Mariana Mazzucato

In the [YOU] on AI Field Guide

The cycle that opened with [YOU] on AI begins in a room in Trivandrum where twenty engineers discover a twenty-fold productivity multiplier. That room is exhilarating. Mazzucato’s framework asks the question the room cannot see from inside: who built the floor the room stands on? The transformer architecture that makes the multiplier possible descends from decades of publicly funded research—NSF grants sustaining neural network theory through two AI winters, DARPA contracts funding natural language processing, Canadian and European public universities training the researchers who eventually produced the breakthrough. The exhilaration is real and so is the plumbing. Mazzucato is the thinker who insists on reading both at once.

Her lens reframes the cycle’s central tensions. The democratization paradox—the simultaneous expansion of individual capability and deepening of platform dependency—is, in her framework, a predictable consequence of technological progress delivered through extractive channels. The builder in Trivandrum who gains a twenty-fold productivity multiplier also acquires a structural dependence on a proprietary platform built on publicly funded foundations. Both dynamics operate through the same transaction. The cycle names the first and celebrates it; Mazzucato names the second and refuses to accept it as inevitable.

She provides the distributional vocabulary the cycle needs at its most serious moments. When her analysis distinguishes between profits from genuine innovation and rents extracted through monopoly power, it supplies a scalpel where most commentary uses a sledgehammer. The AI economy contains both—genuine value creation and genuine extraction, often within the same company and the same product. Her framework is the instrument that separates them, and without it the cycle’s optimism about capability and its concern about concentration cannot be held in the same hand without contradiction.

Her epilogue note to her own book is the one that cuts deepest into the cycle’s concerns: the GPS receipt was for twelve billion dollars of public money that enabled three hundred and twenty billion in annual economic activity—and nobody repaid it. The institutional architecture had no mechanism for repayment. The AI transition is reproducing that pattern at greater scale and speed, and Mazzucato’s career is the sustained argument that the pattern is not a law of nature. It is a choice. And the time for choosing otherwise is now, while the distributional trajectories remain malleable.

Origin

Born in Rome in 1968 and raised partly in the United States, Mazzucato trained as an economist at Tufts and the New School for Social Research, completing her doctorate in 1999. Her early work on industrial dynamics and firm growth was technically focused but already inflected by the question that would define her career: what actually drives the long-run expansion of economic capability, and who deserves the returns from it? The answer, she came to believe, required looking seriously at the state—not the idealised market-correcting state of welfare economics, but the actually existing state that had funded the Manhattan Project, built the internet, and sustained semiconductor research through the decades when venture capital had abandoned the field.

The empirical turn that produced The Entrepreneurial State was prosecutorial in method. She assembled the funding genealogy of individual technologies, tracing each foundational capability—the lithium-ion battery in the iPhone, the touchscreen, Siri, GPS—back through its research lineage to the public grants and procurement contracts that had sustained it. The result was not a theoretical argument but a documented record, and the record was devastating to the prevailing narrative. The myth of the garage—the founding fiction that locates innovation in private initiative and individual genius—was not merely incomplete. It was the active concealment of a massive public contribution, and the concealment had distributional consequences: it delegitimised the state’s claim on the returns from investments it had actually made.

She founded the Institute for Innovation and Public Purpose at UCL in 2017 to translate the analytical framework into institutional practice, advising governments from the UK to the European Commission to South Africa on how to build what she calls the “entrepreneurial state”—not the bureaucratic state that manages market failures, but the patient, risk-taking, mission-oriented state that creates the markets private capital is later celebrated for exploiting. Her concept of conditionality—that public support for private firms must come with obligations attached—and her mission-oriented AI policy framework are now among the most cited frameworks in European and international AI governance.

Key Ideas

The Myth of the Garage. The foundational fiction of Silicon Valley locates the origin of innovation in private initiative and individual genius. Mazzucato’s empirical counter-argument is simple: trace the funding genealogy. The semiconductor industry depended on military procurement. The internet was ARPANET. GPS was the US Navy. Siri was a DARPA-commissioned project at Stanford Research Institute. The myth is not false—work happened in the garage—but it performs a specific ideological function: it makes the public’s contribution invisible, and in doing so eliminates the case for a public return.

Public Risk, Private Reward. The state bears the foundational risk of innovation—sustaining research through AI winters, funding science when commercial outcomes are decades away—and receives no proportionate share of the financial returns when those investments succeed. This structural pattern, documented across pharmaceuticals, the internet, GPS, and AI, is the central distributional injustice of the technology economy. The mRNA vaccine case is Mazzucato’s clearest precedent: thirty years of publicly funded research, billions in government contracts, and the returns distributed exclusively to Pfizer and Moderna shareholders.

Value Creation versus Extraction. Not all private returns from AI are equivalent. Value creation produces goods and services that meet genuine human needs; value extraction captures returns from value created by others without producing corresponding value in return. An advertising optimisation algorithm that exploits informational asymmetries to redirect purchasing decisions is structurally different from an AI tool that expands a builder’s capability to solve genuine problems. The distinction matters for tax policy, competition law, and the governance of AI platforms—and it is systematically obscured by financial metrics that measure the capture of value, not its creation.

Mission-Oriented Innovation. Against the framing of the state as referee, Mazzucato proposes the state as director—using public investment and institutional design to ensure AI capability flows toward societal needs and not only toward the applications that generate the highest private returns. The Apollo programme did not land a human on the moon by regulating aerospace; it set a mission and organised collective effort around it. The mission-oriented AI framework applies the same logic: publicly defined goals, patient public investment, conditionality requirements that link public support to public benefit.

The Algorithmic Rents Programme. Working with Tim O’Reilly and Ilan Strauss at the IIPP, Mazzucato documented how platform companies use algorithmic systems not primarily to create genuine value but to concentrate market power and extract wealth from users and smaller players. The algorithmic rents research showed that the dominant platform business model is partly extractive rather than creative—and that AI amplifies extraction with the same efficiency it amplifies creation. The amplifier does not discriminate.