The question put to the panel at Mining Disrupt Miami 2023 was blunt: is the return on investment really there for hydro and immersion cooling in Bitcoin mining? Over the next hour a manufacturer, a large-scale miner and a cooling-system builder each answered yes, with caveats about capital cost, two-phase fluids and the risk of going it alone.
The session, recorded July 25–27, 2023, was listed with Taras Kulyk, Co-Founder & CEO of Synteq Digital; Irene Gao, VP of Mining at BITMAIN; Taylor Monnig, VP of Mining Technology at CleanSpark; Gary Testa, President and CEO of Engineered Fluids; Paul LI, CEO of Fog Hashing; Tom Rubin, COO of BiXBiT USA; and Stan Dyshko, CEO of Arctic Systems. On the recording the moderator introduces Gao, Monnig and LI by name; the other voices on stage are not identified, so their remarks are credited here only to a panelist. Bitcoin traded at $29,144 during the event. It is $82,535 today.
The Labor Drop That Changes the Math
Gao answered first. BITMAIN has deployed its hydro cooling infrastructure with clients in several regions, she said, the ROI has been good, and the company is offering flexible terms to support expansions built on hydro. A second panelist took the cost objection head on: cost per terahash is higher than air cooling and some operators take that shortcut, but machine longevity and uptime outweigh the premium. The number that stuck was headcount.
We're talking like, you know, 75%, 80% less touches.
Panelist · 3:52
The same panelist said hydro and immersion units do not even need reboots, and that sealed systems keep out the dust, heat, humidity and bugs that kill air-cooled miners. Another panelist framed it as a return to normal: computers were never meant to run in pods with low filtration and no humidity control, and a proper environment lowers chip temperatures, extends lifespan and lets firmware pull more performance out of the hardware. LI, introducing himself from Fog Hashing, agreed that liquid cooling is the inevitable direction but called it only a start, with capex and maintenance problems that the whole supply chain has to solve together.
Automation: Set It and Walk Away
Asked how his team views automation, LI said the real problem is that cooling, mining, hashrate management and site management still run as separate systems with no tight integration, which matters when curtailment or heat reuse require fine tuning across all of them. A panelist running hydro containers from MicroBT and Heat Core described the other side of that gap: set the dry cooler to auto mode with target inlet and outlet water temperatures, set max power and frequency, and let the system handle water pressure and flow.
Once you hit start, you literally walk away.
Panelist · 11:28
That panelist, who said they had mined with air-cooled hardware since 2012, called it a dream come true after years of paying staff to babysit and reboot machines. Another panelist added that automation becomes the cornerstone of the business once a fleet passes tens of thousands of machines, and that holding immersion fluid at a uniform temperature year-round avoids the hot-cold expansion cycles that follow dust and heat as a cause of chip failure.
Battle Scars: Do Not Go It Alone
The moderator asked Monnig for lessons learned. His advice was pre-planning and spending a little more up front; in a previous role he had seen what fixing other people's immersion setups looks like.
Dry coolers aren't going to work in the middle of the summer in Texas.
Taylor Monnig · 14:13
Properly sized pumps and a clear picture of how the electrical system ties together were his other two items. A panelist followed with a plea to new entrants: the capex for hydro and immersion is higher, so do not waste it by going it alone when the community is open about what works. Another panelist warned against proprietary fluids, said both Engineered Fluids and Shell were in the room and that he had deployed tens of megawatts of each, and called any tank, fluid or pump vendor that will not share customer references or performance data a red flag. Gao said BITMAIN is collecting that same feedback: most of its clients have at least a pilot in hydro or immersion, the company now supplies customized hash boards for tank deployments, and it plans to release its own dry coolers so clients avoid expensive mistakes.
So we truly believe that liquid cooling technology is the future of mining.
Irene Gao · 17:55
The $300-a-Gallon Bottleneck of Two-Phase
LI's team tested two-phase immersion two years earlier with a parallel R&D team and dropped it after half a year. Two-phase sheds heat faster and needs no pumps, he said, but the fluid, the quality control and the long-term maintenance cost far more.
So from my point of view, two phase won't be a choice for the mining industry.
Paul LI · 22:25
Monnig took the long view. Two-phase is the most efficient way anyone has found to cool a computer, he said, but the fluid runs about $300 a gallon, 3M has PFAS and regulatory problems and cannot produce it at scale, and today's chip densities do not justify it. He expected that to change over five to ten years as Nvidia and Intel show chips that will not work in single-phase fluid and as startups push two-phase fluid prices down. He also described what density would unlock: a megawatt in a very small tank, and tanks that are themselves the miner, with one control card and a backplane.
I think it'll drive us moving away from the shoe box.
Taylor Monnig · 24:52
Gao said BITMAIN is working on two-phase immersion as well but does not disclose details before release. A panelist who said he and Monnig were aligned again added that the industry always chases higher density and higher hashrate, so two-phase would arrive in time.
A First Megawatt in West Odessa
The most concrete result came from a panelist who had just deployed a first megawatt of MicroBT and Heat Core hydro cooling in West Odessa, Texas, on a dry cooler. At 114 degrees Fahrenheit ambient, hashboard temperatures ran at 60C and chip temperatures in the low 90s, which the panelist put at almost a 20% efficiency gain over air-cooled units.
So we're 20% cooler than the air-cooled units in Texas in full automation.
Panelist · 30:07
It was, the panelist said, the first deployment where everything went exactly right, and a sign that mining hardware is starting to arrive factory-ready the way HPC gear does for a large data center.
The Next Decade: Denser Chips, Colder Rooms
On what excites them, Gao said the next generation of BITMAIN equipment was the question every partner was asking and that the company would announce more at its own conference in Hong Kong in September. Another panelist, not named on the recording, pointed to BITMAIN building an immersion miner, something that panelist said they had been emailing the company about since 2016, to firmware being unlocked for voltage and frequency control, and to two-phase immersion coming back into view. LI, a former university professor who still runs joint labs, said the gains will come at the chip-packaging level, but most of those technologies have no mature supply chain yet, so the industry should pick the most suitable technology rather than the most advanced.
When the audience asked about cooling built into the chip, Monnig described packing boards closer together, laying them out for fluid flow instead of airflow, and building boiler-plate surfaces into the chip package itself; another panelist added that boards could carry more chips if designed only for immersion. On hydro versus immersion, Gao said BITMAIN chose hydro to maximize performance and keep infrastructure affordable, LI said Fog Hashing builds both and considers immersion more mature while hydro carries a leak risk, and a panelist in the hydro camp said the two are close on cost and capacity and called it a Ford-and-Chevy choice. The full session is on the Intelligence Disrupt data centers hub.