasons for Bitcoin mining power’s surge.

There have been fewer power curtailment from major miners as they head into the cooler months in the US and Europe, said Joe Burnett, head analyst at Blockware Solutions. While electricity cost has fallen month over month, it could rise again as people turn on their heaters in the winter, Vera said.
It wasn't as if BTC miners were living on Easy Street before the Merge. David Pan also reports that Bitcoin Miner Argo’s Shares Plummet as Liquidity Concern Rises:
Bitcoin miner Argo Blockchain PLC’s shares are plummeting after the firm announced several measures, including issuing stock at a discount to an unnamed investor for $27 million, to ease liquidity pressures.
...
Argo also plans to amend loan agreements with its primary lender, New York Digital Investment Group, to unlock $6 million. Argo has $84 million outstanding loans with NYDIG, most of which are backed by mining machines. It will also sell 3, 400 mining machines for nearly $7 million, Argo’s CEO Peter Wall said in a video.
...
Bitcoin mining companies have been battered by low Bitcoin prices, soaring energy costs and more competition in the sector. Shares of public miners have been deep in the red this year. Argo has one of the largest Bitcoin mining farms in Texas with one facility planning to have 800-megawatt capacity. “The high power prices in Texas this summer really hurt us.” Wall said.

The company’s Bitcoin production also fell to 215 in September from 235 coins in the previous month, according to an operational update on Tuesday. The firm attributed the decrease to rising mining difficulty due to more computing power for the Bitcoin network.

Other miners have also raised money through new shares. Core Scientific signed a $100 million common stock purchase agreement with B. Riley Principal Capital II, while Iris Energy agreed to sell up to $100 million in equity to the same investment bank.
It looks like the financial problems Paul Butler described in The problem with bitcoin miners are becoming hard to ignore.

2 comments:

  1. This is a rather smarmy and unfair interpretation of the claims. The locus of control for miners to quit mining completely is not within Ethereum dev hands. You seem to put words in their mouth and pretend they made claims they did not. The 99% figure is for energy consumed to secure Ethereum, that did go down by 99%, you redefined it to make hay of the fact that it didn't reduce the entire crypto industry's energy usage, which was never claimed.

    ReplyDelete
  2. Right. That's what the first paragraph of the post says. The Ethereum developers achieved their goal; we can't any longer point fingers at them. But we can't ignore the bigger picture.

    ReplyDelete