ASICs And Future Profitability

[Andijviesoep]

Many people are wondering how profitable bitcoin mining will be in the future. After doing some simple and conservative calculations, I myself have become very pessimistic. I mainly focus on ASICs since those are the major factor to reckon with right now. Below I present my results (some slightly advanced math is involved).

My reasoning is as follows, it can be skipped if it seems too daunting.


-----

First of all check out the daily network growth rate graph from bitcoin.sipa.be:

http://bitcoin.sipa.be/growth-small.png

From this graph one can see that the total network computation speed since january 2013 has been increasing with a daily rate roughly between 0.5 and 2 percent.

Since the total network computation speed is more or less linearly correlated to difficulty, I assume it can be taken as an accurate measure for relative difficulty.

I then determined the relative future difficulty, for varying daily difficulty increases. Basically all it amounts to is using mathematical formulas for exponential growth to determine the relative future total network computation speed, and as such the difficulty and profitability.

I used the following formula to calculate the relative profitability P:

P = 1/[(1+0.01*y)^x] = (1+0.01*y)^-x

Here P is determined by y, the percentage daily increase, and x, the number of days passed since day 0. Using this formula I generated some numbers, as shown in the following spreadsheet:

https://docs.google.com/spreadsheet/ccc?key=0AkVZJqx_r-WWdGlMMGZ3TjE4NmkzMjlscm5QYzdBMHc&usp=sharing

From this spreadsheet I also generated the following more illustrative graph:

http://anonymouse.org/cgi-bin/anon-www.cgi/http://anonymouse.org/cgi-bin/anon-www.cgi/http://anonymouse.org/cgi-bin/anon-www.cgi/http://img404.imageshack.us/img404/4104/pqor.png

Both the sheet and graph show the relative profitability for daily difficulty percentage increases between 0 and 2 percent, and for up to 120 days later. In the graph, the x-axis represents the number of days passed, the y-axis represents the percentage daily difficulty increase. The z axis represents P, it runs from 1 (100%) to 0.

The x-axis represents days, the y-axis represents daily difficulty increase in percentage, from 0.0% up to 2.0%. You can use the graph to visualize the decline from day to day in profitability for a given daily %difficulty increase. Start on a point on the top of the shape, above the green line (y-axis, x=0). Now travel parallel to the red line (increasing x, means increasing days). This will take you downwards into the 'valley'. Your profitability will decline, the speed of which depends on where you started above the green line (which daily %difficulty increase you chose). The only instance where your profitability does not decline is at the very edge of the figure, right above the red line, where y=0 (no difficulty increase). This of course will not happen. If you start from a point with small y (say y=0.1, means 0.1% daily difficulty increase) and travel along x, up to x=120 (120 days later), you will end up with P around 0.89, a relatively mild 11% decline in profit. However, if you do the same but start at the very edge of the graph, at y=2 (2% daily difficulty increase), you will end up in the corner nearest to the point of view, where P is only 0.095, a 90.5% decline in profit.

The implications of all this are as follows. Even for a moderate 0.75% daily increase in difficulty, after 120 days the relative profitability has already decreased by 60%. I mention 120 days since for most people this seems to be the minimum time to be able to get their hands on an ASIC from KnC miner (see https://bitcointalk.org/index.php?topic=170332.0). These guys are probably the least uncertain option on the market at the moment, and they strive to ship by the end of October. This means that by that time, your ASIC will mine only 40% income of what it would now.

If we shift the daily difficulty increase to a more realistic 1%, relative profitability will have dropped to only 30%. Going to 1.5% (still not unrealistic as more ASICs enter the market) drops it to around 17%. Keep in mind that during this time, except for the lucky few who got their hands on an Avalon ASIC or the rare shipped Jalapeno, you haven’t done any mining yet. You have only been waiting for your ASIC to arrive after paying a large sum of money upfront.

Many will probably say that if you obtain a powerful ASIC such as a Jupiter, the profitability will still be large despite the expected decline. This probably true, but keep in mind that the difficulty will keep increasing after your ASIC starts running. The numbers I showed are only before any mining has started.

