AI Capital Expenditure Concerns Finally Becoming Doubts

Hyperscale Capex Doubts Creep in Alongside Soaring Investor Scrutiny

It seems difficult to believe after being front and center for much of 2025, but the density of language we associate with AI infrastructure investment narratives has reached yet another new peak. This week, however, it seems that the language of long-standing concerns about massive capex are finally driving actual doubts about the future of hyperscale builds. Both narratives, in fact, reached their all-time highs. Perscient's semantic signature tracking language expressing doubt about the pace of hyperscale builds hit a z-score of 3.80, reflecting dense language conveying uncertainty that massive AI infrastructure projects can continue at the expected pace. Meanwhile, the semantic signature capturing skepticism about AI capital expenditure returns also reached its highest recorded level at 3.50.

The numbers behind this narrative tension are staggering. Hyperscale datacenter operators nearly tripled their spending on infrastructure over the past three years in response to AI demand, while operational capacity added each quarter has increased by 170 percent. Bank of America analysts estimate that AI capital expenditure from just five hyperscalers will total $399 billion this year and rise to over $600 billion in coming years, with roughly 75 percent of that spend directly tied to AI infrastructure rather than traditional cloud services.

Yet this spending surge is increasingly funded through debt rather than operating cash flows. Global technology companies have ramped up debt issuance to record levels, as the race to build AI capacity forces even cash-rich firms to borrow heavily. Hyperscalers have issued roughly $121 billion in new debt to fund AI and data center expansion in 2025 alone. This represents a fundamental transformation of historically cash-funded business models into ones utilizing leverage.

The semantic signature capturing corporate skepticism about large AI investments rose by 0.34 to 2.96, also approaching its highest recorded level. As we noted in recent AI Pulse reports, BCA Research has warned that markets may be entering a "Metaverse Moment," a phase where investors begin punishing AI companies for raising capital spending. Global data center dealmaking surged to another record this year, driven by the rush to build infrastructure for energy-intensive AI workloads, even as investors grew increasingly wary of inflated valuations.

Comparisons to the 1990s telecom overbuild persist in media coverage, though they moderated slightly to a z-score of 2.6 over the past week. Skeptics continue to highlight similarities to the dot-com buildout, emphasizing hallmarks of past bubbles: excessive speculation, leverage, and momentum-driven prices detached from fundamentals. The coexistence of record-high language about both continued expansion and investment risk suggests media narratives are capturing a market at a genuine inflection point.

Overregulation Concerns Rise Following Federal Preemption Push

The tension between infrastructure expansion and investment skepticism extends into the regulatory sphere, where competing narratives are rising in intensity. The semantic signature tracking concerns that overregulation could harm American competitiveness showed the largest one-week increase among all signatures, rising by 0.51 to 0.56. This intensification coincided with President Trump’s signing of an executive order on December 11, 2025 that seeks to limit states' ability to regulate AI.

The administration has framed this order as a response to what it views as an urgent crisis: a rapidly fracturing AI regulatory environment driven by state action. The White House stated that "to win, United States AI companies must be free to innovate without cumbersome regulation." The order establishes federal policy to create a "minimally burdensome national standard" and directs the Department of Justice to challenge state laws deemed inconsistent with that goal.

Prior to the executive order, the administration had supported legislative proposals that would have imposed a sweeping 10-year moratorium on new state AI laws. When those congressional efforts failed, the administration turned to executive action. Legal analysts note that the practical impact may be limited in the short term, and companies are likely well-advised to continue operating under the expectation that states will legislate and enforce their AI-related laws.

The semantic signature tracking concerns about biased AI content restrictions rose by 0.47 to 0.36, reflecting increased attention to viewpoint-based mandates in AI systems. Meanwhile, the semantic signature tracking calls for government oversight of AI risks remained weaker than average at -0.59, indicating that language demanding regulation of AI dangers is not intensifying despite the broader regulatory debate.

Environmental concerns present a countervailing force in this narrative. The semantic signature tracking calls to regulate AI's impact on energy and water rose by 0.22 to 1.06, consistent with the aforementioned sources of resistance to continued AI capex. By December 2025, over 200 environmental groups have called for a moratorium on new data center construction, and several counties in states like Georgia and Maryland have implemented local bans. This regulatory friction represents a new variable for AI investors navigating the infrastructure buildout.

The competitive framing with China pervades much of this debate. Leaders in Silicon Valley have argued that navigating a patchwork of state regulations could slow innovation and affect America's competitiveness in the global AI race. Europe faces its own crossroads, as analysts describe a "fork in the road moment" between competing meaningfully in AI and maintaining climate commitments.

AI Skills Imperative Intensifies

The regulatory debate over AI governance is in many cases directly related to labor concerns, as the technology's rapid adoption has begun forcing workers and industries alike to adapt. The semantic signature tracking the perception that AI has become a necessary skill rose by 0.31 to 1.13. In spring 2025, nearly 47 percent of workers across all sectors reported using AI tools at least once a month, up from 34 percent the previous year. OpenAI has cited research suggesting workers with AI skills earn about 50 percent more than those without them, making the question of which skills matter feel urgent rather than theoretical.

As it did earlier this year, the consulting industry has once again emerged as a focal point for these workforce concerns. McKinsey & Company's announcement that it plans to cut roughly 10 percent of its workforce has sent ripples through the consulting world, reigniting debate about the industry's future. Major consulting firms including McKinsey, Boston Consulting Group, and Bain have frozen starting salaries for the third consecutive year as AI reshapes how these companies think about their traditional reliance on large cohorts of junior analysts.

Internal assessments suggest McKinsey's AI tool Lilli can cut time spent searching for and synthesizing information by up to 30 percent. Still, these narratives are likely to ebb and flow as media coverage floats between various affected industries. This week, the semantic signature tracking concerns about AI destroying consulting jobs actually declined by 0.3 to 1.2, although still well above average.

The broader employment picture shows AI's growing influence. Artificial intelligence was responsible for nearly 55,000 layoffs in the United States in 2025, according to consulting firm Challenger, Gray & Christmas. Major firms including Amazon and Salesforce cut thousands of roles and cited AI as a factor. The pace of programmer employment decline has accelerated since generative AI emerged, with overall programmer employment in the United States falling a dramatic 27.5 percent between 2023 and 2025, according to Bureau of Labor Statistics data.

Language about AI generating novel employment categories is not keeping pace with job displacement concerns, as our semantic signature tracking expectations that AI will create entirely new jobs remained weaker than average at -0.3, declining by 0.2 on the week. Perscient’s semantic signature tracking broader transformation of white-collar work remained about average at 0.1, indicating that sector-specific concerns in consulting and technology are intensifying faster than generalized professional workforce narratives.

The combination of rising AI skills requirements and consulting industry disruption points to a workforce narrative where adaptation is becoming urgent, particularly for knowledge workers in traditionally high-status professional services. As one industry observer noted, this is not about one firm, one round of layoffs, or one business cycle, but signals an irreversible shift in how value is created in consulting.