Bitcoin Forum

I hypothesize that difficulty increase will DOUBLE every 40 days from now until the market is flooded with ASICs

If we use this as a calculation for the difficulty rate in September we get:


15,000,000 x 1.20 ^ 4 =  31,104,000 (difficulty on 16/07/2013)

31,104,000 x 1.40 ^ 4 = 119,489,126 (difficulty on 26/08/2013)

119,489,126 x 1.80 ^ 4 = 1,254, 349,053 (difficulty on 1/09/2013)

These calculations presume that:
120 days till September
12 lots of 10
every 10 days difficulty increases
every 40 days that increase doubles

I hope i am wrong but at these levels even a 350Ghz miner in September wont really be a profitable ROI, even if the price of BTC was 350 USD

Not insane at all.. my models have been at 67-75% difficulty increase every 1 month now for a while.

Recently, i've started using 75% in my calculations. 

Asics the current models will be non profitable by the end of the year.

People with pre orders will get burnt bad with a product coming later in the year that won't make the greedy profit they thought when they first placed the order.

What's  even worst no resale value when the diff is too high.

They will make a good door stop for a Xmas present this year

I guess what people fail to realize is that mining is not a get rich quick ROI.  Obviously the price of BTC is going down versus fiat ever since the crash and will continue to do so for atleast another month or two until it bottoms out around $30.  All the people who bought miners will probably look for refunds (BFL) and while the difficulty will continue to increase quickly, it will probably still only be 100m in September, maybe less.

really? you think the price of BTC will continue going down? I have to say when i look at the chart it does look like it is in a downtrend, since it went parabolic and hit 200s but i don't know how effective chart analysis is for BTC/USD price. Are there any companies making litecoin FPGAs? that would be a good thing to get into now. 

But by then we will only be able to get peoples hand me downs for reasonable cost.  Current ASIC will likely take a long time to not be profitable but the big guys will keep upgrading and selling off.  1BTC/mo off 20TH is still profitable if the price is right.

http://www.google.com/trends/explore?q=bitcoin#q=bitcoin&cmpt=q is why we are going down in price.  There are not enough people buying into bitcoin which is coupled with a consistent selling brings the price down.  This will continue until supply and demand are equal which is roughly 1/10 of the peak price in April.

Your living in fantasy world. All you have to do is take a look at the diff graphs and see the writing on the wall and that's nothing of what is yet to come . People will not want hand me downs if it won't make money.

Your what if scenarios of btc going up to be profitable is like me walking into the casino saying I'm going to win big.

The only winners are people in the know  who have insider access to these devices before mainstream get them, and the people taking money for pre orders. The winners are also the ones offloading them before most people realise it won't make money.

The average person won't make money from mining by the end of the year in btc . Note in my posts I said by end of year.
That's just under 6 months , do the Maths, it's not rocket science .

Hmm wonder if that's a delayed trend line it looks similar to the bitcoin crash
Parabolic bubble it's interesting :)

That means the next time the trendline goes up time for a Buy
Speculation comment

Regarding the difficulty that may be very possible

This entire thread is terribly depressing  :-[

Its all just speculation at the moment. The difficulty might only be 200,000,000 by September in which case ASIC will still be profitable for quite some time.

Hijacking for a quick noob question; if no additional miners were added to the network, i.e we stayed at the current total hash rate. Would the difficulty still increase?

Unless less miners mine more processing power would mean an increase in difficulty as bitcoin releases blocks at around 10 minute intervals
If the price decreased miners might leave the network as it is less profitable
However that is offset by the increase in processing power coming out soon

no

Looks about right except for the last date said 2013 and I think it should say 2014

(difficulty on 1/09/2013)

Not really history is repeating everyone said it would die when CPUs were on the way out.  There are projects offering decently priced ASICS based on Avalon (which will flood the market late summer) so as long as these or similar chips are produced people that have a will will find a way.  I've done the math even if difficulty is 10x what it is I have 10x the HASH power on its way.  It will at that point be exponentially harder to increase difficulty by that factor.

I remember when people said this about GPUs when difficulty was 2M and the price was $15 or so.   :D

If exchange rate goes up, everything remains profitable longer.  If not, it won't.  Anything can and will happen between now and the end of the year.

For example, if Amazon or Google adopts Bitcoin as a payment method or any number of developments occur that warrant an increase in price that is not merely speculative, your comments will look very silly in short order.  But don't make it seem like it's hopeless.  That's pure FUD.  It is a gamble, but not hopeless.