Now I will demonstrate some ‘real’ numbers. Say that by some magic I got hold of a Jupiter and today have started running it. I got the following result from http://www.bitcoinx.com/profit/:

Bitcoin difficulty 19,339,258 (difficulty as of writing)
Bitcoins per Block (BTC/block) 25.00
Conversion rate (USD/BTC) 103.77
Hash rate (MHash/s) 350,000.00 (stated by KnC miner)
Electricity rate (USD/kWh) 0.2 (quite reasonable average price)
Power consumption (W) 1000 (stated by KnC miner)
Time frame (months) 3 (doesn't matter for this example)
Cost of mining hardware (USD) 6995 (current Jupiter price)
Profitability decline per year 1 (no decline, doesn't matter for this example)

Result: Revenue per day 944.48 USD

This sounds marvelous. However, keep in mind that the Jupiter will in fact be available 120 days later, by which time, using a very conservative 0.75% daily increase, the revenue per day is ~378 USD (1% daily increase amounts to ~283 USD/day).

This still seems enormous. However, lets now plug in some more ‘real’ numbers. 120 days later, 0.75% daily difficulty increase, will mean that by the time your Jupiter arrives the difficulty is actually around 19339258*1.0075^120 ~= 47407426. Moreover, each year (365 days) the difficulty increases with a factor of about 1.0075^365 ~= 15, amounting to a profitability decrease of 1/(1.0075^365) = 1.0075^(-365) ~= 0.065. Assuming that you will run the Jupiter for at least two years, let’s plug in these numbers instead:

Bitcoin difficulty 47407426
Bitcoins per Block (BTC/block) 25.00
Conversion rate (USD/BTC) 103.77
Hash rate (MHash/s) 350,000.00 (stated by KnC miner)
Electricity rate (USD/kWh) 0.2 (quite reasonable average price)
Power consumption (W) 1000 (stated by KnC miner)
Time frame (months) 24
Cost of mining hardware (USD) 6995 (current Jupiter price)
Profitability decline per year 0.065

Net profit first time frame 40765.93 USD

Over a period of two years this still seems like a great deal. However, lets now assume a 1% daily difficulty increase. The numbers are now

Bitcoin difficulty 47407426 (19339258*1.01^120)
Bitcoins per Block (BTC/block) 25.00
Conversion rate (USD/BTC) 103.77
Hash rate (MHash/s) 350,000.00 (stated by KnC miner)
Electricity rate (USD/kWh) 0.2 (quite reasonable average price)
Power consumption (W) 1000 (stated by KnC miner)
Time frame (months) 24
Cost of mining hardware (USD) 6995 (current Jupiter price)
Profitability decline per year 0.065 (1.01^(-365))

Net profit first time frame 18118.85 USD

Lastly let’s put in a more extreme (but certainly possible) 1.5% growth rate. Using the reasoning from above this leads to (time frame of 24 months)

Net profit first time frame 130.94 USD

It is extremely uncertain if even the most powerful of ASICs will be able to get you a lot of profit. Also, the above assumes that KnC miner will deliver on time, that the value of bitcoin will stay the same or higher, that the ASIC in question will actually manage to keep performing non-stop during at least two years (there will probably not be a lot of time for thorough testing and weeding out potential hardware problems), and that governments will not interfere with bitcoin’s future. There are also many other unknowns, such as shady parties who are assembling their own ASICs for private mining operations. No one knows what consequences this might have. Lastly, when GPU mining stops being profitable, you end up with devices which can be used for a great many other things and which, with some luck, can be sold. It is a low-risk investment. When an ASIC becomes worthless, you end up with a machine that cannot do much beyond warming its immediate surroundings.

Long story short, the only ones who will profit majorly from ASICs are those who managed to get in early and already have one running. The rest of us might be expected to make some profits, but the risks involved are large and uncertain. I myself planned on pre-ordering a Jupiter from KnC, but decided the risk is not worth it. If you’re in it for the good of bitcoin then by all means go ahead. I read the reports from Bitcoinorama about KnC Miner’s open day, they seem to be awesome people. However, if you’re in it purely for the money, it’s probably better to directly invest in ASIC miner or something similar. Just my 0.02. :-)

[gambitv]

So going on your reasoning, TAT.VIRTUALMINE shares are a no go too. So shares in the mining hardware companies seems like the smarter plan, as they are actually winning twofold. 1 by selling hardware, 2 by using it for there own pool. Thanks for the detailed analysis. Very timely considering the drop in hashrate for TAT.V today.

