Currently, AI workloads can offer profits 25x higher per kilowatt-hour compared to bitcoin mining.
Source for this claim?
This is a very interesting topic. AI data centers are more profitable business than Bitcoin mining and they also offer much better long-term stable contracts and growing institutional backing. But here is the thing, Bitcoin's price is very low right now. If it reaches $100K and higher, Bitcoin mining might recover and AI pivot might slow down. Honestly, I'm shocked that Bitcoin worth so low right now.
Nothing guarantees that an increase in price doesn't mean an increase in hasing pwoer bring bought back online and thus decreasing the revenue earned per th/s, considering how many have pulled out of the mining game, that would mean a ton of cheap gear is now available and ready to be picked or just monthballed and would be ready in weeks to be bought back online.
Companies follow margins, true, but that does not mean mining disappears. Mining has one big advantage a lot of people keep glossing over: it is far more flexible. ASICs can sit next to ugly power sources, they can curtail fast, and they do not need the same pristine uptime and cooling profile that AI workloads want.
It also had a ton of disadvantages
- A crash in bitcoin price leaves you with gear that has absolutely no usage whatsoever
- Cheap power on the other side of the world can put you out of business
- A new and more advanced ASIC will always outperform you
Also, the part about cooling and uptime is misleading, cooling is the result of power used, a 2MW datacenter will need the same cooling as a 2MW mining farm.
the asic gear can run in dirty hot areas better than servers.
so I would argue it is easy to do a one or two container mine. vs a one to two megawatt server setup.
I read all this and you need to think of 2036 we are under 180,000 coins in 2036.
a block is 1.56 in 2028
a bloc is 0.78 in 2032
a block is 0.39 in 2036
a block is .19 in 2040.
so if i am a 20 megawatt guy with twenty containers mining can i dish off 2 to 4 meagwatts for ai as a hedge.
i know I would want to hedge into ai.
here an other reason to hedge ai.
which a big company mara has done.
mara has a mine in texas.
300 mega watts more than 300 1 megawatt containers.
they have a clever power deal with the grid. they will turn gear off when texas grid is over loaded and texas grid pays them not to mine.
so they keep 240 containers and lease power to ai .
they do not turn the ai off only the miners .
this is cheaper then upgrading from s21xps at 13 watts to hydro a 9.5 watts.
so they certainly were going to do 15 to 20% switch to ai.
this means no expense to make more money. as they will let ai company build a site and lease 50-60 megawatts to them.
this eases the burden of changing over 300 containers of single phase s21xps to three phase hydro s23s
but mara is a giant.
look at diff about 155t in november is now around 136t over 10% drop in hash.
but price dropped a lot since the fall.