Russell Cann, Chief Customer Success Officer at Core Scientific, opened his Mining Disrupt Miami 2024 session by counting the room: about 40 people, 10% of them from his own company. He then promised a talk on hashprice volatility and why it matters, and warned it would be dry. The session was recorded at Mining Disrupt Miami, June 24–26, 2024.
Cann set out the scale behind his view. Core Scientific has mined about 40,000 Bitcoin for itself, and its facilities have mined about another 40,000 for hosting clients, on about 800,000 ASICs since 2017. He said the company has been through two halvings at scale, and he has been through three himself. He also said most of what followed was his own opinion, not a statement from a publicly traded company.
So Core, as a company, has gone through two halvings at scale.
Russell Cann · 1:03
The halving and the silicon chart
His first point was about hardware. In past cycles, he said, chips moved down the joules-per-terahash chart at the same time as each halving, so miners came to believe that every halving meant scrapping the fleet for the newest machines. Cann called that a coincidence. ASIC process sizes have now caught up with the rest of the chip industry, and the gap between 7, 5, 4 and 3 nanometer machines is small.
So everyone thought that at the end of every halving, you had to trash all your miners and replace them with the newest machines.
Russell Cann · 1:44
The result, in his words, is an elongated refresh cycle, with machine life that can be almost in perpetuity. He pointed to the last halving, when Core Scientific, Riot, Marathon and others all replaced their fleets at once and drove up the dollar-per-terahash price of every ASIC. This cycle, he said, that has not happened, because the silicon is so close in efficiency that even Core's 34-joule machines are still profitable on many days. When they are not, they are turned off.
He walked through the numbers: 34 J/TH machines drop to 29 in low power mode and to 26 in super low power mode with firmware; 21-joule machines can do the same; the newest S21 Pros are at 15; the newest machines coming out are at 12. If hashprice is 6 cents, he said, the gap between 12 and 21 joules is almost de minimis. If hashprice is 2 cents, it decides whether a company turns everything off or keeps mining.
Why the miners watch hashprice, not BTC
After the April 2024 halving cut the block reward in half, a larger share of miner revenue comes from transaction fees. Cann said only one person at Core Scientific cares about the price of Bitcoin: the company's trader, Danny, who was in the room. Software, hardware and operations all watch hashprice, because it takes in both network difficulty and the BTC price. He credited Luxor's public hashprice chart for anyone who does not want to do the math.
There's only one person at Core that gives a shit about the price of BTC.
Russell Cann · 3:22
Hashprice, he said, moves like a sine wave. On the day of the talk it was just below 5 cents. When a large ordinal or other layer 2 activity hits the chain, fees jump and hashprice shoots up. He expects more development on Bitcoin, at layer 2 and layer 3, to add to both price and hashprice volatility.
That changes how a site should run. In earlier cycles, he said, the goal was to plug a machine in, keep it on and replace it in 36 months. Now operators need to think about the longevity of the machine and about turning it on and off based on hashprice and on their price of power.
Power: curtailment and AI demand
The fourth factor in his list was the volatility of power. In the United States, he said, AI buyers are now setting that price, and they can tolerate far higher power costs than miners. Core's all-in power price in the U.S. was about 4.25 to 4.5 cents, which he called high for mining; he wants to be closer to 2 cents, with the option to turn it off when hashprice drops. For AI, the same price is cheap, and he expects AI buyers to offer utilities 6 or 7 cents.
At Core right now, our price in the United States is like 4.5 cents, 4.25 all in for power.
Russell Cann · 7:02
He argued that smaller miners have more options. Core does not have a site under 100 megawatts, but a 5, 10 or 15 megawatt miner that is willing to curtail is an easy customer for a grid. Core curtails at every one of its sites, he said, whether for weather, economics or the maintenance of a nuclear plant in TVA's territory. At 100 to 300 megawatts, the relationship with the utility becomes "hand in glove" rather than that of a client.
He asked new miners to plan for 2028 and 2032, the next two halvings, when he expects the block reward to be so small that transaction fees will have to pay for everything. He also noted a risk: a large amount of hash sitting idle on the sidelines can become an attack vector for Bitcoin.
Q&A: the cost of moving to AI
Asked whether AI data centers will take over Bitcoin mining sites, Cann said Core had just announced 500 megawatts of HPC and had set aside 700 of its 1,200 megawatts of power contracts for AI. He framed it as a move to the highest and best use, not an exit from mining: much of Core's power load was put online before Russia invaded Ukraine, leaving it overexposed to natural gas and with some sites priced higher than he wants for mining. Those sites can go to AI, and the company can backfill with new, more efficient mining sites.
He said the barriers to entry are huge. Core builds mining sites for about half a million dollars a megawatt. Converting one for AI runs roughly eight and a half to nine million per megawatt, while a greenfield multi-tier AI site in the U.S. costs 13 or 14 million. Filling it with H100s adds another $25 million a megawatt.
The conversions are around eight and a half to nine million dollars a megawatt.
Russell Cann · 13:28
He does not expect the move to AI to be common among miners, given the skills and capital it takes, and thinks data center companies are more likely to look at mining sites with good power contracts. Mining and AI on the same site was "not just no, but hell no": a PPA for 99.9% uptime would make the power too expensive to mine. He gave the example of a curtailable contract near Atlanta, where the utility needs the load back for some days in July and August; asking for that load full-time becomes a completely different conversation.
Hedging power in and production out
On hashprice and finance instruments, Cann said miners should hedge both ways: their power cost going in and their production coming out, as oil companies did in the eighties. A hashprice hedge or a Bitcoin hedge both work, he said, because a miner knows what its production will be. He dismissed mining at a loss and warned that counterparties need checking.
Mining at a loss is just stupid.
Russell Cann · 16:53
At the time of the conference, Bitcoin traded at $61,272. On October 17, 2026, it was $81,466. Mining Disrupt 2027 runs March 22 to 24 in Irving, Texas.