Dollars spent in increasing network speed

[dbbit]

So I know that conventional graphs has us at 7.5 billion difficulty (58000 ph/s) in a year.

However, there is an aspect to consider. It costs real money to increase the network capacity, while the amount of BTC/month that can be earned through the entire network is approximately constant.

So currently, there is: 25 btc x 10 times per hour = 180'000 btc per month up for grab = $54 million (at $300) per month.

So let's see what month-by-month doubling costs. Starting with a basis of 4phs at beginning of November, and let's try and double that each month.

Nov: +4phs = $44m (KNC at $11)
Dec: +8phs = $40m (BitMine / HashFast at $5)
Jan: +16phs = $56m (CoinTerra at $3.50)
Feb: +32phs = $96m (CoinTerra at $3)
Mar: +64phs = $128m (BlackArrow at $2)
...

After this point, there is a very real prohibitive cost to double the network. Getting 50'000 people to buy $100 worth of ASIC Blocks is one thing. Getting 10 million people to spend $100 with no hope of breaking even is something else. The amount of actual money it will take to grow becomes too large for the growth to keep being irrational.

So one of 3 things have to give:

1) Manufacturers will have to start cutting mining hardware prices by 50% per month. So that means a 2TB miner selling a year from now for $25 (1.25c /ghash). I don't see that happening - laws of physics and all that.
2) The price of Bitcoin has to double each month. So $76'800 / BTC 1 year from now. Fun.
3) The network difficulty has to taper off significantly starting after March.

My bet is on #3.

[dairy100]

+1 on #3

[Rival]

It is already stalling...

[dairy100]

It is already stalling...

you mean difficulty or the price on mtgox?

[Rival]

It is already stalling...

you mean difficulty or the price on mtgox?

diff

[fizzmine]

So I know that conventional graphs has us at 7.5 billion difficulty (58000 ph/s) in a year.

Are you just extending the exponential growth rate of the past month or so out a year?  There's no way its going to continue like that.  Sure there will be high diff jumps down the line, but there's simply not enough silicon being fabbed to continue that growth indefinitely.

I think the more accurate estimates are based on Bitcoin ASIC machines actually in the pipeline no?

[mvidetto]

So I know that conventional graphs has us at 7.5 billion difficulty (58000 ph/s) in a year.

Are you just extending the exponential growth rate of the past month or so out a year?  There's no way its going to continue like that.  Sure there will be high diff jumps down the line, but there's simply not enough silicon being fabbed to continue that growth indefinitely.

I think the more accurate estimates are based on Bitcoin ASIC machines actually in the pipeline no?

7.5 billion actually isn't all that bad, but I think endgame as you've pointed out can be anywhere in the range of 5b-10b.  Also, this depends significantly on how much 1 btc is going for, the higher the price the higher difficulty goes.

[fizzmine]

I wouldn't say there's an endgame, hash power should continue to grow until the end of time (or bitcoin).

But, doubling in power every month is a short term trend.  There's many factors that play into overall network hash rate, but ultimately the limit will be on the technology available.  These rapid increases we're seeing are because Bitcoin ASICs are still relatively new and their chip technology is quite a ways behind mainstream microprocessor tech.

As ASIC fab process get closer and closer to currently available general purpose processor tech, that growth rate will slow.

In one year we've gone from 110nm process ASICs (chip tech circa 2004) to possibly 28nm process ASICs (chip tech circa 2011).  That's 7 years of chip technology crammed in under 12 months!  Of course we're going to see huge hash rate increases.

If bitcoin prices continue to rise, we'll probably see ASIC die processes just about caught up to mainstream processors at some point next year (newer processes are quite a bit more expensive than "ancient" ones, so the demand created by BTC price has to be there).  From then on, hash power growth will be a function of what's left of Moore's "law", the number of miners buying new hardware, and BTC prices. 

Heck at 28nm, that's already caught up to AMD's chip technology.  The amount of R&D necessary to efficiently and reliably package high performance chips at smaller and smaller nodes goes up quite a lot.  Intel is only on 22nm for their mass-market processors.

We may never see difficulty jumps over 40% again, and if we do they will be few and far between.  So I guess I agree with the OP in difficulty jumps tapering off and I would go so far as to say the peak may have already happened for a one-month timeframe.