Bitmain has not publicly released any significant efficiency upgrades since July 2016. Innovation at the mining sector's biggest player seems to have stalled. If the exponential trend had continued we should have reached >40 TH/s/kW by Jan 2018, instead we are at 2016 levels. This is either because Bitmain cannot make more efficient ASICS or because they are keeping secret any innovations they have made.
Looking at the return on investment (ROI) chart. The white horizontal line is the breakeven line. Anything above is profit. The vertical lines are release dates of the various S-model miners. We see that a new miner is released to the public just when the previous model becomes obsolete. Normally there is a lead time before delivery of several months. Bitmain have not made any announcements regarding an S11.
Despite all of this the hash rate is increasing unabated. Have Bitmain been secretly mining with new hardware or have they simply been bringing more S9s online? If they have an S11 why aren't they releasing it?
HORIZONTAL LINE: Breakeven should be at 1 on the right axis
Comment
⋅
Bitmain will hold an IPO to raise USD 12 billion. This will give them wildly more funding than any other ASIC producer. Bitmain's two mining pools already control over 42 % of the global hashrate. By the way, a 51 % share is not necessary to carry out a majority attack, it just ensures its success. In the interview I posted above Jihan Wu says they plan to diversify into "Artificial Intelligence chip products". He also says that one of his major goals for the future is to work together with regulators, it's why they are so heavily invested in Circle (Goldman Sachs). This is what corporate crypto looks like: a company valued at 12 bn controls the network (>51%) and is in bed with investment banks who control the exchanges (guess who regulators will favour). BTC to the moon? Very probable, but currency of the people? Look elsewhere...
There is an interesting study (cell.com/joule/fulltext/S2542-4351(18)30177-6) that tries to estimate power consumption of BTC mining. They also estimate lifetime of miners, which they put at 1 to 2 yrs so it is important to take account of acquisition costs in my opinion. But their figure is slightly lower than mine, because they estimate acquisition costs for the manufacturer (USD 500) where as I just took the current retail retail price (800 unit + 100 PSU)
Anyway if I use their data and recalculate the conversion factor for a 1 yr amortization period I get 0.25 USD/TH/s instead of my retail 0.34 USD/TH/s.
I also chart Bitmain's ROI (area plot) for write off periods of 1, 1.5 and 2 yrs. We see that accordingly Bitmain should still be in the profit zone. Bear in mind this does not include coolin cost (which are substantial) and of ancillaries such as personnel etc.... however for "retail" miners (line) the situation is different with amortization plans of 1 year now in the red at current BTC prices
Very nice educational chart! What are your assumptions on electricity prices? I don't really understand your calcualtions (e.g., 'BCHAIN/MIREV'/'BCHAIN/HRATE'/0.34 ).
where constant is a conversion factor to convert hash rate to dollars (e.g. USD cost per hash power expended). In the case of the Antminer S9 it happens to be 0.34 USD/TH/s
They are estimates and I don't account for facilities and employees etc.
lorschd
⋅
@barclayjames, Thanks, now I understand. I really like that. However, I think ROI is somewhat misleading in this context as you show the average costs of mining rather than a “normal” ROI. A miner who already bought the miner would also earn profit (contribution margin) if only his electricity costs (plus a bit more) would be covered (“alpha_elec” aka marginal costs) . Hence we may well see situations where the ROI -as defined here- is below 1, but mining has a positive contribution margin (covering electricity expenses and a bit more). However, before you buy your miner you want your average costs covered (alpha = alpha_elec + alpha_unit), which implies an average ROI =>1 (1). If we are in a situation where miners expect an average ROI lower than 1, they would not invest. The ROIshould be averaged over the entire period (e.g. 365 d).
RealMcafee
⋅
@lorschd, Thanks for the feedback. I think I do that already. the electrical constant is per day anyway and the unit constant is acquisition costs which are written off over 1 year. It is somewhat arbitrary. But I think it is a stronger requirement than if I chose a shorter time period: that would raise the 1:1 level even higher. Miners older than S9 are completely unprofitable and S9s themselves (even if you just look at electrical costs) are no longer profitable most days. For the bigger megawatt operations they will be replacing and repairing units a lot, so break even is more of an important factor for them. I'm not sure what the average lifespan of a unit is, but I assume it is not more than 2 years.
Tbh I think mining with current hardware is unprofitable. That is part of the weird situation (drive.google.com/file/d/1WE24jXfgbasFhsiqhn2_YIXHPSbTh_Na/view) There are rumours that Bitmain has new technology which it is not sharing. Also there is a lot of different news about massive investment: new farms in N America, or the Bitmain IPO (12 bn ! but they haven't issued a real upgrade since 2016). Just a strange situation in the mining sector.
lorschd
⋅
@barclayjames, if you want to assess the marginal costs only, you should not include the “alpha_unit” in your alpha. If we set alpha_unit to zero for the Antminer S9 we yield alpha = 0.06*0.1*24=0.144. If we use 0.144 for your ROI calculations, the ROI should still be well above 1. In other words marginal costs with an Antminer S9 at 0.06 USD/kWh is much lower than 6000 USD/BTC (~2900 USD/BTC; income from fees not considered). Until recently (~5th of June) even the S7 had still lower marginal costs than the market price of BTC. So, using the marginal costs as an indicator for the short-term price floor, we can see even more downward pressure without killing S9-miners facing these electricity rates. That’s the bad news. The good news is that this can only be a temporary situation and we may expect the price to increase again. The ROI, including the hardware expenses, should eventually climb above 1. Unless we assume that a large player such as Bitmain has the strategy to “dry out” other miners. Bitmain has enough liquidity (and gets more through the IPO) to exercise this kind of market power. I would not be surprised.
RealMcafee
⋅
@lorschd, yeah that's the general thought, Bitmain is trying to acquire more share of sector.
Well I disagree. These miners don't last long running 24 hours a day in cramped and hot environments. The components are temp sensitive and even if they don't get fried their efficiency decreases with use. It is important to include acquisition costs imo that is why Bitmain finally slashed the price of S9s by 50% recently. On top of this you must include cooling costs which are also non-negligible. Btw I'm no saying this is the bottom. It started out that way, but Bitmain is a bn dollar company so they can go very low.
The estimate I provide already underplays the cost of production imo.
Garabet
⋅
@barclayjames, You are right. The reported efficiency numbers are usually valid for extremely cold environments with very good ventilation. The efficiency of Antminer S9 drops considerably when its get hot and the machines starts to draw much more electricity. If chips are at ~60-70 degrees, it draws 1400-1450 watts but it starts drawing 1550-1650 watts if chips are operating at ~90 degrees. I can say that the chips easily reaches around 90 degrees if the weather is at 28-30 degrees even if you have quite good ventilation in the room.
Also i don't agree with lorschd. No one except Bitmain itself gets these machines for free and ASICs are not plug&forget machines, as there are maintenance issues. People buys these devices by paying retail price plus shipping cost and if your country is not in Bitmain shipping list, then you need to pay few 100 dollars more for buying from some other unauthorized importer. As for cooling costs, you need use a 150 W/h powered fan to to properly cool down 3-4 ASICs, 400-500 watts for 10x ASICs
Unfortunately thats the case. People should open their eyes and see the truth.