I disagree, because now we're back to a 300-day payback time with GPU mining. According to data analysis I did on video card prices vs. mining profit since 2016, once payback improvves to 300-350 days, GPU prices plateau or even start rising again. When you can buy an RTX 3060L for $400 and make $1.32/day of profit @ 9¢ power, many people will buy them again for mining, even if PoS is 'coming soon'.

Assuming ETH price stays flat, difficulty will keep slowly climbing at 2-3%/month like it has since February. That will only mean 6-8% lower revenue, so GPU prices should be flat over the summer.

IMO, it's not a bad time to build a mining rig now, especially after the lead ETH dev gave his opinion that the earliest PoS will come is Q3 this year. The real 'suckers' are those who bought RTX 3060's for $600-700 each in December.

I would not buy Radeon with the current uncertainty around ETH PoS. Unfortunately, I'll have to go with Ngreedia GreedForce yet again.

The only exceptions seem to be coin profit switching or pool hopping (with a custom stratum proxy that seamlessly switches pools). There are some algos where these methods can yield a bigger margin even after accounting for the rental 'premium' and a 1% pool fee. If rental isn't available at good prices, you can always try to go ask individual mining farms to rent a bunch of hashrate for a fixed monthly price. Maybe they'll be happy with a 1-2% premium.

Do you mean 20% per year or 20% over the useful lifespan of the miner? If it's 20% per year, that really sucks in comparison to GPUs. I have had a farm with 200 video cards where only 2 of them failed in a 9-month period. I sent these cards to MSI warranty repair and they returned working ones within 1-2 weeks. This is why I have MSI brand loyalty.

Lesson learned for me, then. If I want to buy used MicroBT M20S's or Avalon 1146's, I should price in a 20% failure rate for the unrepairable problems + the money/downtime cost of fixing the fixable ones.

Do you know if failure rates can be lowered by downclocking or running at lower power usage?

I do have to admit I cherry-picked the ideal scenario. If one doesn't have unrealized capital gain equal to $1m, they should depreciate the miners over 2-3 years instead of in 1 year with Section 179.

The impression I had at first when looking at mining during recessions was power & rent rates were important in order to be able to pay the bills while profit was low. But the more research I do, the more I feel like the residual value of the equipment is more important.

Mining farm A: pays 8 cent power, buys a M30S for $6000, then it's worth $5000 by the end of the year (including repairs). Profit = $2600

Mining farm B: pays 4 cent power, buys a M30S for $8000, then it's worth $5000 by the end of the year (including repairs). Profit = $1730

It looks like buying reliable equipment for a reasonable price is the #1 most important factor, as long as you can keep the doors open if BTC price were to fall by half.

Here's a new spreadsheet with more realistic assumptions. I assumed BTC price will slowly rise by 60% over the period, diff will rise by 25%, and the resale value of each working miner will stay the same at the end.



Unfortunately, HODLing still wins by a hair even when price growth is zero.



I tried many more price increases/declines, and in every single case, HODLing won by a small margin over mining.

---

However, as a bright spot, the GPU spreadsheets are a hell of a lot better. Mining almost always beats HODLing Ethereum. So long as you can sell the cards before PoS, the very high reliability & profit margin at even 7 cent power trounces ASIC mining at free power.

The best part in my view is, since GPU mining is light on power, a normal U.S. house with a 200 amp panel can hold 200 RTX 3060 cards. No need to rent a separate warehouse if power is already 9 cents.

That's a lot of merit!

I have spreadsheets that are more detailed than that, but they get really complicated.

Wow, it's great that you bought in at the perfect time, after the great downturn of 2018. It's excellent that you didn't have to beg outside investors for money.

I did exactly the opposite, which is why my ETH farm went bankrupt. It deserved to go bankrupt, especially because I wasted $$ on 4K monitors, gigabit internet and bought cards at the top of the market. On top of that, most of my investors were the 'get rich quick' type who bailed when the market crashed, and weren't interested in coming back even when GPUs were dirt cheap. I was a dumb college kid at the time who was desperate with FOMO to not let my dream of owning a mining warehouse go while investors were still interested. The only good decision I made was getting a 6 month lease.

Thank god I spent a few days digging through this forum and learning from the veterans. I think I've come to the conclusion that I should not build an ASIC farm until/if BTC plummets to $25k while I already have a GPU farm running.