Yep what happens happens and that could possibly mean BTC isn't even necessary/relevant.

Hi, Can you correlate this data with the total hashrate needed to reach this insane amount of difficulty in a so short time frame?

Value of the coin is what matters.  You need more money in the market.  That translates to more people owning bitcoin and not cashing it out.  I think many miners, at first, will cash out as they will need to pay off their investment.  After that point, it's a good bet a lot of them will hang onto coins in order to use them for next generation asic purchases.  (The smart play)  But I expect the market to crash in the short term.  Long term it will stabalize and start growing again as people hang onto coins.

More investment in hardware style wallets is probably a good idea.  Even more the retailers/vendors need some kind of terminal to accept bitcoin and cash out like credit card machines are today.  (Entire services exist doing this making people millions of dollars.  Why not do it for bitcoin?)   If people walk around with a few bitcoins in a digital wallet, that is coins not cashed out in the market, helping boost the market value.

And then, more people have to believe in the currency and invest in it because they know it will grow in value.   Not to mention volatility in fiat markets that cause people to flee to currencies like Bitcoin so they can move money easily.

Instead of looking at doubling every x units of time how about look at how at a point where network will be in equilibrium at current price.

It will require some assumptions but you can also range these assumptions.

Hypothetical assumptions:
* Miners will not buy new hashpower when annual ROIC% is less than 100% (12 month break even).
* Miners operating cost is $Y per TH/s  (take average ASIC efficiency in terms of MH/J and average electrical rate $0.10?)
* Build cost of new ASIC is $X per TH/s.  Look as Avalon chips in bulk, estimated PCB costs, etc. 
* Market forces will eventually push margin on new units to a more reasonable gross profit of 30%? thus the sale price of new units will eventually be X*1.30

You can then estimate the point where the existing hashing power is large enough that the network is unlikely to rise without a change in conditions.  It is also unlikely to fall much because if it does then marginal miners will buy more rigs or turn old rigs back on.   Condition changes could include a significant (say 50% or more) rise in exchange rate, or the release of next generation ASICs (greater than 100% improvement in MH/J and MH/$).

Once you find this "saturation point" you can guesstimate how long it will take to get there and backtrack from there.

It could be some miners are hoping for a breaking point.. ie, it becomes so HARD to mine that MOST miners get out leaving only a hardcore group and eliminating some of the difficulty.  I for one do not think difficulty will continue to climb.. eventually it will go down and only those few left mining will reap any rewards..

It will climb for a while but its not going to be +33% forever.  It will eventually stabilize to a sustainable number. 

Agree, anything parabolic is not sustainable, BTC price or Mining hashing power.

My view is that you shouldn't base your numbers on difficulty increase solely but also look at what is in the order books of Avalon, BFL etc.

Avalon 150T (500K chips on order) + about 80-100T in batch 1-3 miners (not sure whether chips are included in there twice).
BFL - who knows? say 250T (max would be 500T)
ASICMINER could bring another 200T
KNCMINER - say 150T for first 500 (mix n match of 175 and 350GH/s miners).
Bitfury & Others, who knows? say 500T

We're now at around 100T (easy number) and ≃16M difficulty.

So you'd need another 7400T to reach about 1.2B in difficulty. I don't see that happening right now.

um, mining diffiiculty is not exponential, it is linear.

It only looks exponential because we are in a technology transition right now.

It is reasonable to assume that once all the ASIC manufacturers are shipping, they will ship at a constant rate. Thus we won't see difficulty double regularly but rise at a high, but consistent rate.

If you are right, then Asicminer is going to go bust unless they can kill off the competition somehow.

The chips will get much cheaper as volumes rise.
And also as demand falls.
Remember the free market?

I think you are absolutely right. Some of the forecasts around here get to insane numbers not realizing how much hardware would be involved in gaining such astronomical hashes per second.

Every estimate I have seen is based off already paid for devices. These are not numbers pulled from someones ass, they are based on known variables. And the numbers thrown around for BFL are likely on  the low end. The only unknown is when these devices will be delivered. Also just wait til KNCMiner starts production(they have only accepted what? 500 pre-orders?), if they succeed(which looks likely), the network hash rate is going to definitely be stupid high.  Next re-target  is already estimated at 18.2 Million and 890 blocks are still left.  And if the hash rate stays at 150TH, the following re-target would be about 20 million(quite unlikely, I'd put my money on it being 21+ million ).

