Wall Street Bets on AI While America Counts Pennies

Somewhere between the server farm and the grocery aisle, American capitalism has split in two.

On one side: AI stocks have outpaced the non-AI S&P 500 by 121 percentage points since the start of 2024, according to Polymarket data published on 22 May 2026. Zoom Video Communications reported a 10 percent stock surge the same day, citing strong demand for its AI features. The math is unambiguous. Capital loves artificial intelligence.

On the other: U.S. consumer sentiment crashed to a record low on 22 May 2026, with 57 percent of consumers telling surveyors that high prices are actively eroding their personal finances. The same day, prediction markets placed a 92 percent probability on an AI data-center moratorium passing by year-end. The regulatory whiplash is already priced in — or at least the possibility of it.

The two realities coexist without appearing to trouble each other. That is the story.

The Index That Ate Everything Else

The concentration of AI-adjacent returns in a handful of megacap names has become a defining feature of American equity markets. Nvidia, Microsoft, Alphabet, Meta — their combined weight means that the S&P 500 is increasingly a bet on one sector's continued dominance rather than a proxy for the broader economy. When those stocks rise, the headline index rises. When the index rises, policy frameworks — tax treatment, antitrust tolerance, permitting timelines — bend toward sustaining the conditions that produced the rise.

This is not new. Markets have always had sectors that pull the narrative. What is new is the speed at which AI has become the gravitational centre of equity valuation, and the degree to which that gravitational pull has decoupled from any conventional measure of consumer wellbeing.

Zoom's Tuesday surge illustrates the dynamic. The company's AI features — transcription, summarisation, meeting coaching — are genuinely useful products. They are also products priced for businesses and professionals, not for the 57 percent of consumers reporting that everyday costs are consuming their households. Zoom's customer base skews toward exactly the economic stratum that is driving AI equity returns. It is a closed loop.

The Infrastructure Reckoning Nobody Wants to Have

The 92 percent probability assigned to an AI data-center moratorium by year-end is the most revealing number in this week's data. It suggests that prediction markets — instruments used by people who study risk for a living — now consider environmental pushback on AI infrastructure a near-certainty, not a tail risk.

Data centers are not abstract. They consume extraordinary quantities of electricity and water. A single large facility can draw power equivalent to a small town. As the AI industry has scaled, local communities, state regulators, and federal environmental agencies have begun asking the questions that the initial gold-rush framing deferred: who pays for the grid upgrades, who absorbs the water-table impact, what happens to local utility bills when a hyperscaler signs a priority supply agreement?

The moratorium probability reflects those questions reaching a threshold of political seriousness. It does not mean a moratorium will pass. But it means the industry understands the question is live. The fact that AI stocks continue to climb even as this regulatory cloud gathers is itself informative: markets are betting on the long term and discounting the near term, or they are simply not connecting infrastructure constraints to equity valuations. Neither interpretation is comfortable.

The Sentiment Gap Is Structural, Not Cyclical

The record low in consumer sentiment is easy to dismiss as noise — a momentary dip in confidence that will correct as inflation continues to ease. That reading is becoming harder to sustain.

Fifty-seven percent of consumers reporting that high prices are eroding personal finances is not a confidence metric. It is a balance-sheet observation. People are not feeling bad about the future; they are reporting that present spending is outrunning present income in a way that leaves less margin. The distinction matters because balance-sheet problems do not resolve with sentiment recovery. They resolve with income growth, price relief, or debt — and the last of those options tends to create the conditions for the next cycle's problems.

The AI economy is not generating the kind of broad-based employment income growth that would address this gap. Server-farm construction creates construction jobs; the ongoing operations are capital-intensive, not labour-intensive. The productivity gains AI generates accrue primarily to capital owners and to businesses that can deploy the technology at scale. The distributional profile is clear, even if the political consequences have not yet fully arrived.

The Kicker

What prediction markets are quietly acknowledging, in the language of probabilities rather than editorials, is that the AI boom exists inside a social and political context that will eventually assert itself. A moratorium that passes with 92 percent probability is not a surprise; it is a reckoning that has been priced, at least partially, into the odds.

The harder question is what happens to consumer sentiment in the interim — the months or years between now and the moment when infrastructure constraints, equity concentration, and balance-sheet distress produce a policy response. Markets have so far shown no inclination to correct this divergence on their own. The regulatory system is moving slowly, and the political will to impose meaningful costs on AI infrastructure development remains diffuse.

That is the window in which the gap between Wall Street and the grocery aisle either narrows — through wage growth, price relief, or a correction that redistributes capital back toward consumer-facing businesses — or widens until the dissonance becomes politically ungovernable. The Polymarket odds suggest the industry knows which direction the wind is blowing. The record-low sentiment numbers suggest the rest of the country has been feeling it for longer than the indexes admit.

This publication compared Polymarket probability data against conventional consumer-confidence and equity-flow reporting. The gap between the two information sets is itself the story.