Bitcoin Miners Dumping BTC? Mara CEO Fred Thiel On AI Pivot, Next Price Breakout

Here is a chronological list of distinct topics, claims, and statements from the transcript:

1. Quantitative tightening is officially over, and quantitative easing will begin, with money expected to be injected into the economy, particularly in the repo market.
2. Power is the biggest constraint in the AI business, and converting Bitcoin mining sites offers the fastest way to secure power for AI operations. This shift will lead AI data centers to become curtailable loads, increasing competition for power resources with Bitcoin miners.
3. November was one of Bitcoin's weakest months since 2021, prompting questions about whether miners are selling Bitcoin to transition into AI, potentially contributing to the market sell-off.
4. The market was perceived as "frothy" in Q2, driven by institutional interest, ETF inflows, and expectations of Fed rate cuts. However, a series of events in November changed this outlook.
5. Multiple factors contributed to a "risk off" environment in November: a decreased likelihood of Fed rate cuts, concerns about debt in the AI circular economy, and Japanese rate hikes leading to the unwinding of the Japan carry trade.
6. Older Bitcoin "whales," holding Bitcoins in wallets 10-13 years old or more, began selling their holdings or transferring them to ETFs, which market observers interpret as a sale. These older wallets are susceptible to quantum attack as their public keys are known and lack modern protections like Taproot.
7. The combination of these factors, including Jim Chanos' shorting strategies and unwinding of highly leveraged positions, compounded to create a market correction. Bitcoin currently operates much like equity in portfolios during risk-off environments.
8. A large whale successfully took a significant short position in October's perpetual markets, profiting from the market dip. This likely encouraged others to follow a similar trading strategy.
9. The digital asset trust and treasury company model is currently being questioned, with companies like MicroStrategy trading below their net asset value and others like Nakamoto and MetaPlanet seeing significant drops. American Bitcoin's share price also fell 50% after its lockups expired.
10. These combined events generated substantial fear, resulting in withdrawals from ETFs and ETPs. The market has since found stable footing, with Fed rate cuts largely priced in.
11. Looking ahead, quantitative tightening is expected to conclude, making way for quantitative easing and significant money injection into the economy. Given the upcoming midterm election, substantial market stimulus is anticipated.
12. A declining dollar and dramatic growth are viewed as beneficial for addressing the deficit and debt, though these actions pose challenges in managing inflation. Unemployment rates are slightly rising but remain relatively low.
13. A period of uncertainty, exacerbated by a government shutdown and a lack of data, has largely passed. The market is now on more stable ground, and tax-related selling is believed to be complete, leading into a "Christmas lull" before renewed activity in January.
14. An expert from VanEck noted that some miners are selling Bitcoin to fund pivots to AI, which is an expensive upgrade cycle. Tightening credit conditions increase the cost of debt needed for facility repurposing, and if Bitcoin prices drop simultaneously, miners need to raise more debt at higher costs.
15. This expert warned that when AI and Bitcoin become correlated, it can be doubly detrimental to a company's stock, as seen in the previous month.
16. If miners or investors borrow against Bitcoin to invest in AI, and Bitcoin's price profiles turn negative, it creates significant issues. Bitcoin miners should prioritize Bitcoin mining and strategically leverage their existing assets, like sites and power, to convert them into higher-value AI assets.
17. Miners can transition to AI in multiple ways: leasing their sites and power to HPC developers, building facilities with GPUs for rent, or engaging in greenfield land and power deals with developers. A hybrid approach, mining Bitcoin and operating AI at the same large site, is also possible.
18. Marathon is pursuing an inference AI strategy, running it in a containerized, modular, air-cooled format, similar to its Bitcoin mining operations, to avoid building large facilities or requiring water cooling. The company believes training AI models is best left to hyperscalers.
19. Inference AI necessitates private cloud solutions to protect the proprietary data used in models. Marathon's investment in Exion, a French company developed within EDF, provides advanced technologies for managing private cloud data sovereignty and protection, which is essential for offering inference services.
20. Marathon Digital Holdings is recognized as the world's second-largest public Bitcoin treasury company, with its BTC holdings steadily increasing over the past year from approximately 44,000 to over 53,000 BTC.
21. While Marathon has started selling some of its production Bitcoin, it continues to net increase its overall Bitcoin balances monthly.
22. The claim that "70% of top miners are pivoting to a $20 billion AI market" by selling Bitcoin to fund the transition is not entirely accurate for Marathon. Most miners pursuing AI (e.g., Galaxy, Riot) are repurposing *undeployed* Bitcoin mining capacity for HPC AI, rather than dismantling existing sites.
23. Converting an existing Bitcoin mining site to HPC involves deconstruction and reconstruction, which is bypassed by Marathon's modular, containerized approach. Many peers use convertible bonds for AI capital, and there's no widespread decrease in treasury Bitcoin holdings among major miners.
24. Global Bitcoin hash rate has continued to grow, with China re-emerging as the third-largest mining nation, and significant growth also seen in Pakistan and Kazakhstan. Tether's ambition to become the world's largest Bitcoin miner suggests continued viability in the business.
25. The statement "Bitcoin mining profitability plunged to record lows in late 2025" is a future prediction that lacks current basis. Mining profitability depends on Bitcoin block rewards and a miner's proportion of the global hash rate, which has continuously grown, making it harder to win Bitcoin without adding capacity.
26. The industry average cost to mine Bitcoin is around $60,000. While profitability is strong above $100K, it becomes challenging for miners with expensive energy or hosted models if Bitcoin falls to $80-90K. Owned and operated sites, like Marathon's, have significantly lower energy costs.
