The Jetsons Don't Need the Flintstones Anymore

I. The AI Bet You're Already In

Here is a number that should concern every American with a 401(k): 41 AI-related stocks — representing approximately 8 percent of the S&P 500's constituents — now account for 47 percent of the index's total market capitalization, according to J.P. Morgan. The Magnificent Seven alone represent over 35 percent. The top 10 stocks command 41 percent of total weight but generate only about 32 percent of earnings.

This is not an investment thesis you chose. If you hold an S&P 500 index fund — and more than half of American retirement savings are now in passive vehicles — more than $40 of every $100 you invest flows into just 10 companies. BlackRock, Vanguard, and State Street collectively manage approximately $25 trillion in assets and constitute the largest shareholder in 88 percent of S&P 500 companies. Your retirement is a directional bet on AI monetization, placed without your explicit consent, through the passive investing infrastructure that was sold to you as diversification.

Now read the other side of the ledger. Hyperscaler capital expenditure — the spending by Amazon, Alphabet, Microsoft, Meta, and Oracle on AI infrastructure — is projected to reach $660–690 billion in 2026, nearly doubling 2025 levels. Sequoia Capital's David Cahn updated his widely cited analysis: what began as "AI's $200 Billion Question" is now "AI's $600 Billion Question." JPMorgan has calculated the arithmetic: to deliver a mere 10 percent return on the AI buildout requires $650 billion in annual revenue, in perpetuity. Current AI revenue across all providers: roughly $50 billion. The gap is not a rounding error. It is a 13-to-1 mismatch between capital deployed and revenue generated.

Google is spending $180 billion in capex to generate perhaps $18 billion in AI revenue. The hyperscalers collectively hold over $400 billion in liquid cash and are expected to generate more than $600 billion in operating cash flow in 2026 — enough to self-fund the buildout for now. But as MUFG and Morgan Stanley have documented, aggregate capex, after buybacks and dividends, now exceeds projected free cash flow, necessitating external debt financing. UBS projects as much as $900 billion in new corporate debt in 2026. The technology sector, historically defined by cash-funded growth, is leveraging itself to finance an infrastructure bet whose revenue case requires a 13-fold increase in annual income.

Where have we seen this arithmetic before? JPMorgan's own analysts have noted the comparison: the $650 billion annual revenue requirement is equivalent to charging $34.72 per month to every iPhone user on Earth, in perpetuity. Or $180 per month to every Netflix subscriber. This is the financial equivalent of a population-scale extraction — a recurring fee imposed on the installed base of digital consumers to justify infrastructure that was built before the revenue model existed.

And here is the distributional point that the AI analysts do not include in their models: who benefits if the bet pays off, and who pays if it doesn't?

The Federal Reserve's Distributional Financial Accounts answer the question with brutal clarity. The top 10 percent of American households own 87 percent of all corporate equities and mutual fund shares. The top 1 percent alone hold 50 percent of stocks, worth $25.6 trillion. The bottom 50 percent own 1 percent of stocks — roughly $540 billion, split across 165 million people.

If the AI bet pays off — if the revenue arrives, if the productivity gains materialize, if the $660 billion in annual capex produces the returns the market is pricing — the gains accrue to the people who already hold the assets. The retirement savers who were passively enrolled will see modest portfolio appreciation. The asset-holding class will see generational wealth creation on a scale that makes the dot-com boom look like a warmup.

If the bet fails — if the revenue doesn't arrive, if the $900 billion in new tech-sector debt reprices, if the 13-to-1 gap persists — the losses are socialized through pension funds, retirement accounts, and the index-fund infrastructure that channeled trillions of passive savings into a concentrated AI wager. The asset-holding class has already diversified. The passive investor cannot.

Either way, the Jetsons win. The question is only how much the Flintstones lose.

II. The Defense Dividend

The AI concentration is one wealth transfer. The defense surge is another — and it operates through a different pipe.

Rheinmetall, the German defense conglomerate, has delivered a return of approximately 1,800 percent over five years. Its stock price has gone from roughly €70 in early 2020 to above €1,500 by late 2025, with analyst consensus targeting €2,050–€2,250 in 2026. Its order backlog expanded 36 percent in 2025 to a record €63.8 billion, with management forecasting that backlog will more than double to €135 billion by the end of 2026. The company projects revenue growth of 40–45 percent in 2026 alone, to between €14 and €14.5 billion, and has set a target of €50 billion in annual sales by 2030 — a fivefold increase from 2024.

The U.S. defense sector tells a parallel story. The iShares U.S. Aerospace & Defense ETF (ITA) returned 48.6 percent in 2025 — nearly quadruple its ten-year annualized return of 10.3 percent. The SPDR S&P Aerospace & Defense ETF (XAR) returned 46.1 percent. The global aerospace and defense sector projects revenue growth from $800 billion in 2025 to $1.25 trillion by the mid-2030s.

The macro driver is geopolitical rearmament. European Union defense expenditure reached an estimated €381 billion in 2025 — up 11 percent year-over-year and 63 percent from 2020. The June 2025 NATO summit set a new defense spending target of 3.5 percent of GDP by 2035, more than tripling the 2014 Wales pledge of 2 percent. The EU's ReArm Europe plan targets €800 billion in defense spending through 2029, including €150 billion in EU-backed loans through the new Security Action for Europe (SAFE) instrument. Germany has committed to raising defense spending from 1.6 percent to nearly 3.5 percent of GDP by 2035.

This is an €800 billion wealth transfer to defense-sector shareholders, capitalized overnight into stock prices that were accessible only to those who already held positions. The retail investor who bought Rheinmetall at €70 in 2020 saw an 1,800 percent return. The retail investor who bought after the Ukraine invasion, at €200, still saw a 650 percent return. The retail investor hearing about it now, at €1,500, is buying the aftermath.

Defense spending is politically coded as security. It is economically coded as fiscal stimulus. What it is distributionally is a transfer from taxpayers to asset holders. The €800 billion in planned European defense spending will be funded by government borrowing, serviced by taxpayers, and capitalized into equity gains for shareholders of Rheinmetall, BAE Systems, Leonardo, Thales, Lockheed Martin, and RTX. The returns accrue to the people who already own defense stocks. The costs are borne by populations whose social spending will be displaced by the rearmament budgets.

The ITA returned 48.6 percent in one year. European social spending did not grow by 48.6 percent. The difference is the transfer.

III. The Ladder They Pulled Up

The third transfer is the most consequential, because it forecloses the mechanism through which prior generations built wealth: entry-level employment.

Postings for entry-level jobs in the United States plunged 35 percent from January 2023 to June 2025, according to data compiled by Rezi.ai from major job platforms. In the United Kingdom, tech graduate roles fell 46 percent in 2024, with projections for a further 53 percent drop by 2026. Software developers aged 22–25 have seen employment decline nearly 20 percent from the late-2022 peak, according to the Stanford Digital Economy Lab.

The mechanism is not mass layoffs. It is hiring avoidance. About 21 percent of companies have stopped hiring entry-level employees due to AI capabilities. Half expect to stop hiring entry-level workers by 2027. One in three expects entry-level roles to be eliminated entirely at their organizations by the end of 2026. Gartner projects that 40 percent of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5 percent in 2025.

The companies making these decisions are not obscure. Klarna reduced its headcount from 5,527 to under 3,500 — a 40 percent cut — with CEO Sebastian Siemiatkowski projecting the company will operate with fewer than 2,000 employees by 2030. Salesforce cut approximately 4,000 customer support roles after deploying AI agents that now handle 50 percent of customer interactions, then announced it would not hire new software engineers in 2025. Duolingo terminated 10 percent of its contractor workforce and declared itself "AI-first," with CEO Luis von Ahn's internal memo describing a future in which AI handles content creation, performance reviews, and hiring decisions.

The labor economists describe this as "task automation" rather than "job replacement." The distinction matters technically. It is irrelevant distributionally. What the 22-year-old experiences is not the automation of a task. It is the disappearance of the first rung of the ladder.

In a functioning labor market, "vacancy chains" allow mobility: a senior employee leaves, a mid-level employee moves up, a junior employee is hired to fill the gap. AI disrupts this chain by automating the bottom link — the tasks that constitute the training ground for early-career workers. The grunt work that junior analysts, junior developers, junior copywriters, and junior consultants performed was not just labor. It was apprenticeship. The automation of that labor is the elimination of the apprenticeship.

Employers' rating of the job market for college graduates is now at its most pessimistic level since 2020. The National Association of Colleges and Employers projects a marginal 1.6 percent increase in hiring for the Class of 2026 — which, adjusted for the increasing number of graduates, represents a functional contraction in opportunity. The entry-level job market is not recovering at a slower pace. It is structurally contracting.

And here is the distributional lock: the people who lose access to entry-level positions are disproportionately from the bottom half of the wealth distribution — the cohort that holds 1 percent of equities and 2.5 percent of total household wealth. They cannot compensate for lost labor income with capital income, because they hold no capital. The entry-level job was their only mechanism for wealth accumulation. The ladder is being pulled up by the people who already climbed it.

IV. The Prisoner's Dilemma of the Professional Class

The three transfers — AI concentration, defense windfalls, entry-level elimination — do not administer themselves. They require a workforce of professionals who process each transaction within its institutional code, who are rewarded for code-compliance and penalized for cross-code observation. This is the professional class: not the Jetsons and not the Flintstones, but the layer in between that makes the machine run.

Consider the individual decisions:

The HR manager who approves the AI replacement of 12 customer service representatives is operating within a performance code. Her department's metrics — cost per interaction, resolution time, customer satisfaction scores — improve by 30 percent. The AI agent handles 50 percent of interactions. The business case is unambiguous. Her bonus depends on the metrics. She does not model the distributional consequences of 12 people losing their health insurance, their mortgage payments, their children's college savings trajectory. That is not her code.

The fund manager who overweights Rheinmetall in his European equity fund is operating within a return code. The stock is up 1,800 percent in five years. The order backlog has doubled. The geopolitical tailwind is durable. His fiduciary duty is to maximize risk-adjusted returns for his clients. He does not model the social spending that will be displaced by the €800 billion rearmament budget that produces his returns. That is not his code.

The trade lawyer who structures the transshipment arrangement through Vietnam — routing Chinese-manufactured components through Vietnamese assembly to avoid 46 percent tariffs — is operating within a legal code. The arrangement is compliant with rules of origin under current regulations. The 40 percent transshipment tariff applies only to goods "deemed transshipped," and the definition remains legally ambiguous. She is billing $1,200 per hour to navigate the ambiguity. She does not model the effect on the American worker whose job was supposed to be protected by the tariff her client is legally circumventing. That is not her code.

The management consultant at McKinsey or Deloitte who designs the "workforce transformation" — the euphemism for AI-driven headcount reduction — is operating within an efficiency code. McKinsey expects 40 percent of its revenue to come from AI and technology advisory, contributing to an estimated $16 billion in annual revenue. The global AI consulting market reached $11 billion in 2025 and is projected to hit $14 billion in 2026. The consultant is selling the implementation plan for the displacement that the HR manager will execute. Her deck calls it "augmentation." The org chart calls it a reduction. The distributional consequence is identical regardless of what the slide says.

The index fund product manager at BlackRock or Vanguard who maintains the S&P 500 tracking fund is operating within a fiduciary code. The fund must track the index. The index is market-cap weighted. Market-cap weighting means that as AI stocks appreciate, the fund automatically buys more of them. This is not a decision. It is a mechanism. But the mechanism channels $40 of every $100 of passive savings into 10 companies, creating the concentration that makes every retirement saver an involuntary participant in the AI bet. The product manager does not decide the concentration. The code decides. The product manager services the code.

Each of these professionals is making a rational, defensible, code-compliant micro-decision. No one is committing fraud. No one is breaking the law. No one is acting in bad faith. And the aggregate of their individually rational decisions is a generational wealth transfer from the bottom half to the top decile, administered through the institutional infrastructure that each professional maintains.

This is the prisoner's dilemma of the professional class. If any individual professional defects — if the HR manager refuses to approve the AI replacement, if the fund manager underweights defense stocks, if the trade lawyer declines the transshipment engagement, if the consultant refuses to design the "workforce transformation" — they are replaced by a professional who will. The code selects for compliance. The professional who sees across codes is not rewarded for the insight. They are penalized for the friction.

