The first miner has already run towards AI

[FP91G]

Bitfarms targets zero bitcoin on the balance sheet as it pivots to AI
The company is actively selling bitcoin and redeploying capital into AI-focused data centers as part of a broader transformation away from mining.
Link

The shift reflects a broader trend across the bitcoin mining sector, where companies are repurposing energy infrastructure for high-performance computing and AI workloads. Bitfarms is advancing a 2.2 gigawatt development pipeline across North America and targeting AI-driven revenue beginning in 2027, according to its 2025 full-year results.


Quality and Certainty of Power Assets
341 MW 2 energized and 430 MW secured in markets where capacity can’t be easily replicated
https://investor.bitfarms.com/static-files/881289f7-012c-4b4a-b462-0d7832de272e

[blockhilltop]

me reading this while running a few hundred Th/s

[philipma1957]

Well it could be okay.

Back in 2021 the diff dropped like mad but price filled the drop.

and if you were a miner you ended up earning 2x the coin at around 60% of the price.

the price then got to 69k in late 2021

and the diff took a long time to catch up.

[FP91G]

Here's another fresh piece of news
MARA Mining Company Cut 15% of Its Workforce Amid a Shift to AI and Digital Infrastructure
https://incrypted.com/en/mara-mining-company-cut-15-of-its-workforce-amid-a-shift-to-ai-and-digital-infrastructure/
MARA is carrying out mass layoffs.
The Bitcoin miner cut 15% of its team as it restructures the business around AI.
The wave of layoffs in the crypto industry is gaining momentum.


Does AI and Digital Infrastructure require fewer workers?
In mining, everything is highly automated, and one large workshop or network of containers is serviced by several engineers and a small number of junior specialists on a shift basis.

[philipma1957]

I wouldn’t call it abandoning Bitcoin, more like diversifying mining profits can be unpredictable, while AI/data centers offer more stable income it is the same infrastructure, just a different use case

my guess is the building for ai are not the same buildings mara used to mine with. mara had over 300 containers one megawatt a container.

I am not sure what  the ai is housed in but i do not think they are using the containers.

i suppose they could build a metal building at the texas site and route the power into it.

does anyone have a link for the inside of an ai building?


https://edgecore.com/locations/phoenix/

this building is nothing like 300 mining containers.


some shots of a microsoft ai

https://blogs.microsoft.com/blog/2025/09/18/inside-the-worlds-most-powerful-ai-datacenter/

nothing like 300 containers  that mara used

[Somegory]

It's not a bad thing and I hope that many big farms pivot into running data centers, we need them gone so that Bitcoin mining difficulty can go lower.

Are you guys aware that a 4.8TH solo miner hit a block days ago? It's all over the internet, men some people are born lucky, you won't know until you give a try.

Or maybe 🤔 they are pivoting because the remaining amount of Bitcoin to be mined isn't much like the past? Maybe this is a begining to something as new as when Bitcoin mining was new too? I guess ram sticks will keep going up, sadly.

[philipma1957]

It's not a bad thing and I hope that many big farms pivot into running data centers, we need them gone so that Bitcoin mining difficulty can go lower.

Are you guys aware that a 4.8TH solo miner hit a block days ago? It's all over the internet, men some people are born lucky, you won't know until you give a try.

Or maybe 🤔 they are pivoting because the remaining amount of Bitcoin to be mined isn't much like the past? Maybe this is a begining to something as new as when Bitcoin mining was new too? I guess ram sticks will keep going up, sadly.

we are under 1,000,000 coins as of today.

by 2036 we will go under 200,000 coins.

meaning 99 percent will be mined.

chasing those last coins is less attractive then pivoting to a.i.

[joker_josue]

we are under 1,000,000 coins as of today.

by 2036 we will go under 200,000 coins.

meaning 99 percent will be mined.

chasing those last coins is less attractive then pivoting to a.i.

We have to be realistic. I believe that all these mining companies were aware that they couldn't make a living from it for the rest of their lives. Perhaps AI was the opportunity for them to start adapting to other markets and maintain operations.

Now, I think that companies that develop mining hardware should also start creating more viable and economical solutions for small miners and home miners. Because sooner or later, some of the mining work will have to be done directly by the users. Not for revenue purposes, but to ensure the network's functionality.

[FP91G]

we are under 1,000,000 coins as of today.

by 2036 we will go under 200,000 coins.

meaning 99 percent will be mined.

chasing those last coins is less attractive then pivoting to a.i.

We have to be realistic. I believe that all these mining companies were aware that they couldn't make a living from it for the rest of their lives. Perhaps AI was the opportunity for them to start adapting to other markets and maintain operations.

Now, I think that companies that develop mining hardware should also start creating more viable and economical solutions for small miners and home miners. Because sooner or later, some of the mining work will have to be done directly by the users. Not for revenue purposes, but to ensure the network's functionality.

