brain/
← all mechanisms
medium convictionactive · updated 2026-05-12T00:00:00.000Z

Public R&D → private rent capture → state-capacity erosion → AI governance gap

Modern tech was seeded by government R&D, but the rewards were privatized as excess rents. Those rents now outbid the public sector for top researchers — hollowing the state capability needed to regulate, shape, or partner with the next wave of technology (now AI). Without that capability, AI's negative externalities compound while industrial policy stays reactive.

The chain
1
Public R&D (DARPA, NSF, NASA, public universities) created the foundational technologies of the modern stack — Internet, GPS, touchscreen, the algorithms underlying AI.
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "everything in our smartphones that make them smart are not stupid. Internet, gps, touchscreen, Siri and so on were government finance"
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "the investments that went into what we call today artificial intelligence, including LLM models, language models, speech recognition, dates decades and like so many other areas that was led by government"
2
Private firms then capture the rewards from those technologies as excess economic rents — risk socialized, value privatized.
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "we privatized all the rewards from this massive social and collectively created value in this area. They have so much money, right? Trillions. Not billions, trillions"
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "we end up socializing risks and privatizing rewards"
3
Those rents are now used to hire top researchers out of public institutions and universities — hemorrhaging the state's internal capability.
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "The same salaries they are paying to the top researchers in universities, both public and private universities, and to people who used to work in the NASAs, the DARPAs ... they are going now to work in these companies. And that hemorrhaging of talent, of top research expertise, I don't think people are talking about this enough"
4
Without that internal capability, governments cannot regulate, understand, or shape AI — including its negative externalities. Reactive policy fills the gap.
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "if you have a research center that's been thinking about climate change for the last 40 years, use them. Don't ask McKinsey, as Australia did, to design your climate strategy, which ended up by the way, being terrible"
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "if you continue to have the old economic models where we judge things by cost benefit analysis, net present value, all these really static metrics, we would have never even bothered going to the moon"
5
Industrial policy that builds back state capability — mission-driven, conditional procurement, in-house expertise — is the precondition for shaping the next wave of technology. Without it, capital-allocation decisions (CHIPS Act, IRA, NextGen EU) become subsidies and handouts rather than directed strategy.
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "just because government's spending a lot money, investing a lot of money, whether it's the IRA or the Next Gen EU 2 trillion recovery plan in Europe, that that means that industrial strategy is back or that government is being really strategic. It depends"
From 2026-05-08-odd-lots-mariana-mazzucato-thinks-we-need-more-moonshots: "moving away from picking winners to picking the willing if you're willing to work with government around these very difficult challenges"
What would falsify this
  • Step 3: If government R&D agencies (DARPA, NSF) report sustained or growing research output and retain top talent at competitive compensation, step 3 is weaker.
  • Step 5: If the CHIPS Act / IRA produce measurable reshoring outcomes (e.g., Intel-Apple-xAI anchor deals close) without commensurate state-capability rebuild, the precondition framing is too strong.
Contradictions / tensions
  • The 'abundance theorem' framing (regulation/planning held us back) is the direct counter — Mazzucato calls this out and disagrees explicitly. Source flagged this tension.
Implications
  • Long-run political backing for industrial policy (CHIPS Act extensions, reshoring incentives) depends on state-capability arguments like this surviving the next administration's posture. Bullish for reshoring beneficiaries (INTC, semicap) only if the policy posture survives political turnover.
  • AI-platform incumbents (NVDA, hyperscaler-tethered semis) face a long-run regulatory tail that follows the state-capability rebuild — even if near-term capex is uncontested.
  • Industrial policy alone is not sufficient — Mazzucato's argument implies CHIPS Act 'just throwing money' fails without a state-capability rebuild.
Companies
none
Concepts
US fab capacity bottleneck → Intel as the alternative foundry
Open questions
none