Public miners are going in big on immersion technology.
Thousands of megawatts of machinery worth millions of dollars are being dunked in dielectric fluid across North America as big miners predict the future of hashrate and see uneasy times ahead. Difficulty will soar. Hashprice will crash. Stock multiples will suffer. What’s to be done? Deploy more capital now in anticipation of gloomy days.
“The one development that has changed the geographic opportunities in the last 12 to 24 months is there are more and more leaning into immersion,” said Argo Blockchain CEO Peter Wall, speaking at Bitcoin 2022 in Miami, Florida. (Wall announced the imminent online date for Argo's newest deployment of 200 MW of immersion units this past week in a memo to investors).
“And with immersion mining you can go mine in the UAE, you can mine in Texas in the summer when it's hot and not worry about dust and particulate matter in your machines in the same when you are using the air-cooled," he added.
Whether they acknowledge it or not, however, public miner interest in immersion is contextualized by access to public financing. The question then becomes: do public miners know something air-cooled miners do not, or is the immersion play merely a high-stakes bet?
For the uninformed, immersion is the act of dunking an ASIC in a fluid with the intent of increasing performance of a machine over time. Typically, immersion ASICs are aircooled units placed in a dialectic fluid or water. The liquid runs over the hashboards, cooling as it's piped in and out. Some newer generation machines are liquid-cooled from the get-go, like the S19 XP Hydro.
Containers themselves come in a variety of shapes and sizes. Some are proprietary setups using retrofitted air-cooled units. Some are manufactured boxes built for immersion only, such as the M53. And some are completely containerized, like the "liquid-cooled" Antspace and complementary S19 XP Hydro.
Public firms are using both setups, although proprietary models seem to be more common due to the novelty of manufactured containerized options. For example, Riot’s Whinstone facility in Rockdale, Texas, holds some 46,000 S19s in custom computer baths. It’s likely the newly announced site in Navarro, Texas – estimated at 1 GW in size – will be built in a similar fashion.
Other deployments – and those with closer ties to Bitmain, undoubtedly – are setting up Antspaces, such as the 100 MW Poolin deployment this month in Texas.
Are the efficiency changes worth the upfront cost of capital, and in what world do air-cooled miners out-compete immersion units?
It depends on numerous external factors, particularly network difficulty. Public miners foresee difficulty values that price out air-cooled competition. The logic goes as follows: if mining revenue denominated in dollars, or hashprice, squeezes downward, only the most efficient hardware will survive.
“Efficiency going up, and cost going down,” Mike Levitt, CEO of Core Scientific, said at Bitcoin 2022. “It’s a lot less expensive to build an immersion facility than it was a few years ago. The machines themselves are tens of thousands percent more efficient than they were a few years ago.”
Moreover, public miners assume machines themselves will perform better over time, a further savings on machine maintenance.
“Immersion also makes mining hardware last so much longer,” Riot CEO Jason Les said at Bitcoin 2022. “But with immersion, you can help increase that lifespan even longer. From a performance standpoint, that machine is competitive. You can limit the degradation of the quality of the hardware.”
Still, it’s anyone’s guess what will occur with hashrate, and therefore difficulty, by the end of year. Miners straw-polled on the Compass Podcast have frequently called for 300 EH/s-plus by the end of 2022.
Yes, a 250 TH/s water-cooled unit will beat out your M30. It’ll also pay less in daily electrical costs. But there’s one important line item not operating in the watercooler gang's favor: capex.
In many cases, an air-cooled unit will cost between 25% and 50% less than an immersion unit. That initial cost increases the ROI period substantially for immersion-based units. Below, a few simplifying assumptions are made for brevity. The immersion unit is overclocked by 50% (a bit high) and the immersion build is assumed to cost 50% of the ASIC, or $12,000 total. Network conditions include a $40,000 Bitcoin price and 29.79 T difficulty.
Over a 12-month period, an immersion miner can expect to nearly keep pace with an air-cooled unit because of the capex spend. Under different assumptions (lower capex), an immersion miner would certainly beat out the air-cooled unit. But the lesson remains the same: any miner is profitable at today’s hashprices.