A growing chorus of stakeholders is calling for “broad-based ownership” of AI. But what does that mean in practice? And what would move incumbent market powers to share the upside?
This has been top of mind at the Predistribution Initiative, the nonprofit I lead that works with investors to build broad-based prosperity and reduce economic inequality, since at least January, when I listened to a Wall Street Journal interview with Anthropic CEO Dario Amodei.
Amodei projected 5-10% GDP growth per annum with 10-20% unemployment. That didn’t seem possible. Where would the aggregate demand required for economic activity come from if so many people were unemployed? Consumer spending, roughly 70% of US GDP, according to the Bureau of Economic Analysis, would be decimated. Workers who would be displaced carry high marginal propensities to consume — they spend more when their disposable income rises, unlike the wealthier beneficiaries of AI gains.
At Predistribution Initiative, we began to model (with Claude!) macro-financial scenarios with higher unemployment. For simplicity, we scoped our work to focus on the US, while recognizing that future analysis would benefit from integration of global dynamics. Then, in February, Citrini Research came out with its Ghost GDP piece, “a thought Exercise in financial history, from the future.”
Citrini’s 2028 scenario captured the attention of markets, but then was largely dismissed under the premise that, over time, the number of jobs actually has risen with the advancement of technology. Maybe so. But the quality of jobs has been deteriorating. So we added a “light” scenario, where unemployment doesn’t rise, but returns to labor continue to decline in line with historical trends as technologies have advanced.
Part I of the Predistribution AI Lab discussion paper series, Modeling Ghost GDP, should command the attention of investors, policymakers, and regulators alike. Declining returns to labor and the undervaluation of human and social capital are macro-financial risks to markets, pension funds, insurers, and diversified portfolios. We trace the transmission channels of inequality risk through strained mortgage markets, corporate debt defaults, corporate-bond mark-to-market losses, equity-market declines, pension and insurance impairments, an accelerated Social Security drawdown, and a fiscal “doom loop.”
Total economy-wide value at risk ranges from roughly $15–18 trillion in the light scenario (unemployment does not increase, but returns to labor erode further) to $62–72 trillion in the aggressive one (20% unemployment, the “high-end” of Amodei’s prediction). These figures are not forecasts. Instead, they estimate the demand-side economic value at risk if labor income losses of the modeled magnitude are not offset. The model intentionally excludes estimates of potential AI-driven productivity and efficiency gains to firms, isolating the effects of reduced labor income and making that transmission channel visible on its own terms.
Crucially, economic inequality is already on an accelerating trajectory. Whether AI produces mass unemployment or primarily accelerates the long-running shift from labor income to capital income, the demand-side pressures identified in the analysis remain. Even in the light scenario—no net job losses but continued shift of workers into precarious roles—the “fissured workplace” reduces GDP by about 1%. The unprecedented speed of this transition magnifies systemic stress.
Macro risks
It is important to consider that AI is being introduced to an already imbalanced economy. Corporate profits’ share of gross domestic income has climbed, from 7% in 1980 to 12.1%. The labor share of US nonfarm business output fell to 54.1% in the first quarter of 2026—its lowest level since the Bureau of Labor Statistics began recording the figure in 1947, when it stood at 65.8%.
The top 10% of households now hold approximately 90% of equities and mutual funds; the bottom 50% hold about 1%. Cash is depreciating while asset prices rise.
BlackRock CEO Larry Fink notes in his latest annual letter that, since 1989, “a dollar in the US stock market has grown more than 15 times the value of a dollar tied to median wages.” And only about half of Americans earn a living wage.
Returns to capital may be higher in the near term, but they eventually fall as aggregate demand declines and the economy weakens, across the scenarios we modeled.
Decentralized power
Broad based ownership can help narrow the gap, but the devil is in the details of the design. There is a lot of noise around broadening ownership of AI, so it is worth being clear upfront about what is different here.
