Network difficulty can't continue at 75% per month...

[yottahash]

Assuming we stick with silicon chips (i.e. there's no immediate commercialised breakthroughs in optical and quantum computing), increases in mining difficulty will have to taper back from 75% per month...

But if the difficulty were to continue rising at 75% per month (i.e. exponentially), then we'd reach a total bitcoin network hashrate of:

• 1 PetaHash/s (10^15) by October 2013
• 1 ExaHash/s (10^18) by November 2014
• 1 ZettaHash/s (10^21) by December 2015
• 1 YottaHash/s (10^24) by January 2017

The only thing likely is 1 PetaHash/s around the end of this year or early next. The rest is insane.

To put this in perspective, at continuing 75% monthly difficultly increases, in less than 18 short months (around end 2015) we'd be looking at

• 1 ZettaHash/s, or equivalent to almost 13 YottaFLOPS of processing power in the network (if you make some voodoo math FLOP/Hash assumptions akin to the Genesis block calculator).
• China now has the World's fastest supercomputer (Tianhe-2) at 24 PetaFLOPS. 13 YottaFLOPS would equal 288 BILLION of these suckers.
  
• This is equivalent to 7 TRILLION Playstation 3's, or, 1000 PS3's for every single person currently on the planet.

Granted ASICs void this FLOP comparison, but it's just a comparison to put things in context.

Then there's the rare earth metal resources, manufacturing, electricity, etc needed for so much compute power...

We know ASIC manufacturers will be deploying up to 1 PetaHash/s each in product over the next year.  
Once we get to around 10 PetaHash/s in the entire network early 2014, I personally think the monthly increases in network difficulty will taper off very significantly.
But until then, better factor in some heavy increases in monthly difficulty, perhaps doubling or more in the coming months....

[1714892495]

1714892495

[1714892495]

1714892495

[1714892495]

1714892495

[1714892495]

1714892495

[WiFiPunk]

I've heard of yottabytes, but never thought of yottahashes.
I remember reading an old article on Forbes that mentioned a yottabyte datacenter would take up a space the size of Rhode Island and Delaware.

EDIT: Found this.. http://blog.backblaze.com/2009/11/12/nsa-might-want-some-backblaze-pods/

[DeathAndTaxes]

Well I would drop your hash to FLOPs vodoo math because the combined FLOPS of every single Bitcoin ASIC is exactly zero.

However yes difficulty will eventually slow down.

One way to look for a peak is the electrical operating cost.  Using the GH/J efficiency of an ASIC (pick average one to start) you can find the difficulty where the operating cost = BTC produced.  The bad news is that is a very large number. So while, yes you are correct difficulty can't grow 75% per month perpetually it can for quite a long time. 

Also a rise in exchange rate will (eventually after some lag) result in an nearly 1:1 rise in difficulty.  So if you calculate the "break even difficulty" that is based on current exchange rate.  If over the next year the exchange rate doubles then difficulty will also double.  Important: potential miners shouldn't consider future exchange rate increases as a basis for buying hardware.  If the exchange rate rises you will also profit (with less risk) by simply holding BTC.  In other words if your math on how you will make a profit includes the exchange rate rising you are doing it wrong.  I only include this to show that difficulty is really based on three factors a) BTC/USD exchange rate, b) device efficiency (GH/$ and GH/J) and c) average electrical cost.

[Cluster2k]

Currently it costs me $3 to create one bitcoin worth $100 using my ASICs with very high power costs.  Someone in China on a commercial power contract would probably be cranking bitcoins out at the cost of 50c each.  Therefore mining will remain profitable for a very long time to come.  The difficulty could be 30 times higher than it is today with no change in bitcoin's price and I would still mine.

You're right that we can't continue at 75% per month.  Eventually sanity prevails and people see ASICs no longer are rivers of gold.  Currently however, they are.  If you can get your hands on a BFL unit right now it's all good.

[polarhei]

Currently it costs me $3 to create one bitcoin worth $100 using my ASICs with very high power costs.  Someone in China on a commercial power contract would probably be cranking bitcoins out at the cost of 50c each.  Therefore mining will remain profitable for a very long time to come.  The difficulty could be 30 times higher than it is today with no change in bitcoin's price and I would still mine.

You're right that we can't continue at 75% per month.  Eventually sanity prevails and people see ASICs no longer are rivers of gold.  Currently however, they are.  If you can get your hands on a BFL unit right now it's all good.

This relies on where and providers. Currently BFL is started from zero while the other competitions are not. TSMC for Avalon,Something not named. Kncminer is with OrSoc. People trust Kncminer because of OrSOC, well known company which help some people to design video system, especially in on-the-fly ones.

I am still testing if cloudhashing is not doing the job even cloudhashing is working on this. However, the size is getting smaller, the power remain same, also the arithmetic! I think the 75% grow still continues, but not only one one target called bitcoin but others.

No one is stupid through, They obtain the bitcoin box first, first purpose of these machines. After they have paid off in fiat, We shall see the smaller one will be filled with these array.

[qiuness]

i agree... mining is still profitable. Certainly not 10x but a decent 2x ROI should be made off for ex. KNC

[byronbb]

The second reward halving is going to be interesting.

[FeedbackLoop]

There are multiple companies producing multiple petahashes right now. I see no problem with 1 Exahash by 2014.

[Ytterbium]

The next stage in all of this is people starting their own fabs just to make bitcoin hashers, along with extra silicon purification plants. Moore's law gets kicked into overdrive as the profitability from IC R&D goes up.

Soon you'll be able to buy silicon solar panels with embedded 5nm, ultra-low power hasher units that convert sunlight into hashes - by the square meter.

[DeathAndTaxes]

There are multiple companies producing multiple petahashes right now. I see no problem with 1 Exahash by 2014.

At 1 Exahash even assumming free chips you would produce less in Bitcoins then the electricity was worth.  There is an upper bound based on the efficiency of units and the exchange rate.  For the same reason we never saw 1 PH/s using GPU we won't see 1 EH/s using ASICs unless the exchange rate goes up ... a lot.