The hype around cryptocurrencies has cooled, and "smart" money is increasingly flowing into AI infrastructure. Major miners are already repurposing data centers for neural networks, with construction costs skyrocketing.
How the global economy is restructuring and whether it's time to abandon traditional Bitcoin mining was discussed by Arseniy Grusha, founder of the Dataprana data center in the U.S., in the "Podcast Society".
Transition from Mining to AI
ForkLog (FL): Are we witnessing the end of the cryptocurrency mining era, or is this just another transformation?
Arseniy Grusha (A.G.): It's definitely not the end. Mining isn't going away, but the industry is undergoing significant transformation. Over the past five years, we've seen industrialization and the entry of large public companies, which has led to a rally in Bitcoin's network difficulty.
Today, giants have entered the AI market, requiring massive computing power. The profitability figures in AI are much better, prompting miners to convert their data centers for new tasks. This trend will only accelerate.
FL: How feasible and challenging is it to convert mining data centers for AI needs?
A.G.: It's very complex and expensive, as entirely different structures are required. The main value of miners lies in their already allocated power and legal connections to the power grid. Everything else has to be built from scratch with a different approach.
The cost difference is staggering: a Bitcoin data center averages $400,000 per MW, while an AI facility costs around $10 million per MW. The 20-25 times increase is due to strict requirements for full redundancy. AI needs complete backups: massive batteries and diesel generators ready to take over the load in seconds.
The design and liquid cooling systems for GPU power are also more complicated. All these systems require professional design and strict certification to ensure uninterrupted operation.
FL: Are backups necessary for uninterrupted operation?
A.G.: Exactly, uptime is crucial. It must be 99.9999%, allowing for only a few minutes of downtime per year.
FL: So, the miners' advantage is just having electricity, while everything else is built from scratch?
A.G.: Yes, the key advantage is access to large amounts of electricity. Energy companies in the U.S. move slowly, and it can take a couple of years to bring 100 MW online from scratch.
Large IT corporations like Microsoft, Google, and Amazon pay a premium for speed to gain an edge in the AI race. Thus, it's more beneficial for them to use existing mining sites and repurpose them for their needs.
FL: In numerical terms, how much more advantageous is it for a miner to switch to artificial intelligence?
A.G.: At Dataprana, we focus on colocation hosting and invest in infrastructure rather than in ASICs, as equipment depreciates quickly. Traditional mining hosting allows for a return on investment in three to four years, which is quite fast.
Investments in AI data centers are 25 times more expensive. Therefore, their payback period reaches six to seven years.
FL: So, the focus is more on the long term?
A.G.: Yes, but the difference also lies in the scale. Mining is a niche business with high volatility due to its dependence on Bitcoin prices, where contracts are typically for a maximum of a couple of years.
AI is a global market with long-term contracts. A hypothetical 10-year deal with Microsoft provides stability, making it easier to attract bank financing for building such facilities.
FL: What will happen to the Bitcoin network if miners massively switch to AI?
A.G.: I estimate that public miners control about 30% of the Bitcoin network, and they are actively moving into AI. This year, we will certainly see a decrease in network difficulty and a drop in hash rate as major players begin to shut down older equipment.
This will result in an oversupply of mining equipment on the market and a consequent drop in prices. This equipment will need to be placed somewhere, so the demand for traditional Bitcoin data centers won't disappear.
Ultimately, the situation will stabilize, but I expect a slow decline in network difficulty over the next year. Hardware manufacturers will also be forced to move into self-mining to utilize their unsold stock.
FL: Bitmain seems to have started cutting prices.
A.G.: Of course, because mining isn't very profitable at the current price and historically high network difficulty. The market should balance out within the next three to six months.
The best time to invest in Bitcoin mining over the next four years will be in the fall. Equipment will become cheaper, the entry threshold will lower, and investors could see returns of 300-400% on their investments.
FL: Are chip quotas a significant issue now? Is it more beneficial for providers to buy directly from Nvidia with long lead times or from resellers?
A.G.: The main buyers of GPUs are companies from the "magnificent seven," which control 80% of the market. They are buying everything to win the technology race or at least not fall behind competitors.
The remaining 20% is divided among small startups and brokers profiting from reselling scarce graphics cards. Chips are evolving incredibly quickly, accelerating the traditional Moore's Law.
Despite the release of new models, even older cards are still holding their value due to the immense demand for AI. Sooner or later, the market will saturate and a decline will occur, but when exactly is still unknown. Currently, the AI segment resembles the golden age of crypto: everything works and makes money.
Energy Issues
FL: Is there a real electricity problem in the U.S. and the world?
A.G.: In the U.S., there is no shortage of energy—Texas alone produces 85 GW, which is more than all of Germany consumes. The problem lies in outdated power grids that weren't designed to deliver massive amounts of energy to specific locations.
Building data centers requires extensive upgrades to substations and power lines, which takes years. To balance the load, authorities require AI companies to invest in creating their own generating stations.