Their will definitely be a point when the hash rate stabilizes, I'm just hoping not to many devices have been bought and the network is saturated. These devices still use quite a bit of power, so even if they sold dirt cheap, they still need to produce enough money to pay for power and have enough profit to keep operators interested. Before the block halving and the ASIC buzz, ROI was about 6-7 months, I have a feeling we are going to go quite a ways past that and hit at least 12 months ROI, which seems to be above the threshold for operators to keep miners running(at least in the past).

It benefits Bitcoin resilience.

1,254, 349,053 (difficulty on 1/09/2013) correlates to 9,000 Thash. From 150 Thash today.
You really think that's possible in 3 months?
That's 136,363 Avalons!
Or 32,142,857 Avalon chips.
Or 1,800,000 BFL Jalapenos.
Or 180,000 BFL 50Gh miners  ;)
Or 25,714 KnCMiner's Jupiters (not even shipping by that date).
Or 75,000 Bitfury 120Gh miners (also not shipping by that date).
AsicMiner has some 100Th wafers done, but that needs to be assembled yet and put online.

Thanks for that! 
It seems difficult to reach that amount hashrate in place in ~75 days. It will imply 118Thash added daily starting today ....

PS: can you share how did you reach that value?

I can't mine anything at all now with my gaming computer.   I guess unless I have ASICs, or I will just keep what I have and wait for another coin.

Totally agree. Everyone should understand that difficulty is directly proportional to total mining power. Every 1TH/s in mining network equals to 139.698,39 difficulty points, to be exact. For example - if network mining power will increase two times - then difficulty will increase two times too. New difficulty is based on average network hashing power during last two weeks. That's all you need to know about it :)

If difficulty increase would be exponential - that would require all ASIC miner developers to increase their miner production speed every two weeks by the same percent.
Basically, as soon as new mining device is developed and starts to mine - difficulty will grow too. If a lot of these devices are produced - then difficulty will grow rapidly until all devices are connected to network. Then difficulty growth speed will return to previous.

Current rapid increase is caused by some new mining device developer in arena. He has already added mining devices with total power ~40TH/s to network. (first 30 TH/s were added two weeks ago, and that was cause for big difficulty increase last time. And still continues to add devices to network even now). Anyway, next time difficulty will be around 19M (+21%) on 17th of June.

It is exponential, but not on that period.
Moore's law is an exponential factor.
The current difficulty growth also may be transitory and due to other factors than ASICs.
Consider that there are large companies with many tens or even hundreds of thousands of computers that mostly sit idle.
If one or two of those devoted even unused CPU power, it could have such an effect.
Or animation studio render farms between movies with a room full of video rendering hardware + a bored admin and no worries about electricity.  Lots of other possibilities....
This also raises the notion that mining can grow even when it is not profitable under "normal" circumstances.

If we have a 15% growth until September i.e difficulty = ~54M or hashrate = ~392Th...we will need to add ~2.5Th per day to the network starting today. ...hopefully we will not see that. However....thats like ~7 Knc Jupiters per day, i.e ~525 junipers. From yesterday to today only on the Slush pool there was an increment of 4Th

This amount of jupiters are very plausible since they already have more than 800 orders (where 1 order is for example 27 Jupiters)...interesting times!!


excel made by brishtiteveja @ https://docs.google.com/spreadsheet/ccc?key=0AjEbQ5LBKsAEdDYtajZTMi1Ua0ZkUXR0X0NKSnVaSXc#gid=0

BFL have around 400TH/s on their order books. Probably more as this list is inaccurate.  http://bfl.ptz.ro/ so your 2.5TH.s per day averaged increase by Sep should easily be achieved and more.

followed...

This might be interesting for those who want to understand logic behind raise of difficulty:

http://s20.postimg.org/rihqca9yx/ths.jpg (http://s20.postimg.org/56jxiwav1/ths.png)
Network hash power during year (values are taken as average in 5 days). Chart of difficulty per day would look exactly the same. 1TH/s = 139.698,39 difficulty points

http://s20.postimg.org/45jotrtvd/miners.jpg (http://s20.postimg.org/4v2h64uf1/miners.png)
Rough estimation of hashing power distribution between miners

http://s20.postimg.org/p3ijten7d/miners2.jpg (http://s20.postimg.org/70ph26rct/miners2.png)
Known hashing power of miners starting from 01/01/2013

For now most ASIC companies only accept BTC.  Now its nearly required to have ASIC to continue mining.  For this alone I feel the price is about to skyrocket in BTC/$ ratio.  A lot of peoples mining dropped so much they are going to be forced to buy (demand driving up) if they want ASIC.

