1/ Thread on variance.
PoW mining is critical to Bitcoin network security. Mining, in turn, is subject to 3 major sources of variance, from (roughly) easiest to tame to hardest:
i/ Finding blocks via SHA256 hashing
ii/ Market demand (for BTC & transactions)
iii/ Tech innovation
PoW mining is critical to Bitcoin network security. Mining, in turn, is subject to 3 major sources of variance, from (roughly) easiest to tame to hardest:
i/ Finding blocks via SHA256 hashing
ii/ Market demand (for BTC & transactions)
iii/ Tech innovation
2/ Quick note: contrary to popular belief that miners are bad & evil, they are vital to Bitcoin’s survival. So it’s imperative to foster a healthy mining industry.
3/ The best miners would be the ones who understand the nature of these variances & know how to handle them most effectively.
4/ For variance in finding blocks:
For small solo-miners, the variance in finding blocks (& consequently, the variance in payouts) is too great to make mining practical.
For small solo-miners, the variance in finding blocks (& consequently, the variance in payouts) is too great to make mining practical.
5/ Fortunately, hashing is a Poisson process. This means that on average there are 10 mins between blocks.
The larger your hash rate, the closer you’ll get to this ~10 min/block average, and the more stable your payout curve.
The larger your hash rate, the closer you’ll get to this ~10 min/block average, and the more stable your payout curve.
6/ Solution for variance in finding blocks: join a pool.
Currently, Stratum pools are popular. The downside of Stratum pools is that miners give up the right to propose blocks & harm decentralization. With BetterHash proposal, this will hopefully change.
Currently, Stratum pools are popular. The downside of Stratum pools is that miners give up the right to propose blocks & harm decentralization. With BetterHash proposal, this will hopefully change.
7/ For variance in market demand:
In early-stage Bitcoin, block rewards subsidize network growth & hash rate is a direct function of price.
In late-stage Bitcoin, hash rate is a function of price AND transaction (tx) volume - which together make up the fees.
In early-stage Bitcoin, block rewards subsidize network growth & hash rate is a direct function of price.
In late-stage Bitcoin, hash rate is a function of price AND transaction (tx) volume - which together make up the fees.
8/ In both stages there’s an incredible amount of volatility. These things (market sentiment and real world usage) are cyclical in nature: they ebb & flow.
9/ Demand for BTC typically increases during economic & political crises. Tx volume increases where there’s more economic activity done in BTC. Layer 2 adoption (LN) would eventually drive up on-chain activity.
Vice versa, demand for BTC & transactions can also decrease.
Vice versa, demand for BTC & transactions can also decrease.
10/ I suspect that the issue of variance in on-chain tx volume will come to the forefront once block rewards run out (soon).
11/ Potential solution for variance in market demand: use financial hedges such as insurance and/or future contracts.
12/ Similar to how farmers use future contracts to insulate themselves from bad weather/harvest, miners can use future contracts to insulate themselves from price + tx volume variance.
13/ There’s also potential for AI/deep learning applications. E.g.: an AI-driven prediction engine can help a miner decide when is the best time to increase or decrease hedging position w.r.t. price + tx volume.
14/ For variance in innovation:
Mining is inherently a high-risk business due to large CAPEX cost. Your hardware can go out of date before you can recoup the initial investment.
Mining is inherently a high-risk business due to large CAPEX cost. Your hardware can go out of date before you can recoup the initial investment.
15/ ASIC design is a subset of IC design, which has roughly followed Moore’s law for the last four decades.
However, there are signs that Moore’s law is breaking down as we reach the limits of “miniaturization at atomic levels.”
However, there are signs that Moore’s law is breaking down as we reach the limits of “miniaturization at atomic levels.”
16/ It took us 2 1/2 years to go from 22nm to 14nm (April 2012-Sep2014). It took us almost another 3 years to go from 14nm to 10nm (Sep2014-April2017) - not without problems.
17/ GMO is rumored to have started looking into 7nm ASICs. It'd be curious to see performance benchmarks in June, as this seems to be *well ahead* of the chip industry.
18/ Nevertheless, the trend is clear: we’re running into hard limits.
a/ Practical limits of minimization.
b/ Beyond these limits are other limits: energy efficiency limits, computational speed limits.
a/ Practical limits of minimization.
b/ Beyond these limits are other limits: energy efficiency limits, computational speed limits.
19/ So what’s the best strategy for miners here?
Absence of a major tech breakthrough, when we run into these limits, ASICs will likely become commoditized. In which case this variance will resolve itself out.
Absence of a major tech breakthrough, when we run into these limits, ASICs will likely become commoditized. In which case this variance will resolve itself out.
20/ However, in the short to medium term, miners probably need large investments in chip design / manufacturing / supply chain to stay ahead of the curve.
