Episode Summary
Executive Summary: The discussion centered on how AI and high-performance computing are fundamentally re-rating Bitcoin miners. Investors now focus less on BTC beta and more on miners’ power pipelines, contracts, and ability to convert mining sites into data centers. The guests agreed Bitcoin mining remains viable, but future value increasingly depends on HPC/AI buildouts, financing, execution risk, and power economics.
Main Topics: AI/HPC as the new valuation driver for miners (Priority: 5/5): Both guests said investor attention has shifted sharply from Bitcoin exposure to AI and high-performance computing opportunities tied to miners’ power assets and data-center capacity. NeoClouds, contracts, and data-center demand (Priority: 5/5): They explained how CoreWeave, Oracle, and other NeoCloud/HPC players are driving demand for power and capacity, with remaining performance obligations and leases validating the need for new infrastructure. Economics of Bitcoin mining vs. HPC (Priority: 5/5): The panel contrasted declining mining profitability and rising hash rate with the more attractive economics, financing visibility, and longer-duration contracts in powered-shell HPC deals. Execution risk and due diligence (Priority: 4/5): They warned that not all megawatts are equal: investors should scrutinize real power pipelines, supply-chain readiness, contract terms, penalties, and construction timelines. Debt, converts, and capital structure (Priority: 4/5): They discussed how miners are increasingly tapping convertible bonds and other debt markets to fund HPC buildouts, while Bitcoin-mining balance sheets and BTC holdings still matter. U.S. power constraints, policy, and geography (Priority: 4/5): The conversation explored how constrained U.S. power markets may push available megawatts toward AI/HPC, while Bitcoin mining can remain globally mobile and continue in remote or stranded-power locations. Future of proof-of-work and industry coexistence (Priority: 3/5): They debated whether Bitcoin mining could coexist with HPC in the same facilities, with some companies potentially using flexible load management or specialized rack designs to support both.
Key Arguments: AI/HPC has re-rated Bitcoin miners because investors now value power, land, and grid access more than pure BTC exposure. CoreWeave's IPO, additional NeoCloud leases, and Oracle’s large RPO backlog signaled that demand for data-center capacity is real and growing. Not all megawatts are created equal: investors must verify actual power availability, permitting, supply-chain readiness, and whether capacity is truly contractable. The powered-shell co-location model is generally viewed as lower risk than GPU-as-a-service because tenants often provide the interior infrastructure and software stack. Bitcoin mining is still viable where power is cheap or stranded, but in developed markets the same megawatts may earn higher returns in HPC/AI. Miners with strong BTC balance sheets and power pipelines have optionality, but the market is increasingly rewarding those with credible HPC conversion plans. Execution risk is now central: delays, missed milestones, and contractual penalties could erode the economics of HPC deals. Debt issuance, especially converts and project financing, is increasingly tied to HPC buildouts rather than Bitcoin mining expansion. U.S. Bitcoin mining will likely persist, but more marginal megawatts may shift toward AI/HPC as that is the economically preferred use of scarce power. The guests see room for hybrid facilities, but believe true cohabitation of HPC and mining requires more engineering and is not yet widespread.
Data Points: Bitcoin price change (3 months): down about 12% - Referenced in comparison with miner equity performance over the prior three months. Bitcoin miner index change (3 months): up about 120% - CoinShares-style blended miner index used to show AI/HPC-driven outperformance. U.S. data center market size: about 25 gigawatts - Used by John to contextualize how large 1-3 GW demand is. Typical data center building size: 150-250 megawatts - John said most facilities are built in multiple buildings rather than a single full-gigawatt site. Oracle capacity need: 4.5 gigawatts total; 2.6 gigawatts still to procure - Used as evidence of large, real demand for capacity. Build cost per megawatt: about $10 million per MW - Kevin referenced this as a rough rule of thumb for data-center build costs. Build cost per gigawatt: about $100 billion per GW - Derived from the $10 million/MW estimate. AI build cost estimate mentioned: 10 GW could require around $1 trillion - Referenced in connection with Sam Altman’s reported ambitions. Hash price trend: at or around all-time lows - Used to show declining Bitcoin mining profitability. Miner cost to produce BTC: roughly $50 to $80+ per bitcoin - Kevin and John gave ranges depending on power cost, depreciation, and company. Electricity costs mentioned: about 2.7 cents to 5+ cents per kWh - Compared contracts like Cipher’s cheap power with higher-cost mining sites. Cycle comparison: late-1990s buildout analogy - Kevin compared today’s AI infrastructure boom to the dot-com era. Duration of some AI contracts: 2 years to 5 years - John contrasted shorter AI lab contracts with higher-quality longer contracts from major customers. Long-duration co-location agreements: about 15 years - John noted Bitcoin mining co-location agreements can be much longer than AI contracts. Missed-timeline tolerance: 2 quarters - John said some contracts may become cancelable if sites miss milestones by that much.
Pivotal Quotes: "I think there's probably always going to be a place for Bitcoin mining. I don't think we have a clue about what artificial intelligence can bring to us." — Steve Ehrlich: Opening framing of the episode on the long-term coexistence of Bitcoin mining and AI. "For the most part, all the conversations we're having is around HPC and AI." — John Tedaro: Described the shift in investor focus from Bitcoin mining to AI/HPC exposure. "I don't think we have a clue about what artificial intelligence can bring to us. If you look out the next 30 years, life's going to be not incrementally different, but diametric." — John Tedaro: Used to argue that AI infrastructure demand could be transformational and long-lived.
Implications: Bitcoin miners are increasingly valued as scarce power-and-infrastructure platforms rather than pure crypto proxies. Expect more capital, debt, and strategic competition around HPC/AI conversions, while Bitcoin mining survives in cheaper, remote, or flexible-power niches.