Site technicians hauling 35-pound machines all day are what Cassandra Divis called the real proof of work in Bitcoin mining. She made the point on the Mining Operations: In The Weeds panel at Mining Disrupt Miami 2024, recorded June 24–26, 2024. Curtis Harris moderated. With him were Divis of Marathon Digital Holdings (now MARA Holdings), Tim Sandau of Foundry, Cline Kezar of Core Scientific and Bo Ginn of CleanSpark. Between them they described fleets of up to 200,000 machines, and the written procedures, metrics and people it takes to keep them hashing.
At the time of the panel, Bitcoin traded at $61,272. On October 16, 2026, it trades at $82,531. The next Mining Disrupt is March 22–24, 2027, in Irving, Texas.
The Playbook: Replacing Chaos with Discipline
Bo Ginn, director of partnerships and community development at CleanSpark, opened on standard operating procedures. CleanSpark builds sites and also takes over facilities from other partners, and it hires many technicians who have never worked with mining hardware. Its facilities run 24/7, so the procedures cover handoffs between technicians and management teams; without them, he said, a site slides back into the wild west. Safety comes first. Staff are trained to NFPA 70E before they start, "because we acknowledge that they do work in a, you know, highly dangerous setting," and the goal is that technicians go home to their families every day.
Tim Sandau, director of mining operations at Foundry, compared a site without SOPs to backyard football, where teammates run into each other and the quarterback scrambles. Written procedures, he said, turn that into a professional operation in which team members come in every morning, feel confident and know their roles. Foundry's first step with a new site is to write custom SOPs for it: "There's no copy and paste. There's no cookie cutter programs out there."
The way I look at SOPs is like a playbook for the site, right?
Tim Sandau · 5:39
Beyond Uptime: The Metric That Actually Matters
Cline Kezar, senior vice president of data center services at Core Scientific, which he said ran about 745 megawatts and more than 200,000 machines at seven sites in six states, moved the discussion to metrics. A machine can be up while some of its hashboards are down, so uptime alone overstates what a site delivers. Core Scientific now looks at hashrate utilization: how much of the hashrate it owns is actually hashing, and how quickly repaired machines return to the network.
He tied that to economics. With hashprice "sub-five cents," he said, "it is absolutely paramount that you keep everything up and hashing as much as possible." He called utilization versus uptime "a massive piece of alpha" for the audience.
we've started to move away a bit from uptime and really looking at the main metric
Cline Kezar · 7:55
Cassandra Divis said no two Marathon sites have the same expected uptime, because location and curtailment differ. To compare them, Marathon uses a two-dimensional measure, hashrate hours, the way a power bill uses kilowatt hours. A site with one exahash should produce 24 hashrate hours a day, and the company tracks that day over day and month over month.
So something Marathon has done to try to standardize this is like a two-dimensional metric, kind of what you were discussing, where like kilowatt hours, we measure hashrate hours.
Cassandra Divis · 9:42
200,000 Machines and the Art of Asset Management
Kezar spent about a decade at Hewlett Packard, where his biggest year of server deployments was about 10,000 machines. At Core Scientific, he said, that many go in within a week, and Core Scientific had deployed nearly 40,000 machines so far that year.
Tracking that fleet is an asset-management job. Core Scientific has its own software and firmware group and uses a platform built in-house to follow every machine: running on the shelf, in the repair center, going out to be decommissioned, or moving through its logistics hub to be deployed. Fall behind, Kezar said, "and the next thing you know is you're failing your audit" and doing back-end work to fix the inconsistencies.
my largest year of server deployments was about 10,000 machines. We do that in a week in this industry.
Cline Kezar · 16:07
The Grit Factor: Life Inside the Hot Box
Bo Ginn described how CleanSpark hires for that work. New technicians take a standardized course on mining operations, the management system, reports and machine repair before they go into the mining environment. CleanSpark operates mainly in Georgia and Mississippi, in hot and cold weather, and Ginn called it "a labor-dominant position." Applicants are brought out to a site before the later stages of the interview process, so they see the work before they commit to it.
Divis called site technicians the backbone of the business: teams on the ground diagnosing ASICs, swapping parts and hauling machines. On one of her first visits to a container, the airflow pulled her tied-back hair into a miner, "and that's the story of how I got bangs that year." Marathon has worked on single-phase and two-phase immersion so that more of the work is hands-off, she said, but being on site every day still takes grit.
hauling, you know, these are 35 pounds, and the amount of physical labor, that's the proof of work that goes into keeping our hashrate up.
Cassandra Divis · 13:34
Building the Playbook in Real-Time
Asked about challenges, Kezar said there is no established playbook and that no two sites are alike. At minus 40 degrees in Grand Forks, North Dakota, his technicians all want to work in the hot aisle. At 120 degrees in West Texas, nobody does.
Sandau said the hardest part of an operations-as-a-service business like Foundry's is the unexpected. On one site call he expected detailed operations questions; the first thing the team told him was that a cat in the facility had had kittens. He had no SOP for that. The skill is thinking on your feet.
Divis closed on what Marathon brought in-house. It had relied on third parties for work such as logistics, and found that made it hard to adapt quickly, so it now manages more of its fleet itself and keeps a plan A, a backup plan and a backup for the backup. Kezar's version: "We are developing a playbook every single day."
There is no already set playbook. We are developing a playbook every single day.
Cline Kezar · 25:16