Analysis of ASIC earnings, device agnostic

[organofcorti]

I'll just post the basic results here, the full analysis is here: http://organofcorti.blogspot.com/2013/04/913-asic-choices-asic-earnings-17-april.html

Previous post in series here.

4. Conclusions
Earnings estimates are again a bit better than previous estimates - all shipping is a bit delayed compared to my previous best guess, I hadn't counted on ASICMiner dropping their hashrate back to 7Thps. extending the forecast to August (still assuming no batch four from Avalon ) shows that an Avalon that cost 75BTC and started mining on 23rd May should almost have paid for itself by August - about a 2.5 month ROI.

[Viceroy]

Are you considering the shipping of batches of Avalon ASICs in quantities of 10k to 200k chips per run available on demand to anyone? 
And with any tiny degree of accuracy can you extrapolate 24 months?

[organofcorti]

Are you considering the shipping of batches of Avalon ASICs in quantities of 10k to 200k chips per run available on demand to anyone? 

No, I don't have any definitive information about chip batch sales.

And with any tiny degree of accuracy can you extrapolate 24 months?

No. With the inaccurate shipping information available, even extrapolating 3 or 4 months in advance is going to be inaccurate. 24 months in advance, based on current shipment plans? Can't see how I could, sorry.

I do have a guess at how large the network will be once the first generation ASICs reach a sort of "steady state", and that's about 1000 Thps, with a pretty wide margin for error. But I have not idea if this would be achieved in 6, 12 months, 24 months or 36 months, just that it will be the next long term plateau that occurs.

[Minor Miner]

Are you considering the shipping of batches of Avalon ASICs in quantities of 10k to 200k chips per run available on demand to anyone? 

No, I don't have any definitive information about chip batch sales.

And with any tiny degree of accuracy can you extrapolate 24 months?

No. With the inaccurate shipping information available, even extrapolating 3 or 4 months in advance is going to be inaccurate. 24 months in advance, based on current shipment plans? Can't see how I could, sorry.

I do have a guess at how large the network will be once the first generation ASICs reach a sort of "steady state", and that's about 1000 Thps, with a pretty wide margin for error. But I have not idea if this would be achieved in 6, 12 months, 24 months or 36 months, just that it will be the next long term plateau that occurs.

Given that ASICs introduction was a step function game changer, you have a done a great job at modeling the near term.   For the longer term, I think you can only model that by backing into a difficulty where it is no longer that exciting to buy a miner because the ROI is pushed to the 9-12 month range and the return on selling the hardware is pushed down below a 50 GM.   Personally, I think that occurs around 1,800 TH but OBVIOUSLY price steps throw the monkey wrench in this.   Above 250 btc/usd, I think the equilibrium goes higher. 

[organofcorti]

Given that ASICs introduction was a step function game changer, you have a done a great job at modeling the near term.   For the longer term, I think you can only model that by backing into a difficulty where it is no longer that exciting to buy a miner because the ROI is pushed to the 9-12 month range and the return on selling the hardware is pushed down below a 50 GM.   Personally, I think that occurs around 1,800 TH but OBVIOUSLY price steps throw the monkey wrench in this.   Above 250 btc/usd, I think the equilibrium goes higher. 

I think it gets hard to determine ROIs at all - your costs are in fiat, but the cost of the device could be in denominated in BTC or fiat. From the work I've done, a network hashrate at next plateau could easily be 1800 Thps as you suggest. Do you have a time frame for this?

[Minor Miner]

Given that ASICs introduction was a step function game changer, you have a done a great job at modeling the near term.   For the longer term, I think you can only model that by backing into a difficulty where it is no longer that exciting to buy a miner because the ROI is pushed to the 9-12 month range and the return on selling the hardware is pushed down below a 50 GM.   Personally, I think that occurs around 1,800 TH but OBVIOUSLY price steps throw the monkey wrench in this.   Above 250 btc/usd, I think the equilibrium goes higher. 

I think it gets hard to determine ROIs at all - your costs are in fiat, but the cost of the device could be in denominated in BTC or fiat. From the work I've done, a network hashrate at next plateau could easily be 1800 Thps as you suggest. Do you have a time frame for this?

My model is just guess (trying to think of commodities and how people ALWAYS overshoot on capacity -- think chemical industry that even as an oligopoly they never stop increasing capacity until they are two steps too high -- look at TiO2 industry case studies for something that should be rational but goes through irrational capacity expansions).   So, that is just my BS caveat to say (I THINK people should stop at 1800 but I am pretty sure they blow throw it and people get burned and justify their "investment" as a new "hobby").
I think April of Next year.   And I think it happens in STEPS. 

