For years data centers were focused on providing resources for cloud platforms, hosting services and data storage. Then came the rise of cryptocurrencies which became one of the main clients for data center operators both for colocation and for renting of servers and resources.
But over the past couple of years, things have started to change quite a bit. Cryptocurrency mining was one of the main sources of revenue for data center operators, but now it has become a burden. The main reason? – the rise of AI. Artificial Intelligence is now becoming more and more widespread; the use cases are varying and the demand for computing power is rising constantly.
As such, AI is becoming a bigger priority to the IT industry which means crypto has to step back. Not because crypto is bad, it’s just the main “competitor” for computing power while its value is much more localized. It’s only for people and organizations that are actually investing in crypto and working on specific services and products. AI on the other hand affects everyone. Even for people who aren’t using a service, AI can still bring them value. For example, if AI is used to manage and optimize the energy grid or simply to improve the work of a data center to provide better quality service.
What does that matter to data centers? A lot, actually. Here’s why.
Repurposing data centers
AI computing comes with big demands for both hardware resources and as a result – energy. According to data, gathered by Forbes, AI data centers consumed about 4% of the total electricity in the US in 2022. By contrast, forecasts are saying that by 2030, AI may consume 4% of the total energy for the entire world. This would mean the electricity as much as Japan or Russia. Of course, it’s difficult to know how exactly that will pan out, as AI is still in its infancy and will change dramatically over the next few years.
Cryptocurrency mining also needs a lot of energy. Just the mining of Bitcoin consumes more electricity than Norway. But it brings far less value. According to the estimates, gathered by Forbes, AI computing can be up to 25 times more profit-dense than crypto mining. And they carry a lot less capital risk. It’s easier to forecast the computing needs and to schedule the timing for their deliveries. Crypto is much more volatile and also its hardware is far more focused and specific. AI-related hardware can also be focused, but it’s also easier to be repurposed.
AI data centers though have higher uptime requirements and need robust redundancy measures for energy supply, networking and the hardware itself. The hardware is more delicate and also experiences longer periods of high load. Crypto-focused data centers on the other hand can get away with poorer temperature management. Either way, both types of facilities need to be on the top of their game for networking connections, cyber and physical security.
The data center-related news recently has been focused more and more on crypto mining facilities that are being repurposed to AI data centers. And some crypto operators are even selling their data centers. One of them is Argo Blockchain which in March 2024 sold one of its facilities in order to reduce its debt. Also, the company’s shares plunged by 55% in the first three months of the year. Up to the end of September 2024 they haven’t recovered, but at least they have remained mostly flat and haven’t dropped any more.
As a result, a lot of crypto operators are repurposing their facilities towards AI. The process is not cheap, but it allows for more equipment to be saved and used, thus saving costs for a completely new data center. AI is here to stay so it’s a wise business move to focus more on it, especially right now as it’s still booming and there’s a lot of demand and low supply.
Of course, Crypto is also here to stay, but it’s not an industry priority right now. Plus, it’s undergoing its own transformation by moving away from energy intensive algorithms towards a more sustainable model. And if that process continues, in the not-so-distant future, crypto mining won’t be as energy intensive as it is now.
New business opportunities for old miners
Also, mining crypto is becoming less and less profitable. Bitcoin rewards were halved earlier this year. Crypto is as volatile as ever. The demand for computing power for AI is rising exponentially. It seems like a no-brainer. Especially since there’s more in common between AI and crypto than it may seem. For example, mining rigs built into vast farms are the most profitable way to earn from crypto. And, as it happens, such big facilities are the scale that AI also needs. And the mining configurations, for the most part, are capable of providing good computing resources for AI, as well. One of the few other purposes a mining rig could have.
And crypto mining operators are cashing in. In May 2024, Core Scientific announced a partnership to host over 200MW of GPUs for CoreWeave. This will bring in net profits of $3.78 billion for the company. While it’s difficult to say if that would be far more than crypto would, because of the price volatility, it’s far easier to predict and expect over time. So, such collaborations are great for both sides – AI startups get cheaper access to much needed computing power and crypto operators secure themselves regular revenue.
High power consumption leading to new laws
Crypto mining and AI share another common challenge – they both need a lot of energy and thus draw the scrutiny of regulators and governments. In fact, earlier this year, authorities in Norway introduced a new legislation aimed to introduce some controls to cryptocurrency mining within data centers.
Norway’s government views crypto mining as an undesirable activity, reports The Paypers. So, the law aims to make it a bit more of a nuisance for data center operators to offer such services. For example, the law wants the operators to register and disclose detailed information about the services they offer and the management details. This will allow the regulators to pinpoint the most popular data center activities and act accordingly.
The main reason for the new law is concern over greenhouse emissions due to the energy intensive processes for crypto mining. Terje Aasland, Norway’s energy minister, says the government wants to motivate “serious actors” who contribute positively to society and not just being there looking for cheap energy. This is why the government is singling out crypto mining which historically has been trying to locate to countries with cheap energy. Whether or not these facilities are contributing positively to society is a debate that would require a lot of time. According to a lot of institutions, crypto is not really bringing much value to communities.
And Norway isn’t the only country introducing new laws because of data centers, AI and crypto. Germany has its Energy Efficiency Act which introduces rules for energy efficiency, energy reuse, renewable energy sources, energy management systems and more. Basically, it covers the entire energy process with the aim to reduce the strain on the power grid and motivate all businesses to improve their energy efficiency.
Let’s not forget crypto
While seeing all the refocusing towards AI, it’s easy to say that crypto is “done” and that the industry will move on. But let’s not forget that crypto is still a global market with more than two trillion dollars, according to charts by CoinGecko as of the end of September 2024.
Data by the Cambridge Bitcoin Electricity Consumption Index shows that Bitcoin consumes 145.59TWh of electricity per year. By contrast, Ethereum is only going to consume 5.52GWh, not just because it’s not as popular, but also due to its improved algorithm. And with Bitcoin having a somewhat decent year pricewise, it shows that despite the reduced focus on it by the industry, there’s still a lot of financial power going around in this niche.
And some are trying to use this slowdown to get in on the market now. “People are making expensive mistakes. They are buying equipment without considering essential infrastructural issues,” says Phil Harvey, CEO of mining consultancy firm Sabre56 to Wired. The main one – securing power.
“It can be a three to 24-month process for the utility to allocate power. So we will undertake what’s called a load study and the utility will then review its network to understand what power it has in surplus. Then you can start to identify if this utility is going to meet our ESG requirements or our clients’ ESG requirements. If the answer is yes, then you continue down that path, but if they have high fossil fuels, it’s out of consideration no matter how low that price is because you are going to struggle to operate that facility on a long-term basis,” says Harvey.
So, crypto will continue to “fight” for energy and will have to battle with AI demand for the supply. On the other hand, crypto will also help foster data center innovations and optimizations in the areas of cooling, computing and energy efficiency. All of them will be of benefit not only to AI but all other data center uses. This is why AI and crypto shouldn’t be viewed as two competitors but as collaborators.