There are some assumptions built into this analysis that may not hold up.

1) The assumption that the mining is done by ASIC
2) The assumption that there is not free power and free computers (or spare power / spare computers) doing mining.

It is entirely possible that some corporation that has a lot of computers connected and under utilized in facilities with already sunk costs for power, HVAC, real estate, connectivity may have just decided that they want to make use of their spare devices for a very small marginal cost start using them to mine bitcoin.

If this were true, this could have the effect of reducing the BTC/$ rather than raising it.  This is because the mining could be done at such a low marginal cost that the cost per mined coin goes down rather than up.  Further a corp might not need to show a quarterly profit and see that as more important than any long term hold opportunity and have the incentive to sell all mined coins as they are generated.
The secondary effect of this is that ASIC purchases would slow, because they are not as profitable as they once were.

That this may become a more common trend until the ASIC prices and demand falls to a more manageable level.
It would not be a bad thing for Bitcoin, and may even help to entrench it into the existing business environments if this practice continues.

If (for example) someone at IBM, ATT, or Google thought it might be a good idea to make use of their extra systems siting in data centers and rolled out mining installations  to their computing centers globally, you could see a difficulty increase such as this in advance of new ASIC shipments.
Or consider that data center consolidation is an ongoing project, there could be a data center disconnection planned that is delayed for some reason, and enterprising engineers make use of that extra time to get some mining done.
There are many possibilities that exist outside the narrow assumptions that we make thinking that everyone is doing things just in the same way and under the same conditions as we may be doing them.

Think outside the box.
Cognitare Extra Buxum

I'm just talking about currently no ASIC companies accept $ you have to buy them with BTC.  If your smart your lining up access to ASIC if possible.  GPU is already dropping really really quickly.

BFL and KNCminer accept USD

How are you able to determine that GPU use is dropping from these statistics?

It may not yet but I'm already getting 30% of what I was just a few weeks ago.  In another 2-3 months someone running even 10 cards will be barely making anything (like .02-0.04BTC a day).  I think its a safe assumption. There is a MASSIVE amount of Avalon chips coming online.

Those who have large GPU farms that have already been paid off and who receive free electricity will still be mining.

The constant increase in difficulty from December onwards will make it harder for ASIC miners to turn a significant profit.

This discussion is so pointless... There will be no insane difficulty jump to 1.000.000.000 or whatever....
The amount of TH for that would be insanely high; around 9600TH.
If you look at a decent graph you will see that the difficulty rise per day was at most 2,5% since july 12. If you take that as the worst case, which it pretty much is, you will have approx. 791.000.000 at 31/12/2013.
If you take the median, which is around 1,4-1,5%, you will have a difficulty of around 193.000.000 at the same date. I don't think 2,5 percent ís in any way realistic, so here you go.

GPU and even CPU mining may be on the rise rather than on the decline.
Your assumption is that others also have to pay for power, environment, hardware, and etc in the same way that you do.
There exists a tremendous amount of unused computing power sitting idle in data centers with already sunk costs, and owned by corporations.
The amount of already paid for but unused computing power dwarfs the amount of ASICs.  
Corporate IT admins could be turning on mining.

I am trying the wrap my head around how this would affect the market place. So much money is being rammed down the mining pipeline it just screams screw-job. I mean if the price tanks there will be an absolute ton of people with 20k computers that can make $40 a day. I laugh but that's pretty brutal. However assuming they paid in BTC they can just  buy the cost of the rig back in btc and go from there. The whole relationship between miner and price discovery mechanism is hard to figure out for this tired old brain but I keep thinking of a snake eating its own tail.

Price may be more dependent on money flow rather than mining efficiency for the time being.

I really do not understand why some of you are considering for your calculations only 500 Jupiters (+350 GHs by KnCminer).

Wake up guys, KnC may be lying or overpromising, but they have said they will just ship hundreds of units sequentially, and that everybody ordering now will receive his units in September. The "batch 1" myth (a myth KnC itself spread) was debunked by Sam, who said on the record that they will just manufacture as many units as possible and deliver them at a rate of hundreds per day.  In fact, I know for sure that they have some orders for +20 Jupiters and KnC committed to deliver all of them in September, regardless of the position in the preorder queue, etc.

