Redistributed

When AI cuts costs, where does the money go?

Not to workers. Not to lower prices. Increasingly, it goes to shareholders through stock buybacks — corporations purchasing their own shares to boost stock prices and enrich those who already own the most.

This is not an accident. It is not an economic law. It is a policy choice, created by deregulation in 1982 and reinforced by corporate governance norms that prioritize shareholders above all else. As AI accelerates productivity gains, the buyback question becomes a redistribution question: who captures the value that AI creates?

What Buybacks Actually Are

A stock buyback is simple: a company uses cash to purchase its own shares on the open market. This reduces the number of shares outstanding, mechanically increasing earnings per share and typically boosting the stock price.

Before 1982, this was illegal. The SEC considered it a form of stock manipulation — which it is. Companies could use buybacks to inflate prices, mislead investors, and enrich executives whose compensation was tied to stock performance.

Then came SEC Rule 10b-18, which created a "safe harbor" for buybacks. Companies could repurchase shares without fear of manipulation charges, provided they followed certain guidelines (not buying more than 25% of average daily volume, not buying at the open or close, using a single broker).

The floodgates opened. In 1982, US companies spent $5 billion on buybacks. By 2023, the figure exceeded $800 billion annually.

The Scale of Tech Buybacks

The largest practitioners are the companies best positioned to benefit from AI:

Apple: $90 billion authorized in 2024, bringing total buybacks since 2012 to over $700 billion. Apple has repurchased more stock than most companies are worth.

Alphabet (Google): $70 billion authorized in 2024. Google's AI integration into Search, Cloud, and advertising is driving revenue — and those revenues flow to buybacks.

Microsoft: $60 billion program announced in 2021. Microsoft's Azure AI services and Copilot products represent significant new revenue streams.

Meta: $50 billion authorized in 2024, after cutting 20,000 jobs in 2022-2023. The "Year of Efficiency" meant fewer workers and more buybacks.

Nvidia: While investing heavily in capacity, Nvidia has also repurchased billions in shares as AI chip demand soared.

These are not struggling companies returning capital because they lack investment opportunities. They are the most profitable firms in history, generating unprecedented cash flows — and choosing to direct those flows to shareholders rather than workers, communities, or long-term investment.

The Lazonick Critique

William Lazonick, an economist at the University of Massachusetts, has spent decades documenting what he calls "profits without prosperity." His core argument: buybacks represent a fundamental shift in how corporations operate (Lazonick, 2014).

In the postwar era, corporations followed a "retain and reinvest" model. Profits were reinvested in R&D, worker training, and capacity expansion. Wages rose with productivity. Corporations were engines of broadly shared prosperity.

Starting in the 1980s, this shifted to "downsize and distribute." Profits flow to shareholders through buybacks and dividends. Wages stagnate while stock prices soar. Corporations extract value rather than create it.

The numbers bear this out. From 2003 to 2012, the 449 companies in the S&P 500 used 54% of earnings for buybacks and 37% for dividends — leaving just 9% for reinvestment (Lazonick, 2014). The pattern has only intensified since.

Lazonick identifies three mechanisms driving this shift:

Executive compensation tied to stock price. CEOs benefit directly from buybacks. In 2022, median S&P 500 CEO compensation was $14.8 million, with stock awards comprising the majority.
Activist investor pressure. Hedge funds demand "returning capital to shareholders," treating any retained earnings as inefficiency.
Shareholder primacy ideology. The doctrine that corporations exist solely to maximize shareholder value, promoted by economists like Milton Friedman and Michael Jensen.

AI accelerates all three dynamics. AI-driven cost savings create surplus. Executive pay incentivizes directing that surplus to buybacks. Activist investors pressure companies to "return" AI productivity gains. The ideology justifies it all.

The AI Connection

How does AI feed the buyback machine?

Cost savings: AI automates tasks previously performed by workers. Customer service chatbots, automated code generation, AI-assisted document review — each reduces labor costs. Those savings become "efficiency gains" that flow to earnings, then to buybacks.

Meta's 2022-2023 layoffs were explicitly framed as efficiency measures. The company cut workers while reporting record profits. Those profits funded record buybacks.

Revenue growth: AI-powered products (cloud services, advertising optimization, AI assistants) generate new revenue streams. That revenue, after modest reinvestment, flows to shareholders.

Valuation increases: AI hype drives stock prices. Higher valuations make buybacks more attractive — executives can sell their own shares at inflated prices, funded by the company's cash.

