There is one thing Europe imports even more regularly than semiconductors: prophecies of its own irrelevance. The latest comes from Yuval Noah Harari.
To the core: In an interview with Nicola Abé published by Der Spiegel on August 9 — unlikely to circulate in Italian anytime soon, and therefore worth quoting in its key passages — the historian presents Europe with an either-or choice.
- “Become a submissive vassal of the United States” or become “an independent actor; it is not too late yet,” provided Europeans “cooperate and invest billions.” The window, he warns, is narrow: “We have only a few years left.”
Three of Harari’s arguments hold up:
- The first: “This time we are creating actors, not tools.” A knife or a generator waits for a human hand; an artificial agent makes decisions. Technology policy therefore shifts from governing the use of a tool to governing the agent itself.
- The second is his most effective image: “Queen Victoria did not have a button she could press to make the rifle stop firing.” Today, by contrast, whoever supplies technological systems could potentially “paralyze their weapons systems, infrastructure and government administrations.”
- This is precisely the substance of the Strategic Report that the Digital Economy Center (CED) devoted to Technological Sovereignty in 2020, when the expression was still marginal in the Italian debate. On June 3, the European Commission placed it at the heart of a new legislative package.
- The third argument is the most fertile, yet also the least developed: AI agents as “perfect bureaucrats.” Harari argues that “we should fear killer robots less than AI actors rising to power in financial systems and public administration.” He reaches the crucial point — and then abandons it after three lines.
Up to this point, the interview serves Europe well. From here on, it leaves Europe on its own.
The argument begins to falter with the formulation chosen for the headline: “This is the moment when capitalism and humanity part ways,” with AI companies accumulating capital and eventually colonizing the galaxy “without humans.”
- Yet capitalism separated from humanity would amount to a balance sheet without demand: accumulation running in circles, deprived both of the market that gives it value and of the legal order that recognizes it. If artificial agents themselves became both demand and counterparty, that would constitute a new regime, one without a name or legal status.
- The serious version of Harari’s argument lies instead with his “perfect bureaucrats”: algorithmic actors capable of operating through markets and administrative procedures faster than anyone who may ultimately be held accountable for their actions. That is something policymakers can legislate for now. Eschatology is not.
- The interview also runs ahead of the evidence when it comes to employment. Investments “will only be profitable if AI takes over all activities, from the textile worker in Bangladesh to the lawyer in London,” Harari argues, predicting recession for Europe’s service-based economies.
Available macroeconomic estimates suggest a more limited scale and speed. Daron Acemoglu, writing in Economic Policy in 2025, estimates that the cumulative increase in total factor productivity generated by AI will be no more than 0.66% over ten years. The Bank of Italy, in June, estimated an annual impact for Italy of between 0.2 and 1.1 percentage points, depending on the depth of adoption.
- Different metrics, but a converging lesson: technology rewards those who adopt it and punishes those who merely undergo it. The extinction of work remains a theoretical hypothesis. The cost of delayed adoption is already showing up in national accounts.
The strategic weakness in Harari’s argument is its binary logic. Between the vassal and the solitary independent actor lies an approach that global industry has practiced for two decades. Between the 2003-2006 period and 2019-2022, patents jointly filed by competing companies increased by 159%. We call it coopetition.
- Technological Sovereignty means the ability to generate knowledge autonomously or together with trusted partners. Its optimal level never coincides with self-sufficiency. In the age of artificial agents, sovereignty belongs to those capable of deciding where to remain open and where to close themselves off — and who possess the industrial strength to enforce that choice.
Coopetitive Technological Sovereignty is the third option. As for Harari’s warning that “it is not too late yet,” Europe’s construction site is further advanced than the interview suggests — but further behind than the deadline requires.
- On July 30, the call for AI Gigafactories was launched: up to seven facilities, with €10 billion in public funding intended to mobilize at least €20 billion in private investment. Bids close on November 12, with awards expected in early 2027. Italy is preparing a coordinated bid centered around Leonardo, Eni, AI4I and other partners.
The Cloud and AI Development Act, presented on June 3, establishes four levels of safeguards for cloud services used by public administrations, at a time when Europe’s share of the cloud market fell from 29% to 15% between 2017 and 2022.
- On the other side of the ledger is the postponement, agreed in May, of obligations applying to stand-alone high-risk systems under Annex III, from August 2, 2026 to December 2, 2027. A Union that postpones its rules before it possesses the machines tells us something about where we stand.
The decisive lever, meanwhile, remains largely untouched: demand. Across the seven European countries examined in CED’s 2026 Strategic Report, High-Tech Economy, every dollar of high-tech value added generates $3.90 in economic output over three years, compared with $1.28 in traditional sectors. Italy concentrates 70.9% of private research in sectors accounting for just 10.9% of total value added.
The multiplier exists. What is missing is the customer capable of activating it. That is why, including in these pages, we have proposed treating advanced computing capacity as an industrial public utility.
- One final note, because precision is the currency of any argument built around a ticking clock. Argentina, which Harari says “has just decided to allow companies run exclusively by AI,” currently has a proposal from the Milei government before Congress. It is contested domestically and still requires human beings at the top.
- The $80 billion that SpaceX supposedly raised from private investors “last year” also shifts both in nature and chronology. The December 2025 secondary sale established an $800 billion valuation without raising new capital. The $85.7 billion actually raised came from the initial public offering completed in June: public markets, this year.
The bottom line: The underlying phenomenon — private capital financing ideological bets on a sovereign scale — is real. Precisely for that reason, it deserves anchors that hold.
- This autumn will tell us more than the prophecies will, as negotiations over the Cloud and AI Development Act unfold and the race for AI Gigafactories accelerates, with Italy in the running. That is where, long before superintelligence arrives, the distance between vassal and actor will be measured.
- Europe owes Harari’s interview the right question: who holds power in the age of artificial agents? But prophecy counts the years that remain. Industrial policy spends them.
