AI Capex Boom Meets Political Headwinds, Mental Health Concerns and Intensifying Competition
January 6, 2026
AI Capex Boom Meets Political Headwinds, Mental Health Concerns and Intensifying Competition
Hyperscale Infrastructure Expansion Faces Bipartisan Resistance
The artificial intelligence infrastructure buildout has continued that peculiar paradoxical phase in which rapid expansion and skepticism about that expansion are advancing in lockstep. Perscient's semantic signature tracking language affirming continued expansion of massive AI infrastructure registers at a z-score of 3.8, while our signature tracking language questioning hyperscale builds stands at 3.0. This dual elevation captures a media environment where confidence in growth and doubt about its sustainability are being expressed simultaneously.
The political dimensions crystallized over the holiday period when Democratic Socialist Sen. Bernie Sanders and right-wing Gov. Ron DeSantis emerged as unlikely allies in opposing the data center boom. Sanders called for a national moratorium on data center construction in a December 28 CNN interview: "Frankly, I think you've got to slow this process down." DeSantis unveiled an AI bill of rights on December 4 that would protect local communities' right to block data center construction. As one social media observer noted, this bipartisan convergence signals "that a political reckoning is brewing over the AI industry's impact on electricity prices, grid stability and the labor market."
The electricity cost dimension has become acute. Residential electricity prices rose by approximately 5% in 2025 and are forecast to climb by another 4% nationwide in 2026, according to the federal Energy Information Administration. NPR reporting highlighted that average residential rates have jumped from 11-12 cents per kilowatt hour in 2020 to 19 cents in 2025, a 60% increase, and that grid expansion costs are being distributed across all ratepayers rather than falling solely on data center operators. Our semantic signature tracking language calling for regulation of AI's environmental footprint rose to a z-score of 1.5, up by 0.4 week-over-week.
Our semantic signature tracking language asserting that AI infrastructure spending is massive and increasing stands at 3.3, while the signature tracking language characterizing AI infrastructure spending as a risky bet with uncertain payoffs sits at 2.0. Both readings are well above average. A portfolio manager at CenterSquare Investment Management warned that "the largest hyperscale data center developments haven't yet been able to monetize AI technologies" and predicted that "at some point, by 2027 or later, the market will want to see some sort of path to monetization for a lot of this spending because the return on investment capital is so paltry, if not nonexistent."
AI-related services are expected to deliver only about $25 billion in revenue in 2025, roughly 10% of what hyperscalers are spending on infrastructure. Combined capital expenditures across the largest operators are projected to approach $600 billion in 2026. TechCrunch reported that grassroots opposition has blocked or delayed some $64 billion worth of developments, and that growth is increasingly constrained by power availability, equipment lead times, and local resistance. PJM, serving 65 million people across 13 states including swing states Pennsylvania and Virginia, faces what its independent monitor calls a "crisis stage," with the grid projected to be 6 gigawatts short of reliability requirements by 2027.
AI Companionship Use Cases Proliferate, Exacerbate Mental Health Concerns
Perscient's semantic signature tracking language asserting that AI will cause mental health problems rose to a z-score of 1.8, up by 0.5 week-over-week, the largest single-week increase among all AI-related signatures. Our signature tracking language claiming that AI will worsen social isolation maintains an elevated z-score of 1.0, while the signature tracking language arguing against AI for therapeutic purposes stands at 0.9.
The scale of AI companionship use has become difficult to ignore. ChatGPT now has around 810 million weekly active users worldwide, and some reports place therapy and companionship among the top reasons for use. Among younger demographics: roughly three in four U.S. teens have used an AI companion, with around half becoming regular users. One in five reports spending as much or more time with AI companions as with human friends. Research from MIT found that people who are lonely are more likely to consider ChatGPT a friend and spend substantial time on the app while also reporting increased levels of loneliness, suggesting a troubling feedback loop.
The coverage is not uniformly negative, however. Our semantic signature tracking language asserting that AI can provide meaningful mental health support registered at 0.6, up by 0.1, while the signature tracking language suggesting that AI provides meaningful companionship stands at 0.6. The Guardian examined research suggesting that AI companions may reduce feelings of loneliness, noting that one in six people globally experience loneliness, which is associated with a 26% increase in premature mortality. One user told Fortune that her AI companion "reminds me when I'm working to eat something and drink water—it's good to have somebody who cares."
