The knowledge economy is often described as a democratizing force, one in which the primary inputs to production are ideas and expertise rather than land or capital, and in which the barriers to entry are therefore, in principle, low. A laborer displaced from a factory by automation does not lose the mental and social capacities that knowledge-intensive production requires. Yet the gains from the knowledge economy flow to a remarkably narrow group: the founders and early investors of a small number of firms, concentrated in a handful of cities, drawn from a thin slice of the educational and social elite. This is not a paradox to be explained away by appealing to differences in talent or effort. It is an institutional fact—the product of specific arrangements governing property, finance, and organizational form that could be otherwise. An important question of 21st-century political economy then becomes whether those arrangements can be changed, and what would have to replace them. This paper argues that the answer turns on what we do with venture capital (VC), not as a reformable institution but as a diagnostic one. VC’s failures illuminate precisely the institutional functions that alternatives must perform.
Start with what the knowledge economy actually requires to generalize its most productive forms of work. In the sectors where it operates at its best—software, biotechnology, and advanced manufacturing, for instance—production is organized around high-agency teams that experiment rapidly, share tacit knowledge freely, and iterate in ways that blend conception and execution. This mode of production is not only more innovative but also more humane, treating workers as participants in a common project rather than interchangeable inputs to a predefined process. The problem is that it remains confined to a vanguard (Unger 50). The rest of the economy continues to operate under older arrangements: routineized work, hierarchical control, and a distribution of gains heavily tilted toward capital. And the gap between the vanguard and the rest is not primarily technological. The tools of knowledge-intensive production are increasingly cheap and accessible. Rather, it is institutional: who gets to use those tools, under what terms of ownership, and with access to what kinds of financing.
Venture capital sits at the center of this institutional failure. Its defenders argue that it solves a genuine market problem: standard debt finance cannot fund firms whose assets are intangible and whose revenues are speculative, so equity investment with active mentorship fills a necessary gap. This is not incorrect. But the way VC fills that gap is inseparable from a financial logic that systematically produces the concentration it claims only to reflect. A venture fund must generate outsized returns from a small number of winners to compensate for the majority of investments that fail. This means it selects for firms that can achieve rapid scale in large markets—not firms that deepen capability in communities excluded from the vanguard, not firms that create lasting value in forms that resist winner-take-all dynamics, not firms whose success would be incentivized by and measured in the broad distribution of productive capacity over solely the appreciation of equity. The geography and social composition of venture funding are also not accidental features of a basically sound institution, one where capital is overwhelmingly concentrated in a few cities, mediated by networks of elite educational and professional connections. They are expressions of its financial logic. VC not only fails to generalize the knowledge economy, it is also organized around a principle that is structurally incompatible with generalization: maximizing asymmetric returns.
The question is not whether to have institutions that finance knowledge production, but what principle should govern them. This diagnosis generates a specific set of requirements for alternative institutions. They must finance knowledge-intensive production without the constraint of winner-take-all return structures. They must distribute access along dimensions of geography, class, and educational background where VC has historically concentrated it. And they must be compatible with ownership arrangements that give workers a genuine stake in the surplus generated by knowledge work, rather than concentrating equity among founders and investors. No single institutional form meets all of these requirements, but two approaches, taken together, address them in complementary ways.
The first is commons-based peer production of the kind that Benkler analyzes in the context of open-source software and networked collaboration. Such collaboration reveals insights about organizational form. When tasks are modular, contributions are voluntary, and outputs are held as commons available to all, large-scale knowledge production can proceed without firms, without the price mechanism, and without the disciplinary logic of investor-backed growth. Linux, Wikipedia, and the broader open-source ecosystem represents a proof of concept for a different relationship between knowledge production and institutional inclusion. Contributing to the knowledge economy does not require submitting to VC’s gatekeeping. And the gains from this mode of production accumulate as infrastructure rather than equity, available to anyone with the capability to use it. Meaningfully, this is the functional equivalent for the knowledge economy of the widely distributed productive property that underpinned democratic self-governance in the antebellum North: a commons of productive capacity on which a broad range of participants can draw.
The second approach addresses the ownership problem directly through the redesign of property rights. The current intellectual property regime, whereby knowledge-intensive firms accumulate patents, trade secrets, and data assets as moats against competition, is not a neutral background condition but an active institutional choice that amplifies the concentration VC produces. Posner and Weyl’s diagnosis cuts to the root of this: “anemic employment and low productivity growth result from institutional failure rather than changes in technology” (Posner and Weyl 255). An alternative institutional property regime change would make it costly to hold productive assets at values higher than their use warrants, forcing their deployment or relinquishment. Applied to data and intellectual property, the primary inputs to AI systems, a similar system would significantly alter the competitive dynamics of the knowledge economy. Rather than entrenching the first movers who attract VC backing, it would create conditions for more free circulation of productive assets, enabling a wider range of firms and workers to build on them. The specific mechanism matters less than the institutional direction it represents: property arrangements designed not to reward the accumulation of intangible assets but to maximize their productive deployment across the economy.
The urgency of this institutional question has been raised considerably by the emergence of AI as a general-purpose technology. The distributional stakes are unusually high now—the decisions being made now about who owns AI infrastructure, who can access AI tools, and how AI-generated surplus will be distributed are likely to be path-dependent in near-irreversible ways. There exists an opportunity to take the “high road” path, where technology increases broad productivity, instead of the “low road” path, where capitalists “seek to disembody knowledge and locate it in technologically embedded knowledge that capitalists can fully own as property” (Benkler, “Institutional” 5). A world where AI development is organized primarily through VC-backed platforms reproduces the concentration characterized by the knowledge economy and adds to it the particular leverage that comes from owning infrastructure on which the rest of the economy depends. A world where AI development is organized through commons, public investment, and cooperative ownership offers a different path; one in which the productivity gains from AI are available as a shared resource. The choice between these paths is not primarily technical. It is institutional, and it is being made through the same kinds of political and economic decisions about property, finance, and organizational form.
The knowledge economy’s promise—that productive capacity could be more broadly distributed than in any prior economic era—remains unredeemed not because it was false but because the institutions built around it were designed for other purposes. Venture capital was designed to and does generate asymmetric financial returns. Intellectual property law was designed to and does reward and protect innovation by those with the resources to pursue it. Neither was designed to generalize the vanguard, distribute agency, or give ordinary workers a stake in the knowledge economy’s surplus. The 21st-century institutional agenda is to build the alternatives that can do those things: not as supplements to the existing arrangements, but as genuine competitors to them, organized around a different logic and oriented toward a different set of ends.
References Benkler, Yochai. “Law, Innovation, and Collaboration in Networked Economy and Society.” Annual Review of Law and Social Science, vol. 13, 2017, pp. 231–250. Benkler, Yochai. “The Institutional Political Economy of AI: Technology and Class in Capitalism.” Unpublished manuscript, 2025. Lothian, Tamara. Law and the Wealth of Nations: Finance, Prosperity, and Democracy. Columbia University Press, 2017. Posner, Eric A., and E. Glen Weyl. Radical Markets: Uprooting Capitalism and Democracy for a Just Society. Princeton University Press, 2018. Unger, Roberto Mangabeira. The Knowledge Economy. Verso, 2019.