I should probably spend the next 6-8 months keeping track of the mining market, saving up $, while trying other business ideas first (like software). By then, I will have enough funds to buy at least 250 of my own GPUs if PoS happens. If PoS doesn't happen but BTC is $25k, at least I can buy ASICs cheaper than today's prices. If my software ideas succeeds, then there's no point in building a farm in the first place.

I disagree because it looks like manufacturers these days want to make their equipment more reliable, because they know that the die shrink to 3/4nm will take a long time and customers will want the miners to last longer.

In this spreadsheet, the coins are always sold instantly in order to get cash. The risk of losing coins from a hack is practically zero; you can only lose a day of profit. As for the network, the firewall/switch should be configured so none of the machines can talk to each other. Only a few outbound IP addresses, such as the pools, should be whitelisted. This isn't hard to set up.

On the other hand, hacking might not be the biggest problem HODLing coins. What if you lose the HW wallet?

Assuming no loss in value for the equipment, these are the final positions:

Reinvesting for 12mo: $1.39m + $662k = $2.052m

Always withdrawing: $723k + $646k = $1.369m

Buying BTC: $723k * (1.938 ^ 2) - 20% tax rate = $2.314m

I think the spreadsheet lines up perfectly with your price growth figure. If BTC went up by 93% in a year, the equipment would likely be worth the same price and the revenue would still be $500/day per miner.

So yes, HODLing wins by about 13% in the end in the average year after these U.S. tax considerations.

However, if you expected BTC price growth to be flat or only increase to $50-60k in 12 months, mining would be the way to go. HODLing profit would be near zero.

Most people who have enough money to start a mining farm at this scale are people who earn $250k+/year (engineers, doctors, lawyers, executives) and have saved up that 1 million over the past 10-20 years or so. Their marginal tax rate is about 32%. They typically have other investments with unrealized tax gains like stock shares or real estate. Why not take this opportunity to sell those assets and deduct the $1m in miners against that profit? $250k of gains from their regular job, $350k from their vacation house, then $400k from Tesla stock they bought long ago.

The reason I deducted the 32% upfront is as long as they start the farm before December, they can get their tax refund by February, so it's practically instant. I can edit the spreadsheet, but the difference will be tiny.

If the farm is owned by a public company (ex. Riot Blockchain), they already face hefty tax rates of 21% + 20%, and there are plenty of other crypto companies that can benefit from buying the mining company and deducting the initial losses against their insane profit.

U.S tax law gets much more complicated than this, but rest assured I have spent hours talking to my accountant to understand these numbers.

I appreciate it. I have been digging through old posts/threads in order to understand ASIC mining during different market conditions. So far, it seems like a bad idea, unless I can start with GPUs while I wait 5 months for the electric company to install a 400 kW or 600 kW transformer. There are other business ideas I can spend my time on that cost nothing and are more scalable. IMO, the worst part is I only have enough $$ to pay for buildout, but not to fill the place with equipment.  That means I will have to waste time chasing investors and customers and doing what they want, instead of planning for the future.

What do you use to shut the farm down when grid energy price spikes?

BraiinsOS, various mining pools, a stratum proxy optimizer or two, GPU management platforms (Minerstat/SMOS/HiveOS), energy price curtailment, blah blah...

The key is to develop a killer product nobody ever thought they needed.

It's a long story, but I basically built the warehouse at the peak of the ETH price cycle. The total hashrate was 5 Gh because an investor who wanted to buy 20 Gh quit at the last minute after ETH price started falling.

The value of the video cards crashed and the profitability crashed. Because I didn't know anything about energy markets, the power cost was 11-12¢/kWh. I was unable to pay the rent and my business partner fled.

I wish I had understood the 'difficulty' part long ago. I should've kept mining during 2014, 2015 and 2017. I should've saved at least 25% of my DOGE & ETH. Well, at least I learned my lesson now; which is why I want to try again, with 6 cent power  and cheap GPUs this time.

The newest estimate is Q4 2022. The lead PoS developer predicted that PoS would be deployed in Q3 2022 at the earliest. So yes, I would guess before the end of the year. However, PoS was supposed to happen back in 2016-2017. Miners panicked back then. But it has been delayed over & over again. So don't be surprised if the new estimate is 2023.

It seems to be that most medium-size GPU mining farms (50-500 cards) buy mid-range video cards like RX 5700's, RTX 3070's or RX 6700 XT's. However, the best investment is probably the RTX 3060L 12GB. Just dual mine ETH+CFX to get 45 Mh total. I have a rig of them and it has been my most profitable rig.

