As the Bitcoin gold rush dries up, cryptocurrency miners are struggling to make ends meet. But for many, there’s a silver lining—the devices they’ve created are perfect for Silicon Valley’s latest obsession with artificial intelligence.
Cryptocurrency mining can be a profitable but highly volatile endeavor. It involves creating massive data centers loaded with specialized computer chips and using them to solve the mathematical puzzles that underlie the security of various cryptocurrencies. In exchange, miners win some cryptocurrency as a reward.
Most miners make most of their money from Bitcoin. But earlier this year, an event called “halving” took a serious hit on earnings. Every four years, the Bitcoin protocol halves the mining reward — how much Bitcoin miners get in exchange for solving math puzzles — to increase the rarity of the coin. Normally, this causes the price of Bitcoin to jump in response, but this time it did not, which has a serious impact on the profitability of miners.
Fortunately for them, another industry with a voracious appetite for computers arrived just in time. The rush to train massive generative AI models has left companies scrambling for chips, data center space, and reliable access to large amounts of cheap energy, things many miners already have in abundance.
“It (typically) takes 3-5 years to build an HPC-level data center from scratch,” JPMorgan analysts wrote in a recent note. Financial Times. “These power struggles put pressure on companies that have access to cheap energy today.”
While cryptocurrency mining and training AI are not exactly the same, they share fundamental similarities. Both require huge data centers dedicated to doing one particular job, and both consume large amounts of power. But since miners have been playing this game for a long time, and most AI companies have only started trying to train really massive models since ChatGPT launched less than two years ago, companies have a big head start.
They have already spent years looking for countries with abundant cheap energy and plenty of space to build large data centers. More importantly, they have already gone through the time-consuming process of obtaining approvals, negotiating energy licenses and commissioning the equipment.
The rapid expansion of demand for AI training is straining networks in some areas, so many jurisdictions in North America have implemented long waiting lists for new data centers. Time. Already, roughly 83 percent of the data center capacity currently under construction is pre-leased, he says Bloomberg.
This means that the biggest hurdle for many AI companies is finding the hardware to train their models, which presents a new opportunity for cryptocurrency miners. “You’ve seen a number of cryptocurrency miners that have kind of struggled and actually completely nailed it,” said Kent Draper, chief commercial officer at cryptocurrency miner IREN. Time.
Turning a Bitcoin mine into an AI training cluster is not a straightforward exchange. AI training is usually done on a GPU, while Bitcoin mining uses specialized mining chips from Bitmain. Often, though, it’s not so much the chips that AI companies are asking for, but the infrastructure and access to power that the mine has already set up.
In June, cryptominer Core Scientific announced it would host 270 megawatts of GPUs for AI infrastructure startup CoreWeave. “We see today’s opportunity in AI as one where we can convert the existing infrastructure we own to host clients who want to install very large GPU arrays for their clients, who are ultimately AI clients,” said Adam Sullivan, CEO of Core Scientific . Bloomberg.
Some miners also run GPUs themselves. German miner Northern Data had already bought $800 million worth of Nvidia GPUs to mine the Ethereum cryptocurrency, but a major software update to the coin’s blockchain in 2022 ended mining and meant those chips remained idle. The company has now turned them into a 20,000-GPU training cluster, one of the largest in Europe, it says Bloomberg.
Other miners like Hut 8 and IREN are investing heavily in new chips to more proactively chase the AI boom. AI training often goes hand in hand with cryptocurrency mining. “We see them as complementary,” said IREN’s Draper Time. “Bitcoin is coming back immediately, but it is somewhat more volatile. AI is customer dependent – but once you have customers, it’s contractual and more stable.”
This new trend could also provide some modest environmental benefits. People are concerned about the enormous power consumption of both AI training and Bitcoin mining. If the growing demand for AI simply displaces existing mining infrastructure, rather than requiring new energy-intensive data centers, it could help limit the growing carbon footprint of the industry.
But for miners, chasing the latest gold rush can be a risky strategy. Concerns are growing that the AI industry is in a bubble on the verge of bursting. If that happens, the rich new miners who started tapping could dry up very quickly.
Image credit: Traxer / Unsplash