Stories · as read on the AI station
The AI Shift in Immersion Mining
Mike Nixon argues that manual adjustments fail as Bitcoin volatility demands AI-driven, real-time optimization for immersion farms.
Turn a miner's power down and the electricity bill drops. Turn it all the way down and the hashrate goes to zero. Mike Nixon, CFO of Arctic Systems, spent his session at Mining Disrupt Miami 2023 on the space in between. Arctic develops immersion mining systems, and Nixon said the talk grew out of conversations with the company's clients.
The session was recorded in July 2023, when Bitcoin traded at $29,144. On October 16, 2026, it was $82,165.
Nixon's starting point was that immersion is more complicated than air cooling, so it had better pay for that complexity. In his words, "As immersion mining is becoming more and more prevalent, the decisions that you have to make in immersion start to overwhelm."
He listed what an operator has to watch. Outside the tanks: ambient temperature, heat rejection, day and night temperature forecasts, summer against winter. Inside: tank temperatures, inlet and outlet temperatures, fan speeds, pump speeds. Then power consumption, hashrate, and the real-time price of Bitcoin. Profit, he said, is income minus expense, but both change with time, so the real question is what the profit is at any given moment.
Many of these client conversations began in the November and December low in the Bitcoin price, when operators asked whether to overclock or underclock. Most chose to underclock, to earn a return on their hardware without running at a loss when the price was low.
His rule for all of it came from graduate school at Illinois in the 70s, where he worked under Nick Holonyak, the inventor of the LED and the first solid-state laser. His professor told him to always think of the system. Nixon said, "You can optimize individual variables, but if you're not looking at the system as a whole, you can lose sight of the overall vision."
Arctic's answer has two parts. Artificial intelligence and machine learning work out the control settings, from inlet and outlet temperatures to fan and pump speeds, down to overclocking and underclocking through the miners' power supplies. Robotic process automation then makes the adjustments across the whole farm. It is easy to control one machine, Nixon said, but a farm with thousands of machines needs new tools. He said Arctic was seeing very promising results, and he invited operators with questions about their own projects to come and find the team.
The full story is on Intelligence Disrupt at intelligencedisrupt dot com, and many more like it play around the clock on Bitcoin Radio.