Experienced GPU miner considering an ASIC farm; what should I keep in mind?

I've been in the GPU mining game since early 2013, starting out with a lone Radeon HD 7850, then moving on to several R9 270's during the Dogecoin hype. Things really ramped up during the 2017-2018 Ethereum boom when I had around 200 GPUs (RX 570's and GTX 1060's), thanks to some outside investors. But then, in late 2018, things went south when a major investor ditched us after ETH prices started to drop. We ended up with only 200 cards instead of the planned 1000+, had to default on the lease, and I scaled back to just one rig from 2019 to 2021. Now that I can work remotely in my job (software development), I'm really thinking about relocating to a place with cheaper power and setting up a small warehouse again. I'm in NJ right now and initially thought about upstate NY, but the BitLicense nonsense and state income tax put me off, plus it just feels too isolated. I also looked at New Hampshire and Montana since they don’t have sales tax, but I want to have a life outside of mining too. Oregon and Washington have affordable power, but the time zone difference is a hassle with my job in EST. I'm leaning towards the Dallas, TX area as my main choice, unless someone here has a better suggestion. It looks like a growing hotspot for mining farms and Bitcoin enthusiasts, which should make it easier to find investors. The heat might be an issue, but at least it’s a dry climate. The best power rate I found with ONCOR is about 7.0 cents total (5.2 cents for energy and 1.8 cents for demand). I’m aiming to get day-ahead pricing and hopefully bring costs down to around 6.0-6.5 cents by shutting things down during price spikes. I’m kicking things off with a budget of $60k: $50k for the gear and $10k for...

I don't think GPU mining is guna "die" any time soon, eth or not there will always be something to mine with them, not to mention they are versatile and you can run all sorts of other algorithms (AI, chemical modelling, etc) as well if you ever need to pivot. I'm curious why you want to go with s9's? they are quite old at this point, might get better ROI with something newer. It looks like you are trying to go with index power rate? you gotta be real careful with that, if prices spike and you don't manage the miners in time (can happen overnight), your energy bill will basically ruin a years worth of profitability Does your location already have power available? If you need to have oncor install more power, they are guna charge you potentially a lot ($40k+) and they *might* refund you a minimum of 4 years later.

I 100% agree, of course GPU mining will live on, but right now isn't a good time to enter because of the threat of PoS. I definitely will get back into GPU mining when equipment is cheaper. My strategy is more long term. The problem with the AI/rendering compute services is: (1) there are few customers (2) it requires reconfiguring the rig, which will cost 30% more in equipment + 30% more in power per card. I was thinking that I could buy them in order to get to 100 kW as quickly as possible in order to secure lower per-kWh rates to prepare for the upcoming GPUs or newer ASICs, and especially to attract outside investors. I would sell the S9's for scrap after a year after they serve that purpose. I also think scaling to 100 kW will give me some experience with the construction/HVAC part. Should I move up to a slightly higher tier, like the L3 series, Innosilicon ZMaster or T15? I wanted to go with the day-ahead index, where I always know what the rate will be in 24h and it's not subject to instant spikes like the real-time LMP. I'm not worried about the spikes because I will write custom, automated software to cut off the Internet to the ASICs to power them down. Maybe other farms will want the software. I wrote code to analyze historical index prices on ERCOT, and it looks like I can go from 5.3¢ --> 4.5¢ for energy while only losing 5% uptime. That will barely affect the demand charge. That's a very good point. This is the hard part. I need to find a place with heavy power at a reasonable rent. Thanks for reminding me I can't just expect more power and I always have to keep a reserve of capacity. I would rather spend that $40k on replacing the S9's with more efficient ASICs than on ONCOR, which is a waste if I leave that location. I've definitely seen warehouses that have their own transformers, and they tend to have much heavier capacity. The rates are cheaper too. I need to look for those in particular. I think I also have to find buildings that are wider rather than longer, so that hot air exits right out the back. The point is to demonstrate a working, profitable farm to investors before going bigger. At that point, I would try to lease land and install metal sheds with my own transformer equipment.

This day-ahead index is a specific agreement with the energy seller, not oncor, right?

Yes; the day ahead pricing only applies to the energy generation, which is a competitive market. All energy still flows through the ONCOR lines. Here's an example during peak hours. If I can chop off just this 1 hour every day, it goes a long way. The S9's are obsolete anyway so I'm not worried about the risk of heating/cooling cycles. There's a chance I can't get day-ahead pricing. In that case, the energy would be a fixed 5.8¢. The only way to reduce the ONCOR charges is either higher uptime (higher load factor), or to not use ONCOR's transformers. I think it can go as low as 0.8¢ by taking power straight from the utility poles (5-20 kV), although I expect 1.8¢ in most small warehouses.