Much of the current conversation coalesces around a few familiar ideas—universal basic income, a universal basic capital fund, or a centralized sovereign wealth fund. Each has merit, but each leaves concentrated power unchecked and most people dependent on modest, top-down payouts. Households would not hold assets directly, have any involvement in governance, and the payouts wouldn’t consider who took risk and created value in the production process.
How do we instead build direct household wealth, compensating workers (formal, informal, and displaced), communities, and content creators for the risk they take (or have taken) and value that they create (or have created) in economic productivity? How do we center dignity, agency, and respect and design the decentralized approach needed to sustain dynamic markets and democracy? To do otherwise weakens democracy, and capitalism itself.
Corporate governance reform is another key piece of the puzzle. Market-driven wealth has become so concentrated in part because companies are typically governed by a privileged few. Boards of large companies who affect public health, safety, and security are typically comprised of an executive class. Board director training and incentives are largely designed with the interests of investors top of mind, rather than those of other corporate stakeholders who are critical to a company’s success, including workers and communities.
No wonder that trust in institutions – both public and private – is eroding. Surveys from Gallup, Pew, Just Capital, and the Edelman Trust Barometer consistently show that majorities, across the political spectrum, believe the system is rigged. Federal lobbying topped $5 billion for the first time in 2025, dominated by a handful of players (and billionaires who have benefitted comprised 19% of all federal election contributions in 2024).
Equity transition stakes
The second paper in the series, Beyond Ghost GDP, deepens the analysis of how we arrived at our conclusions in Part I and introduces proposed solutions.
The Predistribution Initiative has long argued that workers and communities that host infrastructure and natural resource projects provide essential human and social capital that businesses depend on, alongside executives who provide leadership and investors who provide financial capital. Like executives and investors, workers and communities should participate in and be incentivized with equity upside in addition to living wages.
Top CEO compensation rose by over 1,000% since 1978, compared to 20% to 30% for a typical worker, according to Economic Policy Institute. That’s in part because executive pay was deliberately equity-linked to align incentives with investors, while workers received only cash. This needs to change, particularly through transitions, whether in climate or technology.
Aligning interests through broadening equity-linked compensation can be extended beyond just corporate executives to communities that host infrastructure and to content creators whose work trains AI, and to informal as well as formal workers. The reforms are not novel. Employee ownership models have existed for decades. They demonstrate how broad-based ownership is compatible with operational success.
Part III offers a case study, focused on displaced rideshare drivers as autonomous vehicles are introduced. We propose Driver Equity Transition Stakes, or DETS, under which autonomous vehicle operators contribute a share of gross fare revenue, attached to operating permits. The funds would go into a driver-governed trust split across current income (Pillar 1), platform equity participation (Pillar 2), and diversified investment accounts (Pillar 3). The three-pillar model deliberately includes informal workers.
The DETS model can be adapted for various contexts including for workers who remain employed. Notably, a 15% contribution and capped equity participation of the DETS trust leaves plenty of upside to investors. This offers cities a coalition-building tool (i.e., several large metros adopting at once) and companies a viable deal.
Predistribution Initiative’s AI Lab, a peer learning community, will refine this analysis and solutions alongside companies, investors, and their stakeholders. The new lab will seek to support pilots and implementation with a community of specialists; advise on the architecture for broad-based equity-linked compensation and broadened corporate governance participation; and support companies in advancing living wages and incomes and constructive relations with organized labor.
We urge the impact investing community to reserve some time, attention, and capital for these major economic and societal risks. Investors can encourage these reforms with their investees and consider investments which already center them.
Impact investing has rightly focused on small, entrepreneurial companies. But the largest companies—the six biggest US tech firms alone exceed $20 trillion in combined market value – employ and touch enormous numbers of people and produce headwinds in the shaping of politics, market concentration, and the cultural zeitgeist.
We welcome readers to continue the conversation. Keep an eye out for updates via the Predistribution Initiative newsletter, and join the ImpactAlpha, Confluence Philanthropy, and Predistribution Initiative interactive dialogue on this topic in an upcoming Call.
Delilah Rothenberg is a Co-Founder and the Executive Director The Predistribution Initiative.