They have currently more than 1,000 paid order, that figure is increasing, and you can bet that many of those orders are for multiple units, some of them for dozens of units. If KnC does what they say (shipping sequentially thousands of units in September), they will be pretty much killing any ROI possibility for all their customers.

Link to Sam's on-the-record statement please?

All I found was:
Which was just a quote of a quote, and out of context, but still seems different than your statement above.
It suggests they expect to follow the pre-order queue for shipping, and do not mention any expectation to throttle production capacity artificially.

It's buried in one of those hundred pages threads about KnC. He pretty much said that the fact they were producing in batches is a myth, they will produce as many units as orders they have, they will obviously respect the queue, but nevertheless they plan to manufacture and ship hundreds of units per day, which means that a) an order placed now is shipped in September, b) being in the first 500 will give you a couple of days advantage, that's all.

The 500 thing made sense as they needed that amount of preorders to make the project happen, but in no way means they will produce in batches.

All the above are promises and wishful thinking, but based on what KnC claims if we decide to count with them for our calculations we shouldn't expect hundreds of units but thousands of them. Which seems kinda worrying for their own customers.

Bullshit. They have 890 orders paid, which we know of. Every order after that will be shipped in a first come first serve manner. Of course there are people with multiple orders. Zephyr has ordered 27 of them. So what? Not every order has multiple Jupiters in it, and not every oder is a Jupiter.  
Again with the math: (why can't you people do that yourself?...)  Let's say every order contains 2 Jupiter units (which it certainly does not.) and there are 1000 paid orders right now. 350x2x1000/1000=700th.
Atm there are 150th=19,3mil diff.    Lets say until shipment there will be 400th.
400th+700th=1100th equals approx 141,5mil diff.    That's ROI alone in around 60 days, depending on your power costs and with the btc for a 100 dollars.

So PLEASE tell me again how ROI is impossible.


And, before I have to read any more idle speculation:
One of these babies will still mine 156,6 Bitcoins in a year if the difficulty reaches 400million. That's 3108th or 8451 Jupiter units for you.

They will have 1,890 orders very soon and they claim to be able to ship sequentially hundreds units per day. And you can bet there are many multiple units orders (and with multiple i mean +10) that have not been advertised on the forum. Looking at BFL you will see how quickly orders can pile up but also how easily promises are broken.

The point is KnC plan seems to be to deploy much more than 700TH in September. Seems to be, because everything is pretty messy and unclear. What you can be sure about is that the time of easy and quick ROI is gone for good - quote me on that. This is going to get hyper competitive.

Bitcoin was not designed to be easy for miners, it was designed to facilitate commercial transaction. Miners tend to forget this as they get wrapped up in their own little world. So do traders, as currency exchanges were never part of the Bitcoin design. Bitcoin is designed to be a pain for miners not a money tree.

OP 's predication for today was off well off.....

True, in fact it seems to me people is overly optimistic when discussing difficulty and ROI for September, they might be blinded by the amazing returns had by Avalon batch 1 customers and ASIC miner shareholders,but the harsh truth is that we are returning very fast to the point in which ROI is hard to achieve and saving every penny on electricity/storage is crucial.

Just check the difficulty chart on bitcoin.sipa.be

Well, he said 1.3 billion difficulty by September, that's crazy and its unlikely.

Comparing BFL orders to KNC is like comparing VW to Ferrari.  BFL orders are pretty cheap, not everyone can order a 7000dollar unit, let alone 2 or more.

Also: Source? Where do you see that they will have that many orders? Not every order made translates to an order paid.

Just checked, my order number(apart from my preorder) was 1800ish. That was within the 890 paid orders. Now it's 2250ish. That's not 1800paid orders. They will get more orders once they deliver the first units, that is true, but I doubt they will order 100000 chips more without payment. So the next units will probably be more efficient and maybe even more powerful, but I don't think that they will be able to deliver them within weeks without chips.

AND even if they can, see post above. 400mil difficulty, still ROI in 220days. We are far from that. very far.
AND nobody cares for a quick ROI. If my unit can make me an ROI in 60 days, fine. If it takes 10 months, still fine.
I will make money off it as long as it is profitable, and that it will be a pretty long time after ROI.