Of course, most n00b miners only buy equipment during a gold rush for inflated prices. I was dumb enough to open an ETH warehouse in early 2018. I bulk ordered RX 570 cards for $370 each. When ETH price crashed, profit didn't cover the rent. I defaulted on my lease and the business went bankrupt. I had to sell the cards for $90 at the end of 2018. Whoever bought those cards has made 3-5x their investment back.

The wisest time to buy video cards is right after the ETH PoS crash. The only problem, though, is you probably won't make any profit for a few years. Just HODL the coins, then when the next bull run happens; sell all your coins as well as the GPUs. This is exactly what I did during 2019-2021, and I made a killing. I bought GTX 1060's for $140, made $400 of mining profit, then sold them for $330.

70% is the average, although the actual unlock range can be 60-80%. I recommend dual mining on LHR units since you can attain 90%-95% the profit on the FHR version. Set the power limit to 70-80%.

You are right. Selling equipment that you deducted = considered income, so taxable. Receiving coins without spending them on equipment/power/rent = considered income, so taxable. Selling coins for a price > than the production cost = taxable income.

Yes, the tax does need to be paid at some point in the future. But the point is to delay it for 2-3 years in order to reap the advantages of scaling up. If you can postpone income + sales tax, that means you can buy 30-45% more equipment.

For example, in Texas, the more transformers I buy and the more kW I cram into the warehouse, the cheaper the operating cost is. Let's say rent is $2500/month. At 100 kW, power costs 7¢ and rent is 3.5¢ ($2500/72000 kWh). But at 1MW, power will cost just 5.2¢ and rent is just 0.34¢. By scaling from 100 kW --> 1MW, the total operating cost goes from 10.5¢ to 5.54¢. It's probably 4.5-5¢ at 10MW+ (diminishing returns).

Thus it absolutely makes sense to build up as many miners as possible (preferably during a downturn) in order to get to the megawatt scale. It even could make sense to buy cheaper, older, inefficient ones at first in order to reach that load. After that, one can take profit out of the business (especially during a bull run) and start paying the income tax. That is the best time to break out of the 'infinite loop'.



P.S.: For various reasons, it makes no sense to hide from the U.S. tax authorities and not report the income. Then I can't get all these sweet tax breaks in the first 1-2 years.


Unfortunately in the U.S., HODLing mined coins then selling them later for the 15-20% capital gain tax is not possible. They're considered regular income immediately after they arrive in the wallet.

The only way to take advantage of capital gain tax is if you bought a S19 miner for $10000, deducted the full cost, then sold it for $15000 12 months+. The $10000 is ordinary income while the $5000 is a capital gain. This might make it wise to sell some equipment during a gold rush to not only reduce risk, but cash in on that lower tax rate before BTC price can fall.

I did take a failure rate of 8%/year into account in the spreadsheets. I assumed 8% is the total loss after repairing the miners that can be economically fixed.

Likewise, the HODL scenario accounts for a 1% chance of lost wallet keys or hacks.


I created a new spreadsheet. I changed some of your numbers in order to be more realistic:

1: Reinvesting profit for 12mo, then withdrawing profit for 12mo



2: Withdrawing profit from the beginning



With this simple spreadsheet, reinvesting is a little better than always withdrawing profit.

Of course, it's better to take profit during a gold rush and reinvest during a recession instead of sticking to a single strategy all the time. The depreciation rate, failure rate, S19 price, and revenue will constantly change.

Here's an interesting resource. They've gone 10x deeper than I have, with a complete Monte Carlo analysis:
https://www.aniccaresearch.tech/blog/the-intelligent-bitcoin-miner-part-i

The only shortcoming of their models is U.S. taxes aren't included.

This is why I agree with you. The OP needs to seriously consider changing his system to accommodate video cards with 1 PCIe power connector.

The low-power <75w cards are bad for overall profitability when considering the cost of all the system parts. This is why gaming communities don't like these low-end GPUs either. The best value is usually in the RTX 3060/RTX 3070 or RX 6700 range.

I did build a mining rig with 75w cards in the past. It had eight RX 560 4GB's which did 13 Mh on ETH (the DAG was 3GB back then). I thought I could save $$ on power splitters. But I was totally wrong, since it was 20% more expensive per Mh than a comparable RX 570 rig.

BioFAIL TB250-BTC PRO

Here's why I disagree. Tax laws in some countries give miners an advantage over buying the coin.