Why think so long-term if mining companies can recoup their investments in a few years if the Bitcoin price rises favorably?
I haven't yet seen any real stories of investments in artificial intelligence yielding such a return on investment. In 10 years, more modern ASICs will appear that will be profitable, otherwise, mining equipment manufacturers will go bankrupt.

[joker_josue]

Why think so long-term if mining companies can recoup their investments in a few years if the Bitcoin price rises favorably?
I haven't yet seen any real stories of investments in artificial intelligence yielding such a return on investment. In 10 years, more modern ASICs will appear that will be profitable, otherwise, mining equipment manufacturers will go bankrupt.

Yes, that's true too

Now regarding the efficiency of ASICs, I honestly think that they are not becoming that much more efficient.
Yes, they are better than they were 10 years ago, but I think it is in a very spoiling phase.
Maybe it's just my perception.

[FP91G]

Why think so long-term if mining companies can recoup their investments in a few years if the Bitcoin price rises favorably?
I haven't yet seen any real stories of investments in artificial intelligence yielding such a return on investment. In 10 years, more modern ASICs will appear that will be profitable, otherwise, mining equipment manufacturers will go bankrupt.

Yes, that's true too

Now regarding the efficiency of ASICs, I honestly think that they are not becoming that much more efficient.
Yes, they are better than they were 10 years ago, but I think it is in a very spoiling phase.
Maybe it's just my perception.

The first ASIC Antminer S3 had an energy efficiency of 766 J/Th   Modern ASICs currently on sale have an energy efficiency of 9 to 15 j/Th.
However, any industrial evaporation equipment requires additional energy consumption for ventilation or liquid cooling, so overall energy efficiency values ​​will be slightly higher. If manufacturers are no longer able to improve energy efficiency, they will have to lower the price of equipment.

[joker_josue]

The first ASIC Antminer S3 had an energy efficiency of 766 J/Th   Modern ASICs currently on sale have an energy efficiency of 9 to 15 j/Th.
However, any industrial evaporation equipment requires additional energy consumption for ventilation or liquid cooling, so overall energy efficiency values ​​will be slightly higher. If manufacturers are no longer able to improve energy efficiency, they will have to lower the price of equipment.

That's why I said that in the last few years (I'm thinking of 5-6 years), there haven't been significant jumps.

The biggest difference is that today it is a machine to do what used to take three or four, but the efficiency is almost the same.

Now we need to start lowering the price, but I see this as difficult to happen.

[philipma1957]

I think there are two camps.

btc go up in price enough to keep mining alive.

contain btw's price via paper btc forcing a shift from mining.

lots of ways to argue the angles.

btc should have done 210000 block 1/2ings every 6 years not every 4

so

2012 =  2014. 50      to 25 coins
2016 =  2020  25      to 12.5 coins
2020 =  2026  12.5   to 6.25 coins    we would be at 6.25 coins
2024 =  2032    6.25 to 3.125 coins

all of that is no longer possible.

realistically going to 5 watts instead of 9.5 watts is not much help

So the big boy miners are deciding what to do for
2028 1.56 needs 160k to be equal to today which is tight margins
2032 0.78 needs 320k to be equal to " "
2036 0.39 needs 640k to be equal to " "

  once we are at 2036 next jump needs to be 1.28 million

or buy some coins to hedge by selling off power to ai

the fact is they are going to hedge. because we did 4 year 1/2ings not 6 year 1/2ings

we could see a shift to scrypt mining as ltc/doge works better for decades into the future.

time will tell.

[joker_josue]

realistically going to 5 watts instead of 9.5 watts is not much help

In a way, that's the point I wanted to make.

It's true that it consumes less while doing more. But how much more?

The real breakthrough is being able to do the same thing for half the time I used to.

Incidentally, there's now a proposal from Bitmain that somewhat aligns with what I'm trying to say.
The U3S23H model promises to have twice the hash rate of the S23 Hyd model, but it also consumes twice as much power. In other words, it's true that the equipment does more, but in practice it consumes the same amount of energy.

For me, the real breakthrough would be the U3S23H model, with the same hash rate it has today, consuming the same amount of energy as the S23e U2H model.

Because if we look at the last few years, between the S19 and S23 models, we had an average reduction of 5-6 J/T. Which is very little. Yes, it's true, on a large scale it could become significant. But, the cost of energy will need to come down, otherwise it won't be very meaningful. But in reality, it is already very difficult for large mining companies to lower energy costs even further, even before the problems in Hormuz.

I understand that it may be difficult for hardware companies to reduce consumption even further. But as for BITMAIN and others, are they really investing in research to improve? It seems to me that they are simply taking advantage of the natural improvements that the chip market is presenting. With the money they earned, they could have invested more in research laboratories.

[FP91G]

For me, the real breakthrough would be the U3S23H model, with the same hash rate it has today, consuming the same amount of energy as the S23e U2H model.