There won't be a global collapse; network adaptation is gradual. However, uninterrupted server operation is critically important: if AI data centers go down in the future, global payments and millions of workflows will stop.
FL: If there isn't a global problem and we can manage with our earthly resources, why are Elon Musk and Sam Altman interested in placing data centers in space?
A.G.: Space placement offers two main advantages: automatic cooling and infinite solar energy without atmospheric losses. On Earth, about 30% of data center electricity is consumed just for cooling systems.
The main downside right now is the complex logistics and high cost of delivering equipment into orbit. To build gigawatt clusters there, rockets need to launch as frequently as planes do today.
I believe we will reach that point in 20-30 years. For now, humanity needs to learn how to effectively construct such unprecedented facilities here on Earth.
FL: There were once fears that AI requests consume tons of water. How true is that?
A.G.: Chips are becoming more efficient, but six months ago, a standard request in ChatGPT consumed about 3 Wh. So, 300 requests cost a data center approximately 10-30 cents, considering electricity and infrastructure depreciation.
That's not cheap, but costs will decrease as scaling occurs. At the same time, demand for computing will grow exponentially, as we currently utilize AI capabilities at barely one percent.
Regarding water, data centers use closed-loop cooling systems. Water continuously circulates through pipes, cooling chips, rather than simply evaporating, so the actual consumption is minimal.
FL: Has AI hit a development ceiling?
A.G.: Definitely not; we are at the very beginning. Everyone is waiting for the emergence of AGI (Artificial General Intelligence), which will surpass humans in all tasks.
Elon Musk predicts AGI will arrive this year, while other experts suggest 2027-2028. This superintelligence will radically change the world and take over a significant portion of human functions.
Economy and Market
FL: Is the U.S. a safe haven for infrastructure developers or a battleground for tough regulatory struggles?
A.G.: The main issue in the U.S. is bureaucracy: construction permits take a long time to obtain, and connecting electricity can take years. Neighbors in remote areas are often opposed to industrial projects, requiring land purchases.
However, regulatory policy is very favorable. The government encourages construction through tax incentives, as leadership in AI is fundamental to U.S. national security.
In the global tech race against China, America is currently winning significantly. I hope this trend continues.
FL: In terms of attracting capital, do you see investors turning towards artificial intelligence?
A.G.: To be frank, Bitcoin mining is hardly of interest to investors today. The industry has become akin to traditional oil extraction with a high entry barrier, where quick returns of x100 on investments are no longer possible.
All capital is currently flowing into AI, as this is the new industrial revolution. In 20 years, routine tasks in every household will be performed by robots, and we will fully delegate our domestic functions to cloud agents.
FL: Do you think AI is already in a bubble?
A.G.: There is definitely an economic bubble, and it will burst—that's a standard market cycle. The situation resembles the dot-com boom: back then, Amazon's stock fell from hundreds of dollars to $6, but the internet still changed the world forever.
About 80% of companies will be washed out of the market because they produce blatantly useless products. The major players will survive the crisis and continue aggressive growth.
I expect the downturn to last about a year, but it won't stretch over decades. AI and robotics are developing so rapidly that they will quickly pull the economy out of recession.
FL: In a recent Citrini Research report, a collapse of the economy due to AI was predicted. What are your thoughts on this?
A.G.: This is a realistic scenario. Consulting firms will lose relevance, and routine white-collar functions will be replaced by automation. However, AI won't replace people—it will replace those who can't use it.
To mitigate the social crisis, we will likely see the introduction of a basic income within the next five years. To fund these payments, governments will need to impose a strict tax on robots for corporations.
The economy will undergo painful restructuring over 10-20 years, accompanied by depressions. However, ultimately, the system will stabilize, hopefully in a positive direction for humanity.
The Future
FL: What are your thoughts on the concept of sovereign AI? Will states build their own closed data centers, and will there be room for private business?
A.G.: Sovereign AI will likely emerge, just as sovereign internet has in China. Today, user data is the main currency and a powerful tool for political influence.
Every country will want to store information about its citizens strictly within its borders. Governments will develop their own chips, AI tools, and search engines.
Creating such secure systems will take governments five to 20 years. The public sector always operates significantly slower than private businesses.
FL: Top three technologies that will change the world in the next couple of years.
A.G.: First, AGI. It will replace many employees, and its impact will be enormous. Second—longevity. Medicine is advancing so that, in the coming decades, life expectancy could rise to 120 years. Third—space. In 5-10 years, we will begin traveling there much more frequently, fundamentally changing the consciousness of civilization.
FL: Give advice to entrepreneurs who want to build infrastructure rather than just write code.
A.G.: The infrastructure business is more complex than online ventures, as it requires huge investments in hardware that can't simply be moved in case of problems.
The main advice: build infrastructure only in a reliable legal environment. Don't venture into unstable countries in pursuit of cheap electricity if there are no guarantees for the safety of your assets.
Build where business can be legally scaled, where laws work predictably, and where investors are not afraid to put their money.
This interview has been condensed. For more insights, check out the full episode:
Subscribe to the podcast: