NOTES STORYBOARDS PULSE PORTFOLIO WEBINARS
AI Pulse
WeeklyJanuary 14, 2026

AI Pulse January 14, 2026

Capex Fatigue, Agentic Ambitions, and the Continuing Consulting Shakeout

AI Infrastructure Skepticism Continues Rise as Capex Fatigue Meets Energy Constraints

The collision between Wall Street's patience and Silicon Valley's ambitions grew more pronounced this week as analysts and investors voiced mounting concerns about the sustainability of massive AI infrastructure spending. Perscient's semantic signature tracking language questioning massive AI infrastructure projects registered a z-score of 2.5, while our signature capturing sentiment that AI capital expenditure represents a risky bet with uncertain payoffs stood even higher at 2.7.

The numbers behind this skepticism are striking. Meta recently raised its capital expenditure forecast to between $70 billion and $72 billion, up from a prior range of $66 billion to $69 billion. Across the hyperscaler sector, roughly $400 billion flowed into AI infrastructure in 2025, yet Goldman Sachs warned this week that Big Tech may only generate half the profit needed to justify these investments. The investment bank noted that as AI capex growth begins to slow in 2026, the profit threshold required to justify existing spending will only climb higher.

Our semantic signature tracking resistance to big AI capital projects sits at a z-score of 1.8. One particularly blunt assessment circulating on social media characterized OpenAI's capex plan as something approaching "a pyramid scheme for investors," pointing to spending levels that far outpace any defensible moat. Meanwhile, a survey of investors revealed a repositioning: only 20% still favor large U.S. tech firms for AI investments, while over half now prefer energy providers, recognizing that power generation has become the binding constraint on AI progress.

Coverage from CES 2026 emphasized that the bottleneck has moved from "can we build better models?" to "can we power and cool the infrastructure to run them?" This observation found support in data showing that $98 billion in planned AI data center development was derailed in a single quarter last year by community organizing and pushback, exceeding all disruptions tracked since 2023.

The Trump administration has now entered this arena directly. President Trump announced Friday that Microsoft would make "major changes" to ensure that consumers don't see rising utility bills tied to AI infrastructure. Microsoft responded by unveiling an initiative to curb water usage and limit power consumption impacts. Our semantic signature tracking calls to regulate AI's impact on energy and water rose by 0.1 over the past week to a z-score of 1.5.

The energy constraint is forcing creative solutions. Elon Musk's xAI purchased five massive 380MW gas turbines to power over 600,000 Nvidia GB200-class GPUs, building energy infrastructure independent from public grids. Macquarie's latest memory report suggested that combined supply capacity of major memory makers can only support roughly 15GW of AI data center buildout over the next two years.

Despite these headwinds, only 7% of surveyed investors believe that AI represents a market bubble. The tension captured by elevated readings for both "AI Capex is Huge and Will Keep Growing" (z-score 1.8) and "Companies are becoming more skeptical of big AI investments" (z-score 1.1) reflects a market grappling with the gap between AI's transformative potential and the physical, financial, and operational realities of bringing that potential to fruition.

Agentic AI Emerges as the Enterprise Battleground of 2026

While infrastructure debates dominated one conversation, enterprise technology leaders focused on a different question: when will AI move from augmenting human work to performing it autonomously? Our semantic signature tracking language about improved agentic behavior as the path to LLM advancement registered a z-score of 0.5, but more striking was its trajectory, rising by 0.7 over the past week, the largest single-week increase in our dataset.

The agentic AI field is transitioning from experimental prototypes to production-ready systems. Industry analysts project that the market will surge from $7.8 billion today to over $52 billion by 2030, while Gartner predicts that 40% of enterprise applications will embed AI agents by the end of 2026, up from less than 5% in 2025. IDC forecasts that up to 40% of all Global 2000 job roles will involve working with AI agents this year.

Jason Mendel of Battery Ventures framed 2026 as the moment AI tools evolve into agents that automate full workflows rather than merely augmenting them, "delivering on the human-labor displacement value proposition in some areas." One technology executive noted on social media that enterprise AI deployment reveals capabilities "far more powerful than the chat interfaces consumers are used to," predicting that labor as a percentage of revenue will decline at many companies within five years.

The practical challenges of this transition are becoming clearer. While nearly two-thirds of organizations are experimenting with AI agents, fewer than one in four have successfully scaled them to production. Enterprises are discovering that automating existing processes designed for human workers without reimagining how work should actually be done produces disappointing results.

Our semantic signature tracking the proposition that AI is now a necessary skill declined by 0.6 over the past week, falling from 1.3 to 0.7. This moderation may reflect attention shifting from personal AI skills to system-level agentic deployments. Simultaneously, our signature tracking context and memory improvement as the path to LLM advancement rose from 0.5 to 0.6.

The infrastructure requirements for agentic AI are substantial. One analysis described "Agent Harnesses" as the next layer of infrastructure needed, systems that wrap around AI models to manage long-running tasks, providing prompt presets, human-in-the-loop handling, lifecycle hooks, and sub-agent management. McKinsey emphasized that while agentic AI is powerful, autonomy introduces new security and trust risks that leaders cannot ignore.

Some observers counsel caution. Agents may be following generative AI into what Gartner calls the trough of disillusionment. Our semantic signature tracking the proposition that productivity gains from AI haven't materialized stands at a z-score of 1.5, suggesting that enterprises are betting on agents to finally deliver the productivity improvements that earlier AI implementations failed to produce. As one analyst put it, enterprises have stopped asking "Can your AI summarize this?" and started asking "Can it take action without breaking things?"

Consulting Industry Continued to Bear the Brunt of Expected Disruption

The theoretical debates about AI's impact on knowledge work became considerably more concrete this week with McKinsey & Company's announcement that it plans to cut roughly 10% of its workforce. Our semantic signature tracking language about AI destroying jobs in consulting registered a z-score of 1.1, while the broader signature tracking AI's transformation of white-collar work remained near average at 0.1.

McKinsey's CEO Bob Sternfels explained that AI had led to a workforce shift, with half the business adding jobs and the other half shrinking. The firm now employs what it describes as 40,000 humans and 25,000 AI agents, up from only a few thousand agents eighteen months ago. The goal is for every employee to be "enabled by at least one or more agents" within the next 18 months, with AI initiatives now accounting for 40% of the firm's work.

Analysis of McKinsey's moves characterized this as signaling "an irreversible shift in how value is created in consulting." Traditionally, elite firms thrived on top talent from prestigious schools excelling in problem-solving and insight generation. Digital tools reduced information asymmetry between consultants and clients, but generative AI eliminates the edge in data crunching and recommendations that once justified premium fees.

The numbers on white-collar displacement are sobering. The World Economic Forum's Future of Jobs Report 2025 projects 92 million roles displaced by 2030, pointing to 22% churn in the global workforce with roles eliminated and recreated faster than workers can realistically transition. Over the next five years, nearly 300 million white-collar jobs may be impacted globally, with around 100 million at risk of becoming obsolete. Challenger, Gray & Christmas recorded nearly 55,000 AI-linked job cuts in 2025 alone, a category that barely existed before.

Our semantic signature tracking the proposition that AI will cause massive unemployment remained near average at a z-score of 0.3 and actually declined by 0.2 over the past week. Similarly, our signature tracking the idea that AI will create entirely new jobs remained near average at -0.1.

Some analysts urge caution in interpreting layoff announcements. Forrester noted that financially driven layoffs are being confused with AI-driven layoffs, and many companies announcing AI-related cuts do not have mature AI applications in place. "The complexity here is that many enterprises, despite how ready or not they are to successfully use AI solutions, will say that they are increasing their investments in AI to explain why they are cutting back spending in other areas or trimming workforces. In reality, AI will become the scapegoat for executives looking to cover for past mistakes."

McKinsey is changing how it hires, seeking candidates who can move between traditional consulting and an engineering mindset, people who can work alongside AI. One observer summarized the transformation: client-facing roles growing by 25%, non-client roles declining by 25%, with productivity up by 10% in shrinking functions.

After the rise of AI video tools in 2025, analysts warn that 2026 could bring visible job losses, especially in lower-level white-collar roles. Trades and hands-on professions remain less exposed for now. The combination of elevated readings for our consulting disruption signature alongside moderate readings for "AI is entering a trough of disillusionment" (z-score 1.4) suggests a narrative where AI's enterprise impact is becoming more tangible even as broader hype moderates.