27. Marathon's current diversification strategy is primarily focused on AI and private cloud compute, rather than solely on altcoins. While some altcoins are mined using excess power for marginal profitability, they represent an immaterial portion of revenue.
28. Marathon's mining updates for late 2025 include producing 218 blocks (a 5% increase) and a 1% month-over-month increase in energized hash rate. Its Texas wind farm site has fully deployed containers and miners, on track for full operational status in Q4.
29. Operating behind the meter at Marathon's owned Texas wind farm results in near-zero marginal energy costs, about one cent per kilowatt-hour for maintenance. This enables the "AARP" program, where fully depreciated miners are moved to such sites to generate profit without depreciation expense, even if they run intermittently.
30. Marathon also employs this near-zero-cost energy strategy in flare gas and oil fields in Texas and North Dakota, where data centers generate energy from flare gas.
31. Marathon is finalizing a partnership with MPLX to build three sites attached to their gas pipeline, initially generating 400 megawatts and scaling to 1.5 gigawatts, along with three data center campuses. This large-scale, behind-the-meter energy will be optimized for a mix of Bitcoin and AI operations, with power expected to come online by 2027.
32. The investment thesis for institutional investors in Bitcoin miners has shifted post-ETF launch. While miners previously served as Bitcoin proxies, now Marathon's market cap is primarily valued by its Bitcoin holdings, with little attribution to mining operations.
33. Savvy investors now view Marathon's extensive power assets and strategic locations as highly valuable for potential conversion to AI/HPC. The fastest way to secure power for AI is to convert Bitcoin mining sites, making Marathon an attractive investment if it leverages these assets for AI.
34. Marathon aims to maximize value from its portfolio of assets through Bitcoin mining, hybrid Bitcoin/inference AI, or conversion to HPC. The company will not "outcompete" Bitcoin itself.
35. Marathon adds shareholder value through a strong balance sheet and by leveraging land and power assets, which are crucial resources for both the AI industry and Bitcoin mining. It also fosters relationships with energy majors to convert energy into valuable outputs, whether Bitcoin or AI insights.
36. Marathon's key milestones for the next quarters include closing the Exion investment and finalizing the MPLX deal in Q1. Further news is expected regarding progress at its regional headquarters in Saudi Arabia and EMEA headquarters in France, indicating continued domestic and international growth.
37. The relationship between hash rate and Bitcoin price is direct: higher profitability incentivizes more hash rate, while declining profitability leads to older, less efficient machines shutting down. Profitable mining requires either newer, efficient machines or very low-cost power.
38. Grid-attached miners paying 4-5 cents per kilowatt-hour, coupled with operational and depreciation expenses, face financial strain when Bitcoin prices are low. Owned and operated sites maintain significantly lower costs per kilowatt-hour.
39. Global hash rate and Bitcoin price will eventually reach a "stasis" where Bitcoin mining maintains about a 30% margin. If margins compress further, hash rate declines, but then recovers as remaining miners earn more.
40. A long-term prediction states that by 2028-2032, Bitcoin miners must evolve into energy companies, form joint ventures with energy companies, or own their own generating assets, as grid-attached mining will no longer be profitable. The industry is currently undergoing this transformation.
41. Bitcoin mining is expected to shift towards locations with lower-cost power, including utilizing excess power at hyperscaler sites not fully used by the hyperscalers themselves. Technologies like high-speed switching batteries can support this.
42. Marathon's deployment of inference AI at its sites is part of a strategy to maximize the value of every electron, providing the optionality to switch between Bitcoin mining and AI insights for higher profit. The AI business, while capital-intensive (10x Bitcoin mining capex per megawatt), also offers potential for 10x returns.
43. Regarding jurisdiction, while there's a political desire to make America the crypto capital, global hash rate has grown faster outside the US. US miners face competition from AI sites for existing power, leading to a 2-3 year lag in site development.
44. The future growth of US hash rate will depend on regulatory permitting, energy prices, and the willingness of Bitcoin miners to allocate power to mining versus other uses.
45. In response to BC's permanent ban on new crypto mining connections to the grid due to electricity supply concerns, it's noted that BC has abundant, non-intermittent hydropower. The core issue is the province's allocation of surplus power to create jobs, prioritizing industries like manufacturing over data centers, which create jobs primarily during construction.
46. In regions with intermittent power (wind, solar) and limited transmission, Bitcoin miners can serve as effective load balancers, operating flexibly for fewer hours a day. European politicians, for example, are open to discussions about load balancing with flexible miners.
47. Bitcoin mining can make AI a flexible load. The US, for instance, uses 100% of its generating capacity only 2% of the day, averaging 60% use overall. This indicates a large amount of available flexible energy.
48. A Duke University study suggests the US has 70 gigawatts of available power for the AI industry's 40 gigawatt demand, provided AI data centers are willing to be flexible loads and curtail less than 2% of the time.
49. Currently, AI developers often demand redundant power (grid plus backup), which can complicate grid management. An incident in PJM showed data centers sensing a frequency drop and switching to backup, creating excess energy that flooded the grid and caused problems.
50. Utilities are likely to begin controlling when AI data centers connect and disconnect from the grid to manage load effectively.
51. Regarding concerns about crypto mining draining power from local grids in North America, isolated incidents occurred early in the industry's growth where some miners attached to community grids failed to curtail operations when needed. This forced communities to purchase more expensive power, raising costs for residents and creating local irritation.