The connection is structural, not coincidental. Erik Brynjolfsson's work documents the "productivity paradox" — productivity gains that don't translate to wage growth (Brynjolfsson & McAfee, 2014). Buybacks are a mechanism for that paradox: productivity gains flow to capital, not labor.

What Else Could the Money Do?

Buybacks are a choice. The same cash could be directed elsewhere:

Wages: Apple's $90 billion 2024 buyback authorization could fund a $50,000 raise for every one of its 161,000 employees — every year for more than a decade. Apple chose not to.

Prices: Instead of returning cash to shareholders, companies could lower prices for consumers. Amazon's relentless price competition was once celebrated as consumer-friendly redistribution. That era is over — Amazon now prioritizes profitability and buybacks.

R&D: Buyback critics argue companies underinvest in long-term research. Yet the data is mixed — tech companies do spend heavily on R&D. The question is whether even more investment would produce better outcomes than buybacks. For frontier AI safety research, the answer is almost certainly yes.

Reserves: Companies could retain cash for economic downturns, avoiding layoffs during recessions. Instead, many that received pandemic bailouts used the funds for buybacks (or would have, before restrictions were imposed).

Taxes: Higher corporate taxes would redirect some surplus to public purposes. The 2022 Inflation Reduction Act imposed a 1% excise tax on buybacks — a modest step that generated approximately $8 billion in 2023.

Worker ownership: Employee stock ownership plans (ESOPs) and equity compensation could distribute gains more broadly. Yet equity compensation is concentrated among executives and high-skill workers, not the median employee.

The point is not that buybacks should be zero. Some capital return makes sense for mature companies. The point is that current levels represent a policy choice to concentrate AI gains among shareholders.

Who Benefits: Stock Ownership in America

When companies buy back stock, who captures the value?

Stock ownership in America is extraordinarily concentrated. According to Federal Reserve data:

The top 1% of households own 54% of all stocks
The top 10% own 89%
The bottom 50% own 1%

This concentration has increased over time. In 1989, the top 10% owned 82% of stocks. The rise of 401(k) plans and stock-based compensation has not meaningfully democratized ownership.

When Apple spends $90 billion on buybacks, roughly $49 billion flows to the top 1% and $80 billion to the top 10%. The bottom half of Americans receive less than $1 billion.

Pension funds and retirement accounts provide some indirect exposure. But even here, the wealthy benefit more — they have larger 401(k) balances, more company stock options, and more capacity to invest.

The redistribution math is stark: AI productivity gains, converted to buybacks, flow overwhelmingly to the already-wealthy. This is not trickle-down; it is trickle-up.

The Historical Anomaly

Buybacks at current scale are a historical anomaly. Before 1982, they were effectively prohibited. The SEC viewed them as manipulation because they are manipulation — using corporate treasury to inflate stock prices.

The shift reflects broader changes in corporate governance ideology. In 1970, Milton Friedman published "The Social Responsibility of Business Is to Increase Its Profits," arguing corporations owe nothing to workers, communities, or society beyond shareholder returns.

In 1976, Michael Jensen and William Meckling published "Theory of the Firm," introducing principal-agent theory and arguing managers should be aligned with shareholders through stock-based compensation.

By 1982, these ideas had captured the SEC. Rule 10b-18 was not debated in Congress. It was an administrative decision by Reagan-appointed commissioners.

This history matters because it demonstrates contingency. Buybacks are not natural or inevitable. They are the product of specific policy choices that can be revised.

Policy Responses

Several policy approaches could constrain buybacks or redirect their proceeds:

Buyback taxes: The 2022 Inflation Reduction Act imposed a 1% excise tax on buybacks. This was partly intended to offset the revenue loss from clean energy tax credits, but it also represents a modest disincentive. Some proposals would increase this to 4% or higher.

Buyback bans for subsidized companies: Companies receiving federal subsidies, tax breaks, or government contracts could be prohibited from buybacks. The CARES Act imposed temporary buyback restrictions on companies receiving pandemic aid.

Enhanced disclosure: SEC rules could require more detailed disclosure of buyback timing, executive trading, and the relationship between buybacks and executive compensation. Greater transparency might constrain the most egregious practices.

Executive compensation reform: Limiting the tax deductibility of stock-based compensation, requiring longer holding periods, or capping the ratio of executive to median worker pay could reduce the pressure for buybacks.

Stakeholder governance: German-style codetermination, which requires worker representation on corporate boards, gives labor a voice in capital allocation decisions. Workers rarely vote for buybacks over wages. This is not theoretical — Germany's codetermination system has operated successfully for decades, producing firms that are globally competitive while distributing returns more broadly than their American counterparts.