I would think that an experienced GPU miner might be a little turned off to ASIC mining for a few reasons.  ASICs tend to devalue faster than GPUs in my opinion, so while their volatile prices might give you more opportunities for profit scenarios, you have to be quick to act because you can be left holding worthless bricks in the end if you aren't nimble.  There is also more dependency on a single currency, whereas GPU mining gives you added flexibility if one coin's price dips or something.  You're also left with less usability for your equipment, so if regulation or market conditions causes ASICs to become worthless, GPUs will still have buyers in gamers and graphic artists.  Personally, I think GPU miners will be a bit disappointed moving into ASIC mining, but if you pick the right entry and exit points, it can be massively profitable.

Absolutely, I know that I'll have to pay more attention to the depreciation of the equipment and the revenue:electric ratio. An older miner with $3 revenue/$2 power is much more risky than a newer model with $6 revenue/$2 power. The more bullish I am on BTC price, the more the balance shifts toward older equipment. I'll but ASICs on at least 3 algos in order to manage the risks of just 1 coin going down. If DOGE crashes, my Antminer L3's are worthless. If Zcash crashes, my A9 ZMasters are useless. But a lot of GPU miners will be disappointed and wish they did ASIC mining instead if ETH PoS indeed comes this summer... I totally agree. My strategy will be buy whatever is 'cheapest' at the moment; ASICs or GPUs. The older ASICs will provide the 50 kW of startup demand I need to get index power rates. Since BTC/DOGE have underwent a small correction, I think ASIC prices are reasonable these days (although not great). Then I can double down on GPUs after PoS. I'm much more confident about a GPU holding its gaming value after PoS than before PoS. A $500 RTX 3060 probably has $150 of gaming value, so it could crash to $150-200. Better to buy that card at $200-250 later than $500 today. The high rewards over the next 2-3 months aren't worth $300 of capital loss. If BTC crashes to $20k later on, the strategy would shift to buying used S17/S19 miners at a discount. If the ASIC coin prices boom while GPU coins stay stagnant, then I'll buy GPUs.

a shame that you didnt live in europe, I would potentially partner with you, I have £250,000 GBP capital ready for similar idea to what you speak of.

We will definitely open up to outside investors in May/June with hosted mining (never any contracts!) for investors outside the U.S. U.S. persons (citizens/residents) will be able to buy non-voting stock, for the greatest tax benefits and returns. I sent you a PM.

Let's assume this is your profit from 1 ASIC https://minerstat.com/hardware/antminer-s9 I didn't see in your business plan the cost of a few good engineers to maintain this junk. At the end of 2019, my profit was almost zero at a price of 4 cents/kwt. You have rental premises and employees. Your competitors will buy new equipment and increase the network hashrate.

Good point, which is why I bought a dozen of these older ASICs to run in my own home garage to see how reliable they are. I know I have to expect a much higher failure rate for old ASICs than for GPUs. The point of starting out with old ASICs isn't to make the highest profit. The goal is to get a high power demand so I can get lower electric rates (6.3-6.8¢), and to get a 24% tax write-off instantly. Then we'll shift to GPUs (after PoS) and newer ASICs like the Bitmain S/T17 series. If revenue goes down, we can increase efficiency to get some of it back. For maintenance, I don't even think I will pay for repair. I would just remove the dead hashboard and sell it for a buck or two, then keep mining. I have also lived through bear markets in mining. You're right that breaking even is very important. I know that a smaller operation like Coinfarm can't match the mega miners on cost, but we can offer benefits and services that big farms don't, along with proprietary tricks than can raise profit by 4-8% on ASICs and 10-20% on GPUs. There won't be any employees. As a programmer, I'm confident I can automate operations by stitching together different services and even bypassing devfees. The most the monthly rent+insurance would cost is $2500, which is 2.3¢/kWh for only 150 kW. Of course, I'll try to find a 300+ kW warehouse, which is doable.

by the way, mind sharing how to avoid those US taxes?

As a Chinese practitioner who mines both GPU and ASIC, I now have about 1,000 pieces 3060TI and 3070, 400 units S19 series. My idea is that I will keep digging with GPU. If ETH turns to POS, then I will dig ETC. There will always be a way. Of course, ASIC mining is my main business. Many Chinese investors are now working in Texas for the chinese  policy reasons, and the electricity price in Texas is completely acceptable for mining. As for the S9 is a very stable machine, the failure rate is not high, but $380 is definitely second-hand,The failure rate will be much higher. So I suggest you buy a refurbished S9j 14.5t, it will be much more stable, about $500 a piece.If you are worried about the price of BTC falling, then I suggest you to buy the Ant S19 series. With Texas electricity costs, you'll always be profitable.