You're only comparing hardware specifications, but with that knowledge, anyone could become a miner
The Bitmain Antminer S23 Hyd 3U (1.16 Ph) is a relatively new model and very expensive. Not every miner will buy it until they have more information about its performance and stability. In many cases, it's much cheaper to use older, more proven equipment as long as electricity prices allow.

[philipma1957]

realistically going to 5 watts instead of 9.5 watts is not much help

In a way, that's the point I wanted to make.

It's true that it consumes less while doing more. But how much more?

The real breakthrough is being able to do the same thing for half the time I used to.

Incidentally, there's now a proposal from Bitmain that somewhat aligns with what I'm trying to say.
The U3S23H model promises to have twice the hash rate of the S23 Hyd model, but it also consumes twice as much power. In other words, it's true that the equipment does more, but in practice it consumes the same amount of energy.

For me, the real breakthrough would be the U3S23H model, with the same hash rate it has today, consuming the same amount of energy as the S23e U2H model.

Because if we look at the last few years, between the S19 and S23 models, we had an average reduction of 5-6 J/T. Which is very little. Yes, it's true, on a large scale it could become significant. But, the cost of energy will need to come down, otherwise it won't be very meaningful. But in reality, it is already very difficult for large mining companies to lower energy costs even further, even before the problems in Hormuz.

I understand that it may be difficult for hardware companies to reduce consumption even further. But as for BITMAIN and others, are they really investing in research to improve? It seems to me that they are simply taking advantage of the natural improvements that the chip market is presenting. With the money they earned, they could have invested more in research laboratories.

Gear efficiency does not help.

The big companies buy it the hash rate goes up and back to square 1.

Price price price price is what saves a miner.

But etf's and paper btc tend to drop and suppress btc price.

So I would argue that 150k sooner than later is what miners need.

My 2400 power bill would earn 6000 a month  that would be 6000-2400= 3600 a month an no cost to get there.

Rather than earning 3000 a month

Getting gear that burns 1200 a month means 3000-1200= 1800 profit and I would need to buy a ton of gear

I rather efficiency ends at 9 watts.

Then I never need to buy gear to catch up.

I also rather see prices well over 150k.

[safar1980]

And the Trump family is investing in Bitcoin mining.
Trump-linked American Bitcoin shares spike over 12% after announcing more mining power
The Trump family-linked mining and treasury company announced 11,298 additional bitcoin mining ASICs at its Drumheller site.
The firm said the miners were fully deployed at its facility in Alberta, Canada, increasing its fleet of ASICs (application-specific integrated circuits) to roughly 89,242. It also said that the new bitcoin mining rigs contribute an incremental 3.05 exahash per second (EH/s) at an efficiency of 13.5 joules per terahash (J/TH) to its current operational fleet.

[joker_josue]

For me, the real breakthrough would be the U3S23H model, with the same hash rate it has today, consuming the same amount of energy as the S23e U2H model.

You're only comparing hardware specifications, but with that knowledge, anyone could become a miner
The Bitmain Antminer S23 Hyd 3U (1.16 Ph) is a relatively new model and very expensive. Not every miner will buy it until they have more information about its performance and stability. In many cases, it's much cheaper to use older, more proven equipment as long as electricity prices allow.

I know that, I just wanted to point out that we haven't seen significant performance leaps.

Currently, it seems the idea is to have more machines, instead of having the most efficient machines.

Ultimately, what I'm trying to say is that the mining sector needs to look at things differently and seek a different approach.

[FP91G]

For me, the real breakthrough would be the U3S23H model, with the same hash rate it has today, consuming the same amount of energy as the S23e U2H model.

You're only comparing hardware specifications, but with that knowledge, anyone could become a miner
The Bitmain Antminer S23 Hyd 3U (1.16 Ph) is a relatively new model and very expensive. Not every miner will buy it until they have more information about its performance and stability. In many cases, it's much cheaper to use older, more proven equipment as long as electricity prices allow.

I know that, I just wanted to point out that we haven't seen significant performance leaps.

Currently, it seems the idea is to have more machines, instead of having the most efficient machines.

Ultimately, what I'm trying to say is that the mining sector needs to look at things differently and seek a different approach.

Mining companies haven't yet recouped their old equipment, and new water-cooling equipment requires additional, expensive cooling infrastructure. With such low mining profits, companies won't be willing to invest in expensive equipment. Trump's company is an exception in this case.

[FP91G]

Core Scientific (CORZ) reported a $347.2 million Q1 loss. It sold $208.3 million in BTC and closed a $3.3 billion note offering to fund its shift to AI data center operations.
Colocation revenue for AI data centers surged to $77.5 million, becoming the largest business line. Meanwhile, crypto mining revenue dropped significantly to $30.1 million.
The firm's AI pivot relies on a 590 MW contract expansion with CoreWeave, projected for $10.2 billion in revenue over 12 years.
link

A large company is losing $5 million a month. If this trend continues, when will companies start exiting mining? Or is this a way to lower taxes and get aid?