Antitrust enforcement: Breaking up dominant tech companies would increase competition, potentially directing surplus to lower prices rather than buybacks. This addresses concentration at the source.

Senator Elizabeth Warren has proposed "accountable capitalism" legislation requiring large corporations to obtain federal charters, seat worker representatives on boards, and consider stakeholder interests (Warren, 2018). None of this has passed.

The Redistribution Question

Mainstream economics treats buybacks as neutral — simply a return of capital to its rightful owners. If shareholders don't want the money, they can reinvest it elsewhere. Markets are efficient; this must be optimal.

This framing ignores several realities:

Shareholders are not monolithic. Short-term hedge funds demanding buybacks have different interests than pension funds seeking long-term growth. Buybacks favor the former.

Executive interests are misaligned. CEO compensation tied to stock price creates pressure for buybacks regardless of long-term value. Executives selling into buybacks are extracting value, not optimizing it.

Employees have stakes too. Workers invest human capital in firms. They bear risks. They contribute to productivity. The doctrine that only shareholders deserve returns is ideology, not economics.

Public investment underpins private gains. As Mariana Mazzucato documents, much AI capability rests on decades of public investment — DARPA, NSF, public universities (Mazzucato, 2013). When companies monetize publicly funded research through buybacks, they privatize public investment.

Power matters. The shift to buybacks was not democratically decided. It was implemented by regulators, promoted by economists, and embraced by executives. Workers, communities, and the public were not consulted.

Buybacks are not neutral. They are a mechanism for concentrating the gains from AI — and from technology more broadly — among those who already have the most.

What AI Redistribution Would Look Like

An alternative path is imaginable:

AI cost savings → worker raises. Productivity gains could translate to higher wages, not just higher earnings per share. This would require countervailing power — unions, regulation, or changed corporate norms.

AI cost savings → lower prices. If AI reduces costs, consumers could benefit through lower prices rather than shareholders through higher profits. Competition could drive this, but concentrated markets prevent it.

AI cost savings → public benefit. Higher corporate taxes on AI-driven profits could fund public goods — education, infrastructure, social insurance. The 1% buyback tax is a weak version of this.

AI cost savings → worker ownership. Broad-based equity compensation, profit-sharing, or ESOPs could distribute gains more widely within firms. Some companies do this; most do not. Cooperative ownership models offer a more structural alternative.

AI cost savings → reserves and resilience. Rather than distributing every dollar to shareholders, companies could maintain reserves that protect workers during downturns. This would require different incentives than current ones.

None of these alternatives is automatic. Each requires policy, power, or changed norms. The default — AI gains flowing to buybacks flowing to the wealthy — will continue absent intervention.

The Bottom Line

When AI creates value, the question is not whether value is created but who captures it. Right now, the answer is: shareholders, through buybacks.

This is not inevitable. Buybacks were illegal until 1982. Other countries constrain them more aggressively. Alternative uses for corporate cash exist. Different corporate governance models are possible.

The choice to channel AI gains to buybacks is exactly that — a choice. Made by executives, enabled by regulators, justified by ideology. Choices can be changed.

The redistribution question is not abstract. It is $800 billion per year. It is the gap between productivity growth and wage growth. It is the concentration of wealth in the top 1%. It is the question of who benefits when AI transforms the economy.

Buybacks are how AI gains become shareholder gains. Policy created this path. Policy can redirect it.

Sources

Brynjolfsson, Erik and Andrew McAfee. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W.W. Norton, 2014. Publisher
Federal Reserve Board. "Distribution of Household Wealth in the U.S. since 1989." Distributional Financial Accounts, 2024. Federal Reserve
Friedman, Milton. "The Social Responsibility of Business Is to Increase Its Profits." The New York Times Magazine, September 13, 1970. NYT Archive
Jensen, Michael C. and William H. Meckling. "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics 3, no. 4 (1976): 305-360. ScienceDirect
Lazonick, William. "Profits Without Prosperity." Harvard Business Review, September 2014. HBR
Lazonick, William. "Stock Buybacks: From Retain-and-Reinvest to Downsize-and-Distribute." Center for Effective Public Management, Brookings Institution, 2015. Brookings
Mazzucato, Mariana. The Entrepreneurial State: Debunking Public vs. Private Sector Myths. Anthem Press, 2013. Publisher
Securities and Exchange Commission. "Rule 10b-18." 17 CFR 240.10b-18. SEC
Warren, Elizabeth. "Accountable Capitalism Act." S.3348, 115th Congress, 2018. Congress.gov

General

Stock Buybacks and AI