Absolutely. It's not very complicated. Sales tax on hardware: Get a reseller certificate. Avoid sales tax when buying the equipment as long as you sell the equipment by the end of the year, or the end of next year if it's already September/October. However, if your timing is off, you might owe sales tax eventually. But it's better than paying the tax upfront, because you've already bought more miners and hopefully made returns off them. The amount of that tax has been diluted by inflation anyway. Sales tax on electricity: This is much harder and you need to be larger-scale, but some states (like Texas) have exemptions to the 6.25% sales tax for datacenters and other companies like that. Oklahoma has a similar bill coming up. Federal income tax: Depreciate the equipment (ASICs, GPUs, electrical, ventilation, etc.) with Section 179 in the first year. If you sell equipment, roll the proceeds into new equipment before the end of the year. As long as you re-invest profit into capital equipment, you'll never owe income tax. Just be careful with the timing. State income tax: Same as the above. FICA: This is the worst one of all of them. File taxes as a 'Subchapter S' LLC, and pay yourself with distributions. You will save up to 15.3%. There are more details about this method on the web. Bonus deductions: Set up your home address as the official business address. Now, every trip you make to/from the warehouse is deductible. Also, run a few miners in the garage and a bedroom, then deduct 20-30% of the rent as a business cost. Keep track of power usage so you can deduct that too. Finally, if you use personal credit cards to buy equipment, use 1-2 cards for just the equipment and deduct the interest (if any). With mining, you don't need to cheat/lie/hide in order to pay zero in tax! What's the point of cheating if you pay 0 anyway. Thanks for the encouragement. I obtained wholesale prices and it looks like the best option is a used Bitmain S17 Pro, which has decent efficiency but isn't too expensive. I can get the 56T version for ~$2600 each (shipped from the US). They will still break even at my starting electric rate of 6.3-6.8¢ if BTC goes < $20k, and their value won't go down as much as the S9's. The problem with hoping to mine ETC after ETH PoS is that the ETH ASICs (Bitmain E7, Innosilicon A10, etc.) are all going to mine ETC too. It will be more profitable to sell the video card, even at dirt cheap prices, than mine ETC (or any other GPU coins). GPU mining will go into a depression, just like 2019. But that's the perfect opportunity to buy thousands of them cheaply.

This one in particular is interesting, I have noticed it before, I believe the exemptions is for manufacturing business and was not sure if mining could fit in that category. As far as the "larger-scale" requirement, I do not remember seeing anywhere a size / scale requirement, any more info on that or what exactly constitutes large enough to get that exemption?

Dont get me wrong you can still make good money but not coin, mining BTC with asic's. 90% of BTC is mined whereas retail investors are getting penciled out, and the rest are fighting for the last 10% while difficulty is all time high. Thes best bet right now is to simply buy the coin, by time the next price movement happens difficulty will no doubt increase making it all the much more difficult to obtain. The next best thing is alternative mining resources such as jumping on board new minable coins pretty much so as soon as they are announced. Most retail investors are on nicehash or some platform that provides ease. If you're ahead of the curve, you'll have a bag already of one of these coins by time these platforms/institutions adopt. comparable historical coins are things like. CFX(octopus) Firepro Ergo(autolykos) To be clear I'm not harping bitcoin mining I love it, Maybe ill snag one of these intel based rigs if they ever hit a consumer market Main reason I've exited asic farming is what I like to call the bitmain game, design hardware, build hype , say it wont be out for 12 months while running it until you make the next best thing, release the 2 year old gear. This nifty gear wont be out for another 8+ months yet here's a rack of em running (pic from reddit might be fake) https://ibb.co/JjStKTv By time you get one and have mined it a year and made 13-17K halving happens and other new gears out or announced. 3P upgrades or simply buying buildings as close to you can by substations. At the end of the day its great money, but its an endless cat and mouse game in less than 6 years over 98% of bitcoin will be in circulation. I'm planning ahead and just buying coin, or trading shitcoins for it that are much easier to accumulate and acquire. Or simply selling. Whatever suits my situation.

It looks like the customer needs to go through a long process: https://electricityplans.com/electricity-sales-tax-exemption/ Probably only worth the time/expense for farms larger larger 1 MW. The extra 8% probably comes out to 2-4% less profit. I agree that ASIC mining, especially SHA-256 mining, is a very competitive industry. I already know that a small operation can't compete on SHA-256 in the long run. The ASICs are a means to an end. Power consumption needs to be shored up at first to get good electric rates. My end goal is to sell SaaS, consulting services, GPU colocation, speculative GPU/CPU hosting, etc. These are products that I think Coinfarm can actually compete on, and I have way more experience mining small coins.

antminer s9 380$ per unit?This is too expensive!

This was the price back when I started the thread. The prices have declined obviously. I decided not to start an ASIC mining farm, so it doesn't matter anyway.