[drlukacs]

Many people are wondering how profitable bitcoin mining will be in the future. After doing some simple and conservative calculations, I myself have become very pessimistic. I mainly focus on ASICs since those are the major factor to reckon with right now. Below I present my results (some slightly advanced math is involved).

It was a pleasure to read your analysis, which is (unfortunately) perfectly correct.

It amazes me that people are willing to spend $2-300 for ASIC USB devices that can only do 336 MH/s. Such devices will never break even at anything near the current exchange rate.

[Andijviesoep]

It was a pleasure to read your analysis, which is (unfortunately) perfectly correct.

Thanks! I added a little bit more explanation below the graph, as I realize this stuff might seem indecipherable for those who did not follow slightly more advanced math courses.

[600watt]

nice work, thanks

[ziptie]

Whether it happens sooner or later, I agree completely with your projections.  Owners of cutting edge mining technology have a huge advantage right now and profitability is great, but it's an advantage that a very small minority enjoys.  Eventually, everybody who wants to participate in mining will have to obtain specialized hardware or they will turn to scrypt coins.  Bitcoin mining will be performed by specialized equipment; it simply won't be profitable otherwise.  A few powerful groups will dominate the end-state of the currency.  It’s a lot less utopian than people think it will be.

[jspielberg]

Scrypt coins will eventually go the same route.

Scrypt is following the exact same path as BTC. 
1st CPU
2nd GPU <<<---- currently here
3rd ... some kind of FPGA
4th ... ASIC

There are companies working on the 3rd and 4th stage right now.
When the economic incentive gets strong enough, you will see (expensive) ASICs for scrypt as well (with high speed access to lots of ram etc.).

My guess is that those working on the new hardware will have learned from BTC and won't be offering it to the public.  If everyone else is walking, and you driving... there isn't much incentive to sell cars if you are paid by the mile.

[excalibur]

Hi Andi, that is a good analysis of the situation about mining, which you posted at https://bitcointalk.org/index.php?topic=237277.0 . Thank you for sharing it. I enjoyed the reading.

I bought into the 6th batch of the Avalon ASIC chip group buy in Germany ( https://bitcointalk.org/index.php?topic=187660.0 ) recently. However, these group buys seem to fill up more and more slowly. At their hight, they filled within 4 days for 10000 chips some weeks ago, but now it is uncertain if they fill up at all, or if they will be refunded in the end.

From the slowdown, most people conclude that everybody with money in their pockets bought already and there is no more funds for mining hardware. I was close to believing this until I calculated again and became doubtful if it's still worth buying. Then I clicked on the "Mining Speculation" area in this forum and immediately saw your thread and the analysis, which backs my opinion. Now I'm almost convinced that the group buys stopped filling because most people are too concerned about the risk they are taking. They might not get their money back.

However, let's take this further.

At latest in about 2 months from now, first of all everybody will have realized that it doesn't make sense to order new ASIC equipment which will be delivered another 3 months later. That means that in 5 months, at the latest, the difficulty increase will not continue with 1.5% per day. The growth will slow down, just like the ASIC sales do right now. That means that if your ASIC miner is still doing a good job in 5 months from now, it might still do a good job in a year from now.

Moreover, in 5 months, when the network hashrate X-whatevered, all owners of sole FPGA-miners will have a very bad time, because they have to realize that their miners, which once were so profitable, are almost useless. When the electricity costs more than the FPGA miner produces, the miner will be switched off and the FPGA will be sold on ebay. We are looking forward to a very low price for FPGAs on ebay very soon. This network capacity will simply be switched off. It will not be replaced, because by then, it's too late to buy into new ASIC miners.

[ixne]

It's very silly to try and take recent trends and extrapolate them out for months or years.  ASICs are a game changer and have only been around since March.  Taking recent difficulty increases and trying to extrapolate them to some linear or even exponential trend is ridiculous, unless you think that some new technology will constantly pop up to replace the prior standard every few months.  Once ASICs are established there will be a new stability point, and it's going to be all about efficiency, not raw hashrate.  There will be minor adjustments for more efficient standards, e.g. 28nm, 22nm, and extremely minor adjustments for better ASIC designs, but unless you see another game-changer on the horizon the current situation cannot continue indefinitely.