Yet the clinical evidence increasingly tilts the scales toward risk. A four-week randomized controlled trial found that heavy daily use correlated with greater loneliness, dependence, and reduced real-world socializing. Psychiatric research has documented cases where intense engagement with AI chatbots contributed to delusional thinking or suicidality. Social psychologist Jonathan Haidt warned on social media of "a future that is a combination of two great movies: Idiocracy and The Matrix," urging that we "don't put chatbots into teddy bears, or schools, or children's lives, until we understand how they affect children's development."
Legal and regulatory responses are emerging. MIT Technology Review notes that states face mounting public pressure to push for guardrails, and that California has passed SB 243 to regulate AI companion bots. OpenAI updated ChatGPT's model in October with input from 170 mental health professionals to establish guardrails. Multiple wrongful death lawsuits are proceeding against AI companies, testing whether they can be held liable for what their chatbots encourage users to do.
AI Competition Narratives Emphasizing Efficiency
The infrastructure spending debate connects directly to questions about competitive positioning. Perscient's semantic signature tracking language asserting that Anthropic or Claude leads the AI competition registered at 3.7, up by 0.1 week-over-week, while our signature tracking language asserting that OpenAI leads stands at 2.3, up by 0.3. Both elevated readings reflect intense media focus on the positioning of these firms, though with notably different framings.
Anthropic's president Daniela Amodei articulated a distinctive strategic vision in recent interviews, arguing that "the next phase won't be won by the biggest pre-training runs alone, but by who can deliver the most capability per dollar of compute." While OpenAI has secured approximately $1.4 trillion in headline compute commitments, Amodei emphasized that "Anthropic has always had a fraction of what our competitors have had in terms of compute and capital, and yet, pretty consistently, we've had the most powerful, most performant models."
The demand for financial returns was a consistent theme among experts surveyed about 2026 expectations. "2026 is the 'show me the money' year for AI," said Venky Ganesan of Menlo Ventures. "Enterprises will need to see real ROI in their spend, and countries need to see meaningful increases in productivity growth." Our semantic signature tracking language asserting that AI investment themes remain durable stands at 1.8, down by 0.6 week-over-week.
The labor implications of AI advancement are becoming concrete. Our semantic signature tracking language predicting that AI will eliminate consulting industry positions stands at a z-score of 1.2, elevated compared to other job-displacement signatures. McKinsey & Company announced plans to cut roughly 10% of its workforce, reigniting debate about the future of the industry. Battery Ventures investor Jason Mendel predicted that "2026 will be the year of agents as software expands from making humans more productive to automating work itself, delivering on the human-labor displacement value proposition in some areas."
Our semantic signature tracking language asserting that Deepseek or China leads the AI competition registered at 3.4, while the signature tracking language asserting that large AI spending is required for China competition also stands at 3.4. MIT Technology Review observed that by year-end 2025, "'DeepSeek moment' had become a phrase frequently tossed around by AI entrepreneurs, observers, and builders," an aspirational benchmark for what a relatively small firm in China could achieve with limited resources. DeepSeek's release of its open-source reasoning model R1 in January 2025 challenged assumptions about the relationship between scale and capability. Analysts noted that the real story was not just DeepSeek but the wave of world-class Chinese AI models that followed, including Alibaba Qwen, Moonshot Kimi, and ByteDance Doubao.
DISCLOSURES
This commentary is being provided to you as general information only and should not be taken as investment advice. The opinions expressed in these materials represent the personal views of the author(s). It is not investment research or a research recommendation, as it does not constitute substantive research or analysis. Any action that you take as a result of information contained in this document is ultimately your responsibility. Epsilon Theory will not accept liability for any loss or damage, including without limitation to any loss of profit, which may arise directly or indirectly from use of or reliance on such information. Consult your investment advisor before making any investment decisions. It must be noted, that no one can accurately predict the future of the market with certainty or guarantee future investment performance. Past performance is not a guarantee of future results.
Statements in this communication are forward-looking statements. The forward-looking statements and other views expressed herein are as of the date of this publication. Actual future results or occurrences may differ significantly from those anticipated in any forward-looking statements, and there is no guarantee that any predictions will come to pass. The views expressed herein are subject to change at any time, due to numerous market and other factors. Epsilon Theory disclaims any obligation to update publicly or revise any forward-looking statements or views expressed herein. This information is neither an offer to sell nor a solicitation of any offer to buy any securities. This commentary has been prepared without regard to the individual financial circumstances and objectives of persons who receive it. Epsilon Theory recommends that investors independently evaluate particular investments and strategies, and encourages investors to seek the advice of a financial advisor. The appropriateness of a particular investment or strategy will depend on an investor's individual circumstances and objectives.
