Back AI Macro Nexus Research

Dark Factories, Bright Future: AI’s Physical Bottleneck and the Rare-Earth Standoff

Published on October 19, 2025

Jordi Visser

In this week’s video, I break down how the rare earth battle escalated with public statements and news on China’s dark factories, credit fears morphed into bank concerns as earnings began, and why the singularity is accelerating faster than bubble skeptics realize. The US-China trade situation remains fluid, with both sides signaling negotiation willingness. There was a 27% drop in China exports to the U.S., while China maintains 8.3% overall export growth by offsetting through other markets. Trump and Xi’s summit remains on track, and I do believe in the end, a deal is coming because it is hard to imagine China sustaining a rare earth stranglehold on the entire globe when in three years, the world can balance it out. China would risk broader economic damage.

The investment implication: AI’s binding constraint has shifted from GPUs to physical infrastructure, power, transformers, gas turbines, and grid interconnects. Western executives touring China’s dark factories are returning “terrified” by Beijing’s robotics dominance, with fully autonomous facilities operating without lights or human operators. Tesla’s shareholder meeting on November 6th may showcase Optimus 3, potentially creating an iPhone-moment for humanoids that reshapes the investment landscape as profoundly as the smartphone created the Mag 7. Meanwhile, AI infrastructure companies continue to blow out earnings, with Applied Digital revenue up 84% year-over-year and the semiconductor sector posting its biggest weekly rally since June, up 5.4%.

The AI research reality is stark: OpenAI, Google DeepMind, and others have models far more advanced than public releases, constrained only by compute capacity. Dave Blundin from the Moonshots podcast projects AGI arrival by 2026-27 based on recursive self-improvement already happening behind closed doors. Sam Altman’s “capacity overhang” means breakthroughs exist but cannot be deployed due to insufficient GPU, power, and cooling infrastructure. This creates massive alpha opportunities in utilities, grid equipment, energy infrastructure, and even small caps as the AI build-out diffuses through the supply chain. Bitcoin miners emerge as “virtual batteries,” stabilizing renewable grids while monetizing off-peak power.

For traditional investors anchored in large-cap growth, this AI-power-capex cycle demands portfolio reorientation toward energy, infrastructure, industrials, semiconductors, and autonomous/humanoid plays. The irony: institutions must now buy assets they spent years avoiding, energy, Bitcoin, Tesla, small caps precisely when retail is already positioned there. With PMIs set to rise after a China-US deal, Fed cuts coming, strong earnings growth, and the memory super-cycle extending through 2027 driven by inference workloads combined with continued AI bubble fears, the setup favors risk assets. The singularity isn’t speculation; it’s the investment regime that’s already rewriting alpha sources in real time.

Timestamps

(00:00–02:06) Rare earth battle continues with back-and-forth statements; Trump caring about markets, walking line between strength and stability; tariffs will fast-track if China escalates; Deng Xiaoping’s 1991 quote: “The Middle East has oil, China has rare earths”
(02:28–05:29) IMF meetings with Bessent and Greer signaling course correction; China exports up 8.3% overall but down 27% to US, massive gap; trade deal still likely because China needs it within three years unless planning war
(05:48–07:36) Greer and Bessent had frank discussions Friday night; Trump-Xi summit still on track; Trump acknowledged tariffs unsustainable; 80% chance on Polymarket; economy has massive upside if deal happens
(08:00–09:28) Dark factories in China, Western executives “terrified”; fully autonomous facilities with no humans; Andreessen Horowitz warns America cannot lose robotics race; Elon Musk shareholder meeting November 6th may reveal Optimus 3; China’s self-driving “DeepSeek moment” as Tesla won 36-car autonomous test
(10:16–12:50) S&P up 1.7%, Nasdaq up 2.5%, IWM up 2.4%, best week since August; semiconductors up 5.4%, biggest rally since June; VAR shocks and episodic volatility expected; regional banks down on credit fears; liquidity issues emerging with SOFR-Fed funds spread and standing repo facility usage
(13:32–15:44) Crowded shorts rallying; PMIs likely heading to 60 next year; strong earnings season so far with sales and earnings surprises across financials and tech; capex spend for next year will be massive
(16:14–18:13) Hyperscalers (spenders) unchanged since August while semis up 41%; divergence finally happening; modern price-to-sales acts like leverage, not all hyperscalers will deliver ROI; some multiples will contract as sector broadens; Tesla positioned to outperform dramatically
(18:30–21:24) TSMC, ASML, Applied Digital, SKH all beat numbers; Applied Digital up 84% revenue year-over-year; demand is deep, not shallow hype; arguing AI won’t bring trillions in revenue is “ridiculous”; Matt Sigel’s productivity argument more robust than bubble narratives
(22:26–25:26) Dave Blundin podcast: Two versions of AI progress public models vs. non-public research models further ahead; recursive self-improvement already here; compute is the bottleneck; OpenAI has better models than released but insufficient capacity to deploy
(26:26–28:59) OpenAI securing compute deals for models already built; capacity overhang isn’t warning, it’s roadmap; Dave Blundin projects AGI 2026-27; singularity inflection point arrived; acceleration will be steeper than annual gains; MIT paper shows AI writing its own code
(29:42–31:21) Memory super cycle driven by inference, not training; 24/7 workloads create long-duration demand; edge deployments (phones, humanoids, autonomous vehicles) require more memory; structural super cycle through 2027
(32:23–35:00) Brad Gerstner on All-In podcast: No dark GPUs, every GPU being used; Nvidia’s moat is AI factory, not just chip; performance per watt is key metric; deploying 500,000 GPUs together is miracle, no customer will risk $50B on unproven architecture
(35:18–38:42) State regulation becoming AI issue; midterm elections will focus on AI and jobless growth; autonomous agents inevitable due to capitalism and shareholder value maximization; firms pursue efficiency even at labor’s expense; Charlie Munger: “Show me the incentive and I’ll show you the outcome”
(39:28–41:08) Google major cancer breakthroughs announced two days apart; 2025 will be “last normal year before world went fully exponential”; humanoids coming with hundreds of thousands by 2030; longevity escape velocity math, gaining two healthy years buys time for next wave
(41:45–43:37) 90 gigawatts by 2029; singularity arriving 2025-26; Bitcoin four-year cycle narrative wrong, S&P correlation matters more; S&P at highs while Bitcoin down, opposite of prior cycles
(43:59–46:16) S&P powered by capex, lower inflation, productivity gains from AI; Bitcoin going higher if S&P has good year; disruption year means Bitcoin benefits; bearish on AI = bearish Bitcoin = bearish Elon Musk; disruption feels like deception; John von Neumann quote on singularity from 1950s