AND if you check the Hashrate from the CPU to GPU switch, you will notice that it only rose about 30-45 times (depending on where you mark the swicth), and that was with way more affordable hardware. Most people bought their gpus one after the other.

 
Correct - the problem with larger process and power hungry hardware is it's longevity. This is why BFL is still winning the race even though they've just started shipping.

Nothing in this thread is new information. We've always expected a spike in hash rate and difficulty upon asic release... followed by a low btc price because of miners cashing out to recoup investments.

The only real question is: after how many cycles will it end?

If you got a BFL single in the next 2 week - it would pay for itself in the first week. Even with new prices - and assuming BFL had stock on hand to ship - how many weeks before the difficulty gets so high that you don't think it's reasonable to buy more hardware. This is what we're dealing with.

But then - the weak hands are already buying inferior tech in a mad race to see who bankrupts themselves first. In some period of time (as the btc price crashes low and velocity slows way down) how many of them will turn of their miners - how many will sell them. (hint: you're seeing people trying to unload avalon hardware for insane markups already). How many will simply run the miners in the red... paying electricity out of pocket to accumulate bitcoin?

The only valid moves at this point are to buy as much 'most power friendly' asic gear as you can afford and mine with it until you've recouped all costs. That or just stop. The sooner the idiots figure this out the more stable the global hash-rate will become.

No kidding.
I don't care about dollars. At all.
The more devices that mine bitcoins there are in the world, the better, regardless of how profitable they are in usd.
Also...
Kenya.
/unpopular opinion

sorry there was a typo. that should not have been a prediction for today but for next month. Actually it is spot on. I said that difficulty would increase every ten days and that percentage increase would double every 40 days hence the difficulty x 1.2 to the power of 4 in the calculations. than means i am factoring in 10 day increases in difficulty. At the moment we are in the 20% every ten days. by 16/07 will will enter the 40% increase phase. Then 40 days from then we will enter the 80% increase.

i originally posted this on the 7th/06

difficulty was about 15,000,000. I said in ten days it would increase by 20%.

15,000,000 x 1.2 = 18,000,000

the next leg would be 27/06/2013, estimated difficulty = 21,600,000

Dont bother with difficulty guys.

Its very simple.

3600BTC/day global

so for example now i have 1GHs unit and global hash rate is 100THs

100……..3600
0.001…….x

x=0.001*3600/100

x=0.036BTC/day

So if the hash rate will be 1000THs in september and i will receive Jupiter from KnC on time it will be like this:

1000…….3600
0.350…….x

x=1.26BTC/day


Dont speculate about difficulty its not important, the number of units in network is not important.

The only important think is your H/s and global H/s nothing else !!!!!!

Its easier to predict the global hash rate than difficulty. Predicting difficulty is wasting of time.

Almost correct. Difficulty is directly proportional to hash rate. If you can predict global hash rate - then difficulty can be easily calculated from it.
1TH/s = 139696,25 difficulty point

correct, but first you need to predict the H/s anyway, not difficlutly

Excuse me, but how did you think that we were projecting difficulty? Obviously estimating total hash rate, and total hash rate is estimated taking into account existing hardware and when it will be likely shipped/deployed.

Really don't get your point, you just stated the obvious??  To get difficulty from total hashrate you just need to apply an easy formula.

sorry for that…my point is that my formula seems to be easier for me, i dont need mining calculator, remember the THs/difficulty point... but its only my opinion no flame please im complicated person :)

The "mining calculator" is useful to a) include your costs (electricity, hardware) and b) factor the network hash rate increase (which equals to profitability decline in terms of constant BTC exchange rate), in order to project the results on a given timeframe.

You obviously do not need any online calculator to know how many BTC will generate X GH/s at X total network hashrate.

I made similar predictions, albeit a lot more conservative. The below graph shows relative profitability from day 0 up to 120 days, for variable daily difficulty % increases. I copied the explanation from my thread here (https://bitcointalk.org/index.php?topic=237277).

http://img404.imageshack.us/img404/4104/pqor.png

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 OP's estimate is on average 5.8% per per day ((2.5+5+10)/3). I think this is extremely unrealistic considering the past daily growth hardly ever exceeded 2%. Despite this I also think ASICs really won't be as profitable as they might seem. Above all they are not really worth it considering the major unknowns in delivery time/durability and also Bitcoin's future, now that it has caught the attention of the main stream media, and more importantly mistrustful governments and shady unkown parties.