[Viceroy]

No, I don't have any definitive information about chip batch sales.

Here ya go:
http://store.avalon-asics.com/?page_id=9605

Avalon ASIC Chip
The no bullshit, no fine print terms of sale
the only payment accepted is Bitcoin.
these chips will be available until at least end of the year, 2013.
spec and various information is available on the wiki
the chips being sold are packaged and tested.
the lead time on the chips is 9 to 10 weeks.
made to order from TSMC foundry and then packaged and shipped.
the minimum order quantity is 10,000 chips and the maximum order quantity is 200,000.
the chips are identical to those in Avalon, clocking 282Mh/s per chip.
communication protocol, reference board design provided in early May.
everything will be open source from FPGA to PCB design.
sample chips will be provided 4 weeks into ordering.
we do not offer technical support of any kind, this is final.
if you do not know what to do with the packaged chips, please do not purchase.
Please read the above carefully, as with all things Bitcoin one should treat this as an investment and make the decision best for you based on the liquid-able funds available at the moment when placing an order.

[organofcorti]

Given that ASICs introduction was a step function game changer, you have a done a great job at modeling the near term.   For the longer term, I think you can only model that by backing into a difficulty where it is no longer that exciting to buy a miner because the ROI is pushed to the 9-12 month range and the return on selling the hardware is pushed down below a 50 GM.   Personally, I think that occurs around 1,800 TH but OBVIOUSLY price steps throw the monkey wrench in this.   Above 250 btc/usd, I think the equilibrium goes higher.  

I think it gets hard to determine ROIs at all - your costs are in fiat, but the cost of the device could be in denominated in BTC or fiat. From the work I've done, a network hashrate at next plateau could easily be 1800 Thps as you suggest. Do you have a time frame for this?

My model is just guess (trying to think of commodities and how people ALWAYS overshoot on capacity -- think chemical industry that even as an oligopoly they never stop increasing capacity until they are two steps too high -- look at TiO2 industry case studies for something that should be rational but goes through irrational capacity expansions).   So, that is just my BS caveat to say (I THINK people should stop at 1800 but I am pretty sure they blow throw it and people get burned and justify their "investment" as a new "hobby").
I think April of Next year.   And I think it happens in STEPS.  

So is 1800Thps  and April the first step (plateau)? Or after several?

[organofcorti]

No, I don't have any definitive information about chip batch sales.

Here ya go:
http://store.avalon-asics.com/?page_id=9605

Avalon ASIC Chip
The no bullshit, no fine print terms of sale
the only payment accepted is Bitcoin.
these chips will be available until at least end of the year, 2013.
spec and various information is available on the wiki
the chips being sold are packaged and tested.
the lead time on the chips is 9 to 10 weeks.
made to order from TSMC foundry and then packaged and shipped.
the minimum order quantity is 10,000 chips and the maximum order quantity is 200,000.
the chips are identical to those in Avalon, clocking 282Mh/s per chip.
communication protocol, reference board design provided in early May.
everything will be open source from FPGA to PCB design.
sample chips will be provided 4 weeks into ordering.
we do not offer technical support of any kind, this is final.
if you do not know what to do with the packaged chips, please do not purchase.
Please read the above carefully, as with all things Bitcoin one should treat this as an investment and make the decision best for you based on the liquid-able funds available at the moment when placing an order.

Yes, saw that. But I can't make any guesses as to how many they'll sell since I don't know if 200000 chips is the max per order or the max they'll sell ever, and I don't know what demand is like. It also isn't clear when units based on the chips will arrive because this depends not only on shipping times but also on r&d time for the bulk purchaser, and whether or not the bulk purchaser will be selling retail.

So I don't have sufficient information to even guess the number of chips that will be purchased or the time periods over which they'll appear. If Avalon release information regarding chip sales I'll try to add the chips in to the network and difficulty calculations, Estimates of the time when the chips start hashing would be inaccurate, but probably no more than the current estimates are due to BFL's delivery schedule.

[Minor Miner]

I think every two weeks for the next three months we start see steps as difficulty adjusts and avalon plus ASICMiner constantly ship asics.   This gets extended (or the steps get larger) if UPS ever gets called to come to BFL.  I think we plateau next april after the people who get delivery of devices in december take months to realize they may never get their capital back......

