AI Spending on Trial

Executive summary

Wall Street's grilling of Meta, Amazon, and Apple over AI spending has landed with the public — and most people think the pressure is justified. As big tech collectively commits nearly $720 billion in 2026 capital expenditures on AI infrastructure, a new pulse survey finds that nearly half of Americans (46.3%) view analyst scrutiny as necessary oversight, not noise.

The results reveal a public that is neither anti-AI nor blindly optimistic. People want AI spending to produce jobs and economic growth above all else — but open-ended responses expose a deep anxiety that the opposite is happening, as companies like Oracle cut 30,000 workers explicitly to fund AI data centers. Consumer trust in big tech's AI intentions is moderate at best, with respondents leaning toward the belief that benefits will flow to shareholders and enterprise partners before they reach everyday users.

Four signals stand out: the oversight instinct is bipartisan and accountability-driven, not anti-technology; job creation is the dominant success metric but is shadowed by displacement fears; the 'too early to judge' camp is patient but transactional, expecting eventual consumer cost relief; and structural competitive dynamics make the spending slowdown that many respondents prefer essentially impossible for any single company to enact alone.

Context

The survey was fielded in April 2026 — the exact week Meta, Amazon, and Apple released first-quarter earnings that put AI spending at the center of every analyst call. Eighty-two respondents answered four questions: two multiple-choice on sentiment and preferred outcomes, and two open-ended probing concerns and trust levels. The study is a pulse survey, designed to capture a real-time public reaction to a live financial and policy debate rather than to track long-term trends.

The financial backdrop is staggering in scale. Meta guided 2026 capital expenditures to $115–$135 billion, with Q4 2025 expenses growing 40% against 24% revenue growth — a gap that rattled investors. Amazon committed $200 billion in capex for 2026, the largest single-year corporate capital program in US history, anchored by AWS revenue hitting a $15 billion quarterly run rate. Apple, meanwhile, is betting roughly $1 billion per year on a Google Gemini partnership intended to redesign Siri for iOS 27. Together, the five largest US hyperscalers project $720 billion in combined 2026 AI infrastructure spending — a figure Bloomberg pegs at roughly 2% of US GDP.

The public signal this survey captures sits at an uncomfortable intersection: institutional investors are demanding proof of return on investment, corporate executives are arguing the payoff is real but long-dated, and workers are watching Oracle eliminate 30,000 jobs to fund data centers in the same month these earnings drop. The question the study probes is not whether AI investment is happening — that is settled — but whether ordinary people believe it will work for them, and whether they think anyone is watching to make sure it does.

External benchmarks sharpen the stakes. Stanford HAI's 26-nation survey found only 36% of people globally believe AI will improve the national economy, and just 31% expect a positive job market impact. Only 26% of enterprises successfully scaled AI pilots to measurable ROI in 2025, per Stanford's Enterprise AI Playbook. This is the credibility environment in which big tech is asking the public — and Wall Street — for patience.

Findings

Nearly Half Back Wall Street's AI Spending Scrutiny — And They Want Results, Not Restraint

The plurality view is clear: 46.3% of respondents describe Wall Street's pointed questioning of Meta, Amazon, and Apple's AI budgets as necessary oversight to ensure responsible spending. But dig into what these respondents actually want, and the picture is more nuanced than a simple skeptic label suggests.

Respondents in the oversight camp are 89% more likely than the overall sample to list 'better products and services for users' as their top preferred outcome from AI spending. That means the loudest bloc calling for accountability is not asking companies to spend less — they are asking companies to spend smarter and prove it reaches consumers. The oversight impulse tracks with real institutional pressure: a shareholder resolution demanded IBM detail its AI bias protocols, and BofA analyst Justin Post cut his Meta price target from $885 to $820 while maintaining a Buy rating, citing investor apprehension about near-term returns.

A substantial secondary group — 32.9% — says it is simply too early to judge AI returns. This is not apathy; it is a patient but transactional posture. Respondents who hold this view are 63% more likely to prioritize 'lower costs for consumers' as their preferred outcome, suggesting they are willing to wait for ROI but expect the eventual payoff to arrive as savings in their own bills. That expectation has some market support: GPU rental rates have collapsed from roughly $8 per hour in 2024 to below $1 in early 2026, a price signal that could eventually flow through to consumer-facing AI services if competitive dynamics hold.

Only 11% think companies should prioritize profits over AI spending — meaning anti-investment sentiment is a fringe position, not a mainstream one.

Jobs Are the Scoreboard — But the Score Is Already Going the Wrong Way

When asked what outcome they most prefer from big tech's AI investment, 42.7% of respondents chose more jobs and economic growth — the single largest response by a wide margin. Lower costs for consumers ranked second at 34.1%. Better products and services trailed at 19.5%, with just 3.7% selecting other options.