V. The Machine That Nobody Designed

Mark Blyth has spent a career documenting how ideas function as weapons in institutional change. In Great Transformations (2002), he showed that economic categories are not neutral descriptions of reality — they are deployments by actors who benefit from specific framings. In his 2025 book Inflation: A Guide for Winners and Losers, co-written with Nicolò Fraccaroli, he extended the analysis to the asymmetric distributional consequences of macroeconomic policy: inflation harms debtors and workers; the policy response benefits creditors and asset holders. The choice of which inflation to fight and which to tolerate is itself a distributional decision, disguised as technical expertise.

Blyth would recognize the code-compliant wealth transfer immediately. The separation between "AI policy," "defense policy," "trade policy," and "labor policy" is not a natural division of reality. It is an institutional architecture that determines who is qualified to speak, what data is relevant, and — critically — what connections are invisible. The HR manager operates in "workforce optimization." The fund manager operates in "asset allocation." The trade lawyer operates in "compliance." Each domain has its own conferences, its own journals, its own professional certifications. None rewards the observation that the aggregate of all three is a class divide.

The separation benefits specific actors. The AI company does not want the AI investment debate connected to the entry-level employment debate, because the connection reveals that the productivity gains are extracted from labor displacement. The defense contractor does not want the rearmament debate connected to the social spending debate, because the connection reveals the opportunity cost. The trade lawyer does not want the transshipment debate connected to the labor protection debate, because the connection reveals that tariff evasion undermines the stated purpose of tariffs. Each separation is a load-bearing structure in the architecture of denial.

Niklas Luhmann would describe the structural mechanism that makes the alibi work. Each professional operates within a functional subsystem — economy, law, education, politics — and each subsystem processes the world through its own binary code. The economic code processes everything through payment/non-payment. The legal code processes everything through legal/illegal. The HR code processes everything through performance metrics. The fund management code processes everything through risk-adjusted return.

"Every system uses its own distinction to observe the world," Luhmann wrote in Social Systems (1995). "The system cannot observe what it cannot observe. It cannot observe that it cannot observe this."

The HR manager cannot observe the distributional consequences of her decision because her code does not process distributional consequences. She processes cost-per-interaction. The fund manager cannot observe the social cost of the defense windfall because his code does not process social cost. He processes alpha. The trade lawyer cannot observe that her transshipment structure undermines labor protection because her code does not process labor protection. She processes compliance.

Each code is extraordinarily sophisticated within its own domain. And each code is structurally blind to everything outside it. The blindness is not a failure of intelligence or ethics. It is a feature of functional differentiation. The system is designed this way.

Jürgen Habermas, Luhmann's great intellectual rival, identified the force that keeps the codes locked. The "steering media" of modern society — money and power — have colonized what Habermas called the "lifeworld": the domain of shared meaning, communicative reason, and democratic deliberation where people could, in principle, evaluate the aggregate consequences of institutional decisions.

The colonization thesis describes exactly what happens to the professional class. The HR manager's performance review is a steering medium. The fund manager's bonus structure is a steering medium. The consultant's utilization rate is a steering medium. Each medium rewards code-compliant behavior and punishes deviation. The professional who raises distributional concerns in a workforce optimization meeting is not engaging in communicative reason. She is introducing noise into a system optimized for signal. The system ejects the noise.

As Peter Verovšek argued in Political Studies in 2023, applying Habermas's colonization thesis to post-2008 economic governance: "The whole program of subordinating the lifeworld to the imperatives of the market must be subjected to scrutiny." The scrutiny still has not come — because the professionals who could provide it are the same professionals whose compensation depends on not providing it.

The code-compliant wealth transfer is the machine that nobody designed but everybody services. Blyth names the political interest that maintains the categorical separations. Luhmann names the structural mechanism that locks each professional inside their code. Habermas names the force — money and power as steering media — that punishes any professional who tries to see across codes. Three diagnoses. One machine.

VI. The Passive Trap

There is a specific mechanism through which the code-compliant wealth transfer reaches the broadest population: the passive index fund.

More than $13 trillion in U.S. assets are now managed in passive index strategies. BlackRock, Vanguard, and Fidelity together manage approximately 50 percent of all fund assets in the United States. The Big Three constitute the largest shareholder in more than 40 percent of publicly traded U.S. firms and in 88 percent of the S&P 500. Vanguard alone manages nearly half of all passive assets under management.

