Unlock stock picks and a broker-level newsfeed that powers Wall Street.
Tom Carreras
7 min read
Frank Holmes has had a long career as a money manager: financing gold mining companies; getting involved in the creation of gold royalty companies; developing financial products for the airline industry — all of this with U.S. Global Investors (GROW), the publicly-traded asset management firm he’s been leading since 1989.
He’s also the chairman of HIVE Digital Technologies (HIVE), a bitcoin mining company with a $345 million market capitalization and a rapidly expanding footprint in Paraguay, thanks to a recent deal in which the firm acquired facilities previously owned by another miner, Bitfarms. The firm was born, he said, after he attempted to launch a spot bitcoin exchange-traded fund (ETF) in 2017.
HIVE has been green from the get-go. Its first facility used geothermal energy in Iceland; another used hydro-power in Sweden, only 100 kilometers south of the Arctic Circle. Now, the company expects to have roughly 430 megawatts (MW) of infrastructure up and running by the third quarter of 2025 — meaning enough energy to power a city of 86,000 homes.
Holmes will be speaking at the BTC & Mining Summit at Consensus 2025, in Toronto on May 14-15.
In the lead-up to the event, Holmes shared his thoughts about HIVE’s place within the broader mining industry, the company’s decision to recycle its GPUs for AI purposes, and what the future holds.
This interview has been condensed and edited for clarity.
CoinDesk: HIVE has been repurposing some of its GPUs for AI. Can you tell me about that?
Frank Holmes: At one time we had 130,000 AMD chips and we were mining ether (ETH). We were about 6% of the world's ether mining and it was very profitable. When that went away [with Ethereum’s transition to Proof-of-Stake in 2022], we had this expertise in GPU chips and we replaced a lot of our AMD chips with Nvidia chips. That allowed us to start going down the AI path.
The difference between a basic ASIC miner and Nvidia chips is like driving a Bronco and a Ferrari. The delicacy of the motor, the engines, all the gearing that goes into a high performance car — all of that relates much more to a GPU. When Antminers S21 Pros show up, it takes us six hours to unwrap them and plug them in. When the Nvidia chips show up, like an H100, it’s six weeks before you’ve built the brain and it’s working. So it's a completely different skill set.
When you're building infrastructure for bitcoin mining, you're spending a million dollars per megawatt of electricity. When you go into high performance computing (HPC), the facilities need so much redundancy that you spend $10 million per megawatt. That's excluding the equipment. You've got much higher logistical engineering requirements, and you've got much higher capital expenses.
Terms and Privacy Policy Your Privacy Choices More Info