This result reframes the AI spending debate entirely. The public is not primarily evaluating these investments on the product roadmap or the quarterly earnings beat — they are using macroeconomic outcomes as the scoreboard. Big tech's AI bets will be judged, in the court of public opinion, by whether they make the labor market better or worse.

The problem is that the early evidence cuts the wrong way. Oracle's March 2026 mass layoff — 30,000 workers, the largest tech layoff of the year — was explicitly tied to funding AI data centers, not to financial distress. Oracle's net income jumped 95% in the same period. Approximately 12,000 of those cuts fell on workers in legacy software roles that the company said AI would replace. Open-ended survey responses surfaced the same fear repeatedly: respondents cited loss of jobs, reduced job security, and the concern that AI investments come at the cost of layoffs. Stanford HAI's global survey found only 31% of people expect AI to have a positive job market impact — aligning closely with the anxiety visible in this study's free responses.

The tension is sharp: 42.7% of respondents name job creation as their top hoped-for outcome, while available evidence suggests displacement is the more immediate reality. That credibility gap is the central challenge big tech has not yet answered.

Consumer Trust Is Moderate, Conditional, and Quietly Skeptical

Open-ended responses on trust reveal a public that has not been won over. The dominant concern is distributional: respondents doubt that AI gains will be passed down to everyday users rather than captured by shareholders, enterprise clients, and large tech partners. The Beneficiary Focus dimension — derived from free-response scoring — leans toward consumer protection with a mean of -0.32 on a -1 to +1 scale, a statistically significant result.

The Spending Scale dimension tells a parallel story: respondents show a modest but statistically significant tilt toward limiting AI investment (mean -0.25), though the distribution is polarized rather than uniformly cautious. Voices range from 'total waste of money' to 'as long as the innovations help society, I am all for it' — which means the center is thin and the camps are real.

One personality signal worth watching: respondents who score higher in Agreeableness show meaningfully higher trust that AI investments will benefit everyday consumers (Spearman r = 0.251). More agreeable people extend more benefit of the doubt to big tech. This suggests that trust is partly dispositional — meaning companies trying to move the needle will need to do more than communicate better; they need to demonstrate tangible consumer benefits that convert even the skeptically-inclined.

Fathom's bipartisan national poll offers an external anchor: 84% of Americans support retraining programs for displaced workers, and 71% support a sovereign wealth fund that shares AI-generated wealth publicly. The public 'decisively rejects a let-the-market-sort-it-out approach,' in Fathom's words — a posture that mirrors this study's Beneficiary Focus lean toward consumer protection over tech-partner benefit.

The Structural Trap: Public Wants Restraint, Markets Demand More

The most consequential finding may be the one that is hardest to act on. Respondents lean toward limiting AI spending and demand accountability — but the competitive structure of the hyperscaler market makes unilateral restraint a strategic liability, not a viable option.

The five largest US hyperscalers have collectively committed $720 billion in 2026 capex. Analysts warn that any company that hesitates risks being commoditized — reduced to a utility while faster-moving rivals capture the next wave of AI value. Schroders' Q2 2026 CIO outlook describes a risk scenario where an AI bubble bursts in Q3 as expectations go unmet, dragging consumer spending down and lifting unemployment — the exact downside the 'necessary oversight' plurality implicitly fears. But Schroders also outlines a bull scenario where rapid AI adoption drives investment-led growth, mirroring the optimism of the 'too early to judge' camp.

The polarized Spending Scale distribution — with a mean only slightly negative at -0.25 — reflects this structural ambivalence. There is no strong consensus for dramatic cuts. Respondents are uncomfortable with the scale of spending and uncertain about the returns, but they are not calling for a halt. What they are calling for is accountability — which puts the pressure squarely on earnings calls, shareholder resolutions, and regulatory frameworks rather than on voluntary corporate restraint.

Conclusion

The April 2026 earnings season has become a public referendum on whether big tech's AI gamble serves anyone beyond its own shareholders. The verdict so far: conditional patience, with a clear scoreboard. People want jobs and cheaper services. They are not getting obvious early proof of either — and they know it.

Watch three developments in the months ahead. First, Q2 and Q3 earnings calls will test whether Meta's ad-targeting AI and Amazon's AWS AI services begin producing the near-term ROI that the 'necessary oversight' majority is demanding. Any quarter where capex growth outpaces revenue growth again will widen the credibility gap. Second, legislative and shareholder pressure on AI accountability — from bias disclosure resolutions to retraining fund proposals — will accelerate if displacement news like Oracle's layoffs continues to dominate the headlines. Third, GPU rental price collapse below $1 per hour is a real signal that compute costs are falling fast; if that deflation reaches consumer AI pricing, the 'too early to judge' camp's expectation of cost relief could be validated sooner than the bears expect.

The companies that close this credibility gap fastest will be the ones that stop describing AI's potential and start showing receipts — in jobs created, costs lowered, and benefits that reach people outside the enterprise software stack.