Frank Holmes, initially immersed in the gold sector, began exploring the potential of Bitcoin in 2017. Facing regulatory hurdles in launching a Bitcoin ETF, specifically related to anti-money laundering and customer identification protocols, he established HIVE, a cryptocurrency mining operation that now boasts a market capitalization of $627 million.
Following his first experience with a “Bitcoin halving” – a cyclical event occurring roughly every four years that diminishes the creation rate of new Bitcoins by half – Holmes strategically invested in data processing centers. This proactive measure was designed to buffer the financial impact of halvings on miner profitability and ultimately paved the way for HIVE’s entry into the artificial intelligence (AI) arena.
The chairman of the board at HIVE Digital Technologies recently discussed with Sherwood News the evolving landscape of the Bitcoin mining industry. The conversation also covered the synergy between Bitcoin and high-performance computing centers for AI, and the financial implications of utilizing Nvidia processing units.
The ensuing exchange has been edited for the sake of brevity and clarity.
Sherwood News: What is your assessment of the current health of the mining sector?
Frank Holmes: I am observing a trend where companies are issuing convertible debt to acquire Bitcoin without proportionally expanding their operations or upgrading facilities with more efficient technologies. This implies a speculative bet against the industry’s underlying dynamics. It’s vital to have a decentralized network with a diverse set of nodes globally. Since I started in 2017, the validation network has significantly expanded, from about 11,000 to 21,000 nodes worldwide.
We’re also witnessing activist pressures on Bitcoin miners in Texas, urging them to halt expansion and transition to AI businesses. This signals a shift in focus. HIVE anticipated this trend and has been making this transition in a careful and profitable manner. The Bitcoin mining industry will likely undergo consolidation between now and the next halving. Many businesses have failed to invest sufficiently in energy-efficient chips or secure access to low-cost energy sources.
If you lack inexpensive energy, you must continually reinvest in more energy-efficient ASIC processors, and this level of investment is not universally evident. Consequently, many energy-rich companies are forging partnerships with entities like CoreWeave to enter the AI domain. This exodus will ultimately benefit the remaining miners by improving industry economics. We are currently navigating this rebalancing phase.
The key takeaway is that Bitcoin miners laid the groundwork for the AI boom. They identified and harnessed underutilized, wasted, and surplus energy. Bitcoin miners pioneered methods to extract value from these previously untapped resources, which is significant. We served as the foundation for the current explosive growth in artificial intelligence.
Sherwood: How would you characterize the current distribution of HIVE’s business activities between Bitcoin and artificial intelligence?
Holmes: At the close of last year, it was approximately 90% Bitcoin and 10% AI.
Sherwood: Do you foresee that percentage shifting significantly in the near future?
Holmes: Yes, I do. We anticipate generating $500 million from Bitcoin mining and $100 million from high-performance computing (HPC), bringing the AI portion to around 20%.
Sherwood: Could you explain more directly how Bitcoin mining operations are preparing the pathway for the massive expansion of artificial intelligence?
Holmes: Bitcoin mining fundamentally involves data centers and high-performance computing. AI also relies on data centers, but they differ significantly based on tiers. Tier 4 is reserved for military or governmental use. Tier 1 costs about a million dollars per megawatt.
Upgrading from Tier 1 to Tier 3 can cost an additional $10 million due to the need for extensive air conditioning and backup systems. The financial investment required for high-performance computing (HPC) for AI is considerably greater. Furthermore, the processing chips are substantially more expensive, around $35,000 each compared to approximately $3,000 for an ASIC chip.
It takes about six hours to unpack and configure an ASIC chip. However, setting up an Nvidia chip takes six weeks because of the intricate system building required to make it function – six weeks spent connecting the CPUs and GPUs. The complexity involved is significantly higher.
Once you have an operating data center in place with the infrastructure investment already paid off, it becomes quicker and more economical to upgrade to Tier 3. Starting from scratch with Tier 3 could take up to three years to obtain all necessary permits in North America.
Having existing power infrastructure and a power station allows for rapid scaling and facility upgrades within nine months instead of three years. Bitcoin miners have effectively paved the way, enabling the rise of high-performance computing for AI.
Sherwood: What is the direct difference in output between using Nvidia chips versus ASIC chips?
Holmes: Looking at hourly economics, an ASIC Bitcoin mining machine generates about $0.15 per hour with costs of $0.05, resulting in a $0.10 gross profit. An Nvidia chip, conversely, generates roughly $2 per hour. Despite the higher initial expense, the revenue generated on an hourly basis is far superior.
Sherwood: What is HIVE’s shareholder composition?
Holmes: Institutional investors hold about 25%, with the remaining 75% held by retail investors. Within that retail segment, a significant portion consists of family offices. We have a substantial number of large-scale investors, particularly from Europe and Canada. Initially, our institutional funding came from Fidelity mutual funds and others, but retail investors provided the liquidity and overall support.
I have a significant following within the gold investment community. Many of my gold investors were hesitant to enter the Bitcoin ecosystem due to the failures of FTX and Celsius, which were extremely damaging. HIVE emerged as a suitable proxy for these gold investors to gain exposure.
Sherwood: Does HIVE sell newly mined Bitcoin to cover expenses, or does the company prefer to hold (HODL) and borrow funds?
Holmes: Unlike other companies, we have avoided borrowing at high interest rates like 9% and 14%. We executed a unique transaction with Bitmain where we pledged our Bitcoin against machines without incurring interest costs. This saved shareholders approximately $20 million compared to our peers.
We’ve used at-the-market offerings (ATM), but we deploy them with a cash-flow return on invested capital model. We ensure that machine purchases immediately generate cash flow.
Sherwood: Are there any significant challenges or advantages that you see on the horizon?
Holmes: The primary difficulty revolves around the logistics of procuring and installing equipment.