Remember that bitcoin mining is a zero-sum game - it's a different model from the vast majority of economic systems where more in = more out, independent of everyone else.

[Andijviesoep]

It's very silly to try and take recent trends and extrapolate them out for months or years.  ASICs are a game changer and have only been around since March.  Taking recent difficulty increases and trying to extrapolate them to some linear or even exponential trend is ridiculous, unless you think that some new technology will constantly pop up to replace the prior standard every few months.  Once ASICs are established there will be a new stability point, and it's going to be all about efficiency, not raw hashrate.  There will be minor adjustments for more efficient standards, e.g. 28nm, 22nm, and extremely minor adjustments for better ASIC designs, but unless you see another game-changer on the horizon the current situation cannot continue indefinitely.

Remember that bitcoin mining is a zero-sum game - it's a different model from the vast majority of economic systems where more in = more out, independent of everyone else.

My predictions (which are indeed based on very rough assumptions; that's why I tried to remain conservative) are based on a period of at least five months, at least certainly a longer period than 'recent trends'. If I were to take 'recent difficulty increases' (say, the last three weeks) then a daily %difficulty increase of 2%-2.5% is much more reasonable. If you would have read until the fifth line of my post you would see why I did not do that.

Now that Bitcoin has turned into a full-fledged gold rush, the ASIC market seems like a volcano ready to erupt. The pressure is building. It seems to be set to pump out as many ASICs as possible, as soon as possible.

In the end, no one knows what is going to happen. No one knows when and in what manner 'the volcano' will erupt. This thread is simply aimed at those who wildly plug the specs/numbers of prospect ASIC products into profitability calculators and see the dollar signs dancing in front of their eyes.

[k9quaint]

It's very silly to try and take recent trends and extrapolate them out for months or years.  ASICs are a game changer and have only been around since March.  Taking recent difficulty increases and trying to extrapolate them to some linear or even exponential trend is ridiculous, unless you think that some new technology will constantly pop up to replace the prior standard every few months.  Once ASICs are established there will be a new stability point, and it's going to be all about efficiency, not raw hashrate.  There will be minor adjustments for more efficient standards, e.g. 28nm, 22nm, and extremely minor adjustments for better ASIC designs, but unless you see another game-changer on the horizon the current situation cannot continue indefinitely.

Remember that bitcoin mining is a zero-sum game - it's a different model from the vast majority of economic systems where more in = more out, independent of everyone else.

You can easily extrapolate current trends with variations out 3-4 months (delivery schedules for the major players in the Bitcoin ASIC space).
That is enough to eliminate many ASIC purchases and auctions as likely to be unprofitable.

Back of the envelope incoming!!!
We got to 20 TH/s on GPUs. Roughly speaking, Gigahashes are the new Megahashes.
So I would imagine the stabilization point of the curve looks like 5,000 - 20,000 Terahash, presuming Bitcoin itself does not unravel and most miners upgrade their GPUs to ASICs.
It will take 40 weeks or so at 10% per week to hit that range.

[madams]

Thank you for the great analysis. Although I don't have an aptitude for math, I can follow your thinking/logic.

To make this more practical/real for us right-brained thinkers, I thought it might be helpful to calculate a real example (mine).

I have been trying to determine if it makes sense for me to cancel my order with Butterfly Labs.

In May 2013, I purchased (1) 7 Ghs (5 with 2 upgrade) and (1) 30 Ghs (25 with 5 upgrade) miners for a total of $3,300 USD. At BFL's current shipping rate, I should receive the units in ~2 months.

With the increasing difficulty and new ASIC miners coming online, does it make sense for me to cancel or keep my order?

Any insights would greatly be appreciated.

[ujka]

@madams:
BFL new policy is 'no refunds':
https://bitcointalk.org/index.php?topic=246182.0

[k9quaint]

@madams:
BFL new policy is 'no refunds':
https://bitcointalk.org/index.php?topic=246182.0

Yep. That makes your decision easy.

[madams]

Well...that makes it easy

I guess I could sell my pre-orders on Ebay for an obscene amount to an unsuspecting miner as well...

Would still appreciate any feedback on profitability, ROI, etc.

Thank you.

[madams]

Thank you! This is very helpful.