Recent Pulses

Markets PulseWeeklyNov 10, 2026

AI Pulse November 03, 2026

Local opposition is now the loudest frame in the set, and it has moved from rhetoric to the ballot. Coverage this week centered on 41 data center measures across ten states, 21 Ohio initiatives alone, and a steady drumbeat of municipal moratoriums from South Carolina to Minnesota to New York. The reporting increasingly frames elected officials as pricing political risk rather than courting investment — Pennsylvania's governor reversing course inside two years being the emblematic case — while capital allocators such as Larry Fink counter that siting restrictions themselves pose macroeconomic danger. The narrated bottleneck has migrated away from accelerators toward land, power, and memory. Language asserting that memory chip shortages constrain AI growth posted one of the week's sharpest increases, anchored to reporting that DRAM prices have roughly doubled since the end of 2025, while claims about GPU scarcity sat flat and below average. Grid interconnection and construction-delay framing held at an elevated plateau, and doubts about specific hyperscale projects climbed even though expectations of continued expansion did not move — a combination suggesting that skepticism is attaching to individual builds rather than to the build-out as a whole. Market coverage rotated from contagion talk toward line-item discipline. The density of language claiming that businesses doubt large AI spending produced the single largest move in the set, and framing of infrastructure spending as a dangerous gamble strengthened, while predictions of an AI-driven market-wide crash retreated meaningfully from recent highs. Supporting detail was granular rather than systemic: CFO surveys showing consumption costs overshooting budgets, absent cost-governance mechanisms, Microsoft reportedly trimming internal model spend, and credit markets charging AI-linked borrowers above 15 percent. Notably, the critique stayed financial rather than technical, since signatures tracking failed deployments, missing productivity gains, and disappointment cycles all remained subdued. Taken together, Sections 1 and 2 describe the same pressure arriving through two different doors. Voters are raising the political cost of siting at precisely the moment that investors are raising the evidentiary bar on returns, and both pressures converge on the same question put to vendors: what does this specific project cost, who absorbs the memory and power pass-through, and when does it pay back. The revival of 1990s telecom-overbuild comparisons fits that mood, implying durable assets with impaired equity returns rather than stranded capacity. Leadership framing has narrowed to one name tied to one event, while the societal debate has gone largely silent.Anthropic dominates competitor-leadership language on the strength of its S-1 disclosures, xAI's leadership framing posted the week's steepest decline, and both OpenAI and Google remained quiet. Scarcity framing meanwhile extended into training data, where restricted access to high-quality web sources is drawing fresh attention. Against this, signatures tracking job destruction, job creation, universal basic income, and most domain transformations sit well below average — leaving the jobs, education, and health conversations with unusually little media competition for any organization seeking to be heard there.

Private Credit PulseBiweeklyNov 3, 2026

Private Credit Pulse

Next release: November 17, 2026

Markets PulseWeeklyNov 3, 2026

AI Pulse October 27, 2026

Coverage has relocated the bear case from technology to finance. Perscient's semantic signature tracking the density of language predicting that an AI investment collapse will crash overall markets posted the largest gain among market-framing signatures, carried by the IMF's stability warning, Dalio's "classic bubble" language, and the circular-financing critique now reaching Capitol Hill. Yet the signature tracking claims that AI hype is giving way to disappointment stayed flat and negative, which indicates that reporters are questioning who earns a return rather than whether the technology works. Conviction and alarm are growing simultaneously, not sequentially. The signature tracking language predicting a long-term AI investment supercycle recorded the single largest weekly gain in the entire set, moving out of negative territory at the same moment bubble language climbed. The press is staging a genuine two-sided argument heading into the Alphabet, Amazon and Apple prints, where the sell-side test has moved from growth to cash conversion. The bottleneck vocabulary has migrated from silicon to permission, power and memory. Language tracking GPU scarcity as a growth constraint sat below its long-term mean and has effectively left the conversation, while claims that opposition to large AI investments is increasing produced the highest reading in the entire set, powered by Oregon's statewide pause, Charlotte's extended moratorium and a flood of California local measures. Memory-shortage language cooled sharply on the week even though IDC and Michael Dell both describe worsening DRAM and HBM4 supply, suggesting that press attention is lagging the underlying constraint. Siting fights are still being covered as local disputes rather than a competitiveness emergency. The signature tracking arguments that national energy buildout will determine AI leadership rests near its long-term mean, which leaves a gap between the permitting friction documented in Section 2 and the geopolitical framing one might expect. Should interconnection queues and moratoriums persist, that gap is the most likely place for a new narrative to form. The leaderboard story has narrowed to one name while buyers are portrayed as disciplined rather than disillusioned.Anthropic's October 26 S-1 pushed the signature tracking claims of Claude's competitive lead to the dominant position, with OpenAI and Google framings both flat and negative; the filing's $80.7 billion half-year loss headline simultaneously feeds the risky-bet language described in Section 1. Meanwhile, the signature tracking corporate doubt about large AI spending strengthened even as claims of outright project failure weakened, and predictions about job destruction or creation remain deep in negative territory, leaving the week's coverage almost wholly financial and infrastructural.

Browse the Pulse Archive