Here in the U.S., a farm owner can deduct 100% of the equipment cost in the first year. This is a huge tax advantage worth up to 37%. The owner can keep re-investing all profit back into more miners, then deduct those. At the end of the day, the effective tax rate can be zero. Also if the electric rate is 6¢ while the owner's income tax rate is 24%, the real cost of power is 4.5¢.

In the U.S., there is no upfront deduction for buying a coin. The best case scenario is a capital gain later on, which will be taxed at either 15% or 20%. You may think this isn't so bad, because no tax is owed upfront. But with mining, tax can be deducted upfront, which is much better.

My own spreadsheets say that mining returns are equal to HODLing returns if coin price goes down by 50% or doubles. But when BTC price goes sideways, that's when mining has a big advantage over HODLing. HODLing returns are zero while mining return is 50%+.

Although I can't help with the actual boots-on-the-ground buildout, I can write custom software to automate your operation:

• Optimizing your power cost by changing ASIC power usage based on real-time energy market prices, which can raise profit by 3-7%
• Improving profitability with a coin-switching Stratum proxy (+1-4%)
• Removing the pesky firmware Devfees (+2-5%)
• Developing a custom modeling dashboard to help you choose the best equipment or whether or not to buy equipment, based on crypto market conditions
• If you're interested in GPU mining, building custom rigs for you and optimizing profit by 15-20% over market levels

I have 9 years of experience mining crypto and I started an Ethereum warehouse in 2018. I've been a software engineer for 3 years.

The barrier to entry can be as low as $400 if you buy a single RTX 3060 video card for altcoin mining. Avoid 'cloud mining' contracts, since they will either be a scam or you'll be locked into hardware that can become unprofitable. The cloud provider, even if they're honest, can't adapt to market conditions.

Another alternative if you're in the U.S. is to buy stock market shares of mining farms, like Riot Blockchain. But the problem is double taxation, since the company has to pay 21% if it makes profit, then you have to pay 15 or 20% on top of that.

The cheapest price for a brand new ASIC is probably $3000-6000. The higher your power cost is, the newer the model of ASIC you should get. Be wary of sellers that offer a price cheap enough that the payback period is < 300 days. It's probably a scam.

Use this site to see which ASIC models are the best deal at your energy cost. Older, less efficient miners are higher risk if coin prices go down, but higher reward if coin prices go up.


If you have a gaming PC, you can use your existing video card, or add another video card. But I recommend building a dedicated mining rig with 3 to 10 GPUs.

When it comes to ASIC miners, you just need to connect an Ethernet cable and power cord. The ASIC usually has its own onboard computer

The smaller your mining operation, the less likely you are to get caught. What do you mean by 'not yet legalized'? Most countries either do not regulate crypto mining, and a few outright forbid it (like Communist China). Most nations treat mining as a regular business with regular tax laws. Here in the U.S., the IRS treats coins as regular income that is taxable the moment it's deposited into a wallet.

According to this source, Butterfly Labs sold pre-orders for 60k units at first. I vaguely remember that the average price of the 50 Gh model was $500 or so. That comes out to $30m of funding in total. However, many of these funds went to Butterfly's own warehouse in Missouri. They later failed to pay the power company, then folded in 2013 or so for whatever reason.

Hah, I'm a newbie to ASIC mining and I thought the Bitmain S17/T17 series was a good idea. I thought the price was good for the efficiency. Then I read some threads on this forum, and I discovered what an idiot I was! Better to go with the T19 or a 2 year old Whatsminer/Avalon machine.

I was also dumb for assuming that ASICs are as reliable as GPUs. Even the most reliable ASIC models seem to have failure rates of 5%+ per year. Compare that to 2-3% on GPUs.

Thanks for sharing your real-world experience.

Designing a new ASIC chip, let alone an ASIC miner, will take a team of electrical/electronic/computer/software engineers. I'm thinking at least 5 people in total. It will cost you hundreds of thousands at the lowest.

If you want to build something new for mining, why not write GPU mining software for a new coin or for a CPU coin? Or a stratum profit-switching proxy? There are so many ideas that you can bring quicker to market by yourself, as long as you know programming.

That sucks. I think low-power ASICs like Goldshells or more GPUs are a better choice because you can fit more investment into that power capacity. If money is a problem, though, and the grid power is cheap, maybe you could buy some used Bitmain S9j's then run them for 8-14 hours a day or whenever you have power.