[Entropy-uc]

I think every two weeks for the next three months we start see steps as difficulty adjusts and avalon plus ASICMiner constantly ship asics.   This gets extended (or the steps get larger) if UPS ever gets called to come to BFL.  I think we plateau next april after the people who get delivery of devices in december take months to realize they may never get their capital back......

BFL buyers from last year will never get their BTC back.

Mining equipment purchases are always speculation on the exchange price of BTC.  As such you always need to compare buying equipment against market purchases of BTC directly.

People also tend to substantially underestimate the work required to maintain a mining farm with reasonable uptime.

[Minor Miner]

I think every two weeks for the next three months we start see steps as difficulty adjusts and avalon plus ASICMiner constantly ship asics.   This gets extended (or the steps get larger) if UPS ever gets called to come to BFL.  I think we plateau next april after the people who get delivery of devices in december take months to realize they may never get their capital back......

BFL buyers from last year will never get their BTC back.

Mining equipment purchases are always speculation on the exchange price of BTC.  As such you always need to compare buying equipment against market purchases of BTC directly.

People also tend to substantially underestimate the work required to maintain a mining farm with reasonable uptime.

that is basically my point, people wade in hearing there are fish in the water but they do not stop to think if they should learn to swim until they are at least out 8 feet deep.   With a rip tide to boot.   they they complain they were not given a life vest.
that is why these things ALWAYS swing way past where they should have rationally stopped...

[organofcorti]

I think every two weeks for the next three months we start see steps as difficulty adjusts and avalon plus ASICMiner constantly ship asics.   This gets extended (or the steps get larger) if UPS ever gets called to come to BFL.  I think we plateau next april after the people who get delivery of devices in december take months to realize they may never get their capital back......

Maybe. But the idea is not to attempt plot the jumps in hashrate due to deliveries - I did that previously and tbh the inaccuracies in the estimations makes this no more accurate for my method. Instead, I'd use the estimates as a guide over time - I aim for the cumulative earnings per time period to be as accurate as possible rather than the hashrate to get there.

I agree that if the hashrate steps this will have a significant effect on the network hashrate and difficulty and possibly also cumulative earnings, but I don't think anyone can accurately estimate when these jumps will occur. As long as the hashrate shipping estimates are accurate, I think the method I use here is the simplest one that any miner can reproduce for themselves in order to plan for the future.

I'd be interested in seeing how you make your assessment though - I'm always keen to learn new things!

[Viceroy]

Insignificant initially but the ultimate impact will be HUGE.  
ASICMiner is shipping it's first 10 cards.  


Significant shipment of ASICS has begun.  
Avalon has sold more than 30,000 of their chip delivery in 10 weeks
10,000 more here delivery in 10 weeks
Generation 2 Bitcoin mining ASIC? delivery in ? weeks

Of course it will certainly take 10 weeks to build pcb's to mount them too, but let's assume the buyers know what they are doing and they have a plan to use their new toys.  Please tell me, oh great seer of seer's, what will the addition of 100,000 avalon chips (@282mh/s) do to the data and difficulty in 90 days.

[organofcorti]

Thanks for the update, viceroy - I appreciate you going to the trouble of collecting the data and putting it all in one place.

Do you have any clear information about the ASICMiner upgrade schedule? 200 THPS by when?

[Viceroy]

I can't believe it... but it looks like the first BFL's are out. 

I do not, sir.

[Silv0r]

Wow. I can not believe this.. BFL Jalapeno unboxing and demo video oO.. Congratulations to this guy! Buuut.. "1 of 10.000" BFL's out there for customers

Hmm.. Difficult will rise in the next 4-6 weeks I think.

[organofcorti]

I've already taken into account 2Thps of BFL ASICs per day from May 15th. At worst, I may have to bring that forward a couple of weeks. I'll see if many more BFL devices start arriving before I change that date though.

[John Self]

What does the difficulty look like if it hits 1000THps by year end?

What about 5000THps?

[organofcorti]

What does the difficulty look like if it hits 1000THps by year end?

What about 5000THps?

You can approximate difficulty as follows:

Difficulty = network hashrate / 2^32 * 600

At 1000e12 hashes per sec, D ~ 140 000 000
At 5000e12 hashes per sec, D ~ 700 000 000


Of course it is really dependant on the previous retarget's 2016 block hashrate, but this at least will give you an idea.