Watch here

DISCLOSURES AND DISCLAIMERS

Analyst Certification

The analyst, 22V Research Group, primarily responsible for the preparation of this research report attests to the following: (1) that the views and opinions rendered in this research report reflect his or her personal views about the subject companies or issuers; and (2) that no part of the research analyst’s compensation was, is, or will be directly related to the specific recommendations or views in this research report.

Analyst Certifications and Independence of Research.

Each of the 22V Research analysts whose names appear on the front page of this report hereby certify that all the views expressed in this Report accurately reflect our personal views about any and all of the subject securities or issuers and that no part of our compensation was, is, or will be, directly or indirectly, related to the specific recommendations or views of in this Report.

22V Research (the “Company”) is an independent research provider. The Company is not a member of the FINRA or the SIPC and is not a registered broker dealer or investment adviser. 22V Research has no other regulated or unregulated business activities which conflict with its provision of independent research.

22V Research, LLC is a professional services and independent publication organization. 22V Research, LLC is not a securities broker-dealer, not a member of the Financial Industry Regulatory Authority (FINRA), not a registered investment advisor (RIA) and not a member of SIPC.

Securities transactions, when offered, are offered by 22V Securities, LLC through LPS Capital, LLC. Certain employees of 22V Securities, LLC are dually registered as securities representatives of LPS Capital, LLC or Analyst Hub Securities, LLC. 22V Securities, LPS Capital and Analyst Hub Securities are members FINRA, SIPC.

https://brokercheck.finra.org/

Current Ratings Definition.

SECTOR OUTPERFORM: An “outperform” rating anticipates the company will outperform the S&P Regional Banking Index (peer group).

SECTOR PERFORM: A “market perform” rating anticipates the company will perform in line with the S&P Regional Banking Index (peer group).

SECTOR UNDERPERFORM: An “underperform” rating anticipates the company will underperform the S&P Regional Banking Index (peer group).

Limitation Of Research And Information.

This Report has been prepared for distribution to only qualified institutional or professional clients of 22V Research Group. The contents of this Report represent the views, opinions, and analyses of its authors. The information contained herein does not constitute financial, legal, tax or any other advice. All third-party data presented herein were obtained from publicly available sources which are believed to be reliable; however, the Company makes no warranty, express or implied, concerning the accuracy or completeness of such information. In no event shall the Company be responsible or liable for the correctness of, or update to, any such material or for any damage or lost opportunities resulting from use of this data. Nothing contained in this Report or any distribution by the Company should be construed as any offer to sell, or any solicitation of an offer to buy, any security or investment. Any research or other material received should not be construed as individualized investment advice. Investment decisions should be made as part of an overall portfolio strategy and you should consult with a professional financial advisor, legal and tax advisor prior to making any investment decision. 22V Research Group shall not be liable for any direct or indirect, incidental or consequential loss or damage (including loss of profits, revenue or goodwill) arising from any investment decisions based on information or research obtained from 22V Research Group.

Reproduction And Distribution Strictly Prohibited.

No user of this Report may reproduce, modify, copy, distribute, sell, resell, transmit, transfer, license, assign or publish the Report itself or any information contained therein. Notwithstanding the foregoing, clients with access to working models are permitted to alter or modify the information contained therein, provided that it is solely for such client’s own use. This Report is not intended to be available or distributed for any purpose that would be deemed unlawful or otherwise prohibited by any local, state, national or international laws or regulations or would otherwise subject the Company to registration or regulation of any kind within such jurisdiction.

Copyrights, Trademarks, Intellectual Property.

22V Research Group, and any logos or marks included in this Report are proprietary materials. The use of such terms and logos and marks without the express written consent of 22V Research Group is strictly prohibited. The copyright in the pages or in the screens of the Report, and in the information and material therein, is proprietary material owned by 22V Research Group unless otherwise indicated. The unauthorized use of any material on this Report may violate numerous statutes, regulations and laws, including, but not limited to, copyright, trademark, trade secret or patent laws.