Luke Rossy opened his Mining Disrupt Miami keynote by calling it "a bit of a technical talk." The chief operating officer of Hive Digital Technologies, speaking at the June 24–26, 2024 event, spent most of it on one question every miner faces: what equipment to buy, and when. His answer ran through a single metric, hashprice, and ended at a number Hive had already committed to: a fleet efficiency of 25 joules per terahash.
Rossy introduced Hive as the first publicly traded crypto mining company. It started mining Ethereum in Iceland in 2017, expanded to Sweden and Canada, and began mining Bitcoin in 2018. It runs on geothermal power in Iceland and hydro in Sweden and Canada. After the Ethereum merge, Hive moved its GPUs into AI compute.
What the Merge Taught Hive
Rossy walked through Hive's recent production. The company mined nearly 2,900 Bitcoin in the 12 months before the talk. Over fiscal 2023, he said, it took twice as much hashrate to mine one Bitcoin at the end of the year as at the start; Hive went from about 2 exahash in January 2023 to 4 exahash by year end, yet production rose only slightly. In 2024, difficulty climbed about 20% to 88 trillion right after the halving, and Hive's growth past 5 exahash again did not show up fully in output.
The sharpest drop in Hive's results came after September 2022, when Ethereum moved to proof of stake. Rossy said Hive was generating $150,000 a day mining Ethereum at the time. The company prepared in two ways: it pointed its GPUs at the most profitable altcoins with no additional capex, and it started a beta project using its best GPUs for AI compute. Through the Luna collapse, the Celsius and FTX bankruptcies and Bitcoin's fall to $16,000, he said, Hive stayed profitable every quarter, across the 2020 halving, the merge and the 2024 halving.
It's really important that when you have a miner, not only do you ROI, but you free cash flow as long as possible so you could fund your future upgrades.
Luke Rossy · 4:47
Hashprice, Not Price or Difficulty
Rossy defined hashprice as "how much the network is going to pay you for a specific amount of hash on a given day." On his chart in US dollars, it hit about $0.06 per terahash per day after the 2020 halving and about $0.05 after the FTX crash in 2022. In the 2021 bull run it sat between $0.30 and $0.40. Since 2020, he said, the floor of about $0.05 to $0.06 had been tested time and again, and after the 2024 halving it broke below it.
He explained why the floor does not simply hold. When hashprice falls that low, miners with older equipment unplug and difficulty readjusts every 2,016 blocks, about every two weeks. But miners also sign long-term power contracts, buy equipment months or years ahead and keep plugging it in, and ASIC makers keep releasing more efficient machines that can mine at lower hashprices. "So over time, these floors are going to keep going down," he said.
Price alone, he argued, is misleading. In November 2021, with Bitcoin at $65,000, it took one-eighth of the hashrate needed at the time of his talk to mine one Bitcoin. Measured in Bitcoin, Hive's chart showed about four to four and a half Bitcoin per exahash in 2022, about 0.8 the week before the conference and about 0.6 on the day. Three spikes on that chart came from transaction fee rallies driven by ordinals and runes.
So clearly, you can't just look at Bitcoin price or just difficulty and make decisions.
Luke Rossy · 9:30
Reading the Profitability Matrix
Rossy showed a matrix that turns hashprice and ASIC efficiency into revenue per megawatt hour. With the inputs from the week before the talk and power at $45 per megawatt hour, an M30 was making roughly $10 per megawatt hour, "just squeaking out a profit"; a J Pro a little over $20; an S21 over $60. A second matrix gave break-even efficiency by power cost, and at $0.02 power, he noted, an S9 could still run. Using the inputs from November, six or seven months earlier, the same M30 made $30 per megawatt hour and the J Pro $50. His point: a single snapshot is never enough, and purchase decisions need past patterns and future scenarios.
He also pointed to the hardware coming to market: the S21 XP, announced at 13.5 joules per terahash, or 12 on hydro or immersion, due at the end of that year or early the next, and S21 Pros being installed as he spoke. The ETF rally earlier in 2024 took Bitcoin to an all-time high of $73,000, yet hashprice reached only 10 to 13 cents, against three to four times that in 2021.
My personal prediction is that from now to the next four years before the 2028 halving, is we're going to see a new floor at three cents a terahash.
Luke Rossy · 14:03
Why Hive Picked 25 J/TH
Hive set its fleet target more than a year before the conference, Rossy said, and had reached it. At 25 joules per terahash, Hive breaks even at a hashprice of 2.7 cents per terahash per day, which means a 3-cent floor would still leave the whole fleet mining profitably. Hive bought more than 30,000 ASICs over that period, including S21 Pros, S21s, XPs, K Pros and Pro Pluses, and had announced another 1,000 S21 Pros a couple of weeks earlier. It operated a bit over 5 exahash on 140 megawatts in three countries, and could draw down to 100 megawatts or downclock miners to improve efficiency further.
Buying the newest model is not always the answer. Sometimes, he said, it is smarter to buy a generation or two behind, because what matters most is how fast a machine pays for itself: ideally four to six months, so it can free cash flow to buy the next round of equipment.
Competing With AI for Power
Rossy closed on two pressures. The first is power. Even an S21 brings in maybe $100 to $130 of revenue per megawatt hour, the latest machines perhaps $160, and M30s about $60, while next-generation AI data centers are "in the thousands of dollars a megawatt hour." Bitcoin miners, he said, will be pushed to more remote sites with less reliable power, and the mega sites of the past "are not happening anymore." That is why large miners are looking at new revenue lines; for Hive, with its Ethereum-era experience running GPU fleets and 38,000 Nvidia GPUs bought in 2021, AI was the natural move.
We can't compete with them for the same power resources.
Luke Rossy · 16:41
The second is fees. Realistically after the 2028 and 2032 halvings, he expects most miner revenue to come from transaction fees. Today miners run most of the time and draw down when profitability falls; he imagines that flipping, with fleets switched on for short windows when a fee rally lifts hashprice above a floor that could sit at two cents.
When this session was recorded, Bitcoin traded at $61,272; on October 16, 2026 it was $82,531.