The S&P 500 is market-cap weighted. This means the largest companies receive the largest allocation of new inflows. When AI stocks appreciate, the index mechanically buys more. When passive investors contribute to their 401(k)s, the money flows disproportionately into the 10 companies that already command 41 percent of the index. This creates a feedback loop: passive inflows support the largest stocks, which increases their market cap, which increases their index weight, which attracts more passive inflows.

As of March 2026, the top 20 stocks account for nearly 45 percent of the S&P 500's total weight — a level of concentration that surpasses the peak of the 2000 dot-com bubble. If the rumored IPOs of SpaceX and OpenAI proceed in 2026–2027, the concentration of the top 10 holdings could push toward 50 percent of the entire index.

The passive investor — the teacher saving for retirement, the nurse contributing to her pension, the truck driver with an employer-matched 401(k) — has been enrolled in this bet without choosing it. The investment advice they received was correct within its code: passive indexing outperforms active management over long time horizons, fees are lower, diversification reduces risk. Every piece of the advice was accurate. None of it disclosed that "diversification" now means 47 percent exposure to AI-related stocks, that the index has become a concentrated sector bet, or that the teacher's retirement savings are financing the AI infrastructure that is eliminating the entry-level jobs her students will graduate into.

The financial advisor who recommended the index fund was operating within his code. The code says: low fees, broad exposure, long time horizon. The code does not say: you are funding a $660 billion annual infrastructure bet with a 13-to-1 capex-to-revenue gap, and your retirement depends on whether that gap closes before you need the money.

This is not a conspiracy. It is a mechanism. The mechanism has no author. It has no intent. It has only a distributional outcome: the people who already hold concentrated positions in AI and defense stocks capture the upside. The people whose savings are passively channeled into the same stocks absorb the downside if the bet fails. The professional class that maintains the mechanism — the product managers, the financial advisors, the retirement plan administrators — is compensated for servicing the code, not for questioning it.

VII. The Question That Matters

Article 1 described the five-gauge feedback loop — the structural coupling between tariffs, deficits, Fed independence, immigration, and energy-compute that institutional specialization renders invisible. This article describes a different machine: not a feedback loop between policy instruments, but a wealth transfer administered through the aggregation of code-compliant professional decisions.

The two machines are connected. The tariff regime from Article 1 feeds the transshipment industry that employs the trade lawyers in this one. The defense spending triggered by geopolitical instability produces the equity returns captured by the fund managers. The AI infrastructure that consumes the energy in Article 1 eliminates the entry-level jobs in this one. The institutional blindness that prevents anyone from reading five gauges simultaneously is the same blindness that prevents anyone from reading five transfers simultaneously.

The code-compliant wealth transfer does not require malice. It requires only specialization. It requires only that each professional do their job well, within their code, without looking at the aggregate. It requires only the institutional architecture that Blyth identified as political, that Luhmann identified as structural, and that Habermas identified as colonized.

Here is the arithmetic of the aggregate:

The top 1 percent hold $55 trillion in wealth — equal to the bottom 90 percent combined. The top 10 percent hold 87 percent of equities. AI capex of $660 billion annually is financed by the capital markets that the top 10 percent dominate. The returns from that capex accrue to the shareholders. The labor displacement from that capex falls on the bottom 50 percent, who hold $60,000 in average wealth and 1 percent of equities. The defense windfall — €800 billion in planned European rearmament, 48.6 percent one-year returns in U.S. defense ETFs — accrues to defense-sector shareholders. The cost is borne by taxpayers whose social spending is displaced. The entry-level hiring collapse — 35 percent decline in postings, 20 percent decline in junior developer employment — forecloses the mechanism through which the bottom 50 percent historically accumulated wealth. The passive index fund ensures that retirement savers finance the displacement without choosing it.

The Jetsons do not need the Flintstones for labor anymore. AI provides the labor. The Jetsons do not need the Flintstones for capital anymore. The capital markets provide the capital. The Jetsons do not need the Flintstones for consumption anymore — not yet, but the revenue models are being restructured to extract recurring fees from the installed base rather than rely on mass-market purchasing power.

What the Jetsons need from the Flintstones is consent. And the professional class provides it — one code-compliant micro-decision at a time.

The five-gauge feedback loop from Article 1 will determine whether the machine destabilizes. The code-compliant wealth transfer from this article will determine who absorbs the damage when it does.

The boiler does not care which floor you live on. But the building is designed so that the penthouse has a fire escape and the basement does not.