In 1996, the renowned British political economist Susan Strange heralded the modern era as one characterised by the retreat of the state, arguing, perhaps prematurely, that "the heads of government may be the last to recognise that they and their ministers have lost the authority over national societies and economies they used to have." The rethinking of economic interdependence now proceeding in Washington, Brussels, and Beijing is the institutional codification of a view that supply chains, semiconductor flows, and artificial intelligence capabilities are strategic assets, and that the firms controlling them are consequently instruments of statecraft, whether they choose to be or not. Governments are invoking national security over domains that would have seemed eccentric to classify that way a decade ago. As such, a new corporate environment has emerged, in which the relevant counterpart for political strategy is no longer the Federal Trade Commission (FTC) or the Securities Exchange Commission (SEC) but the National Security Council.
The NVIDIA case best reveals the nature of this new geoeconomic reality. When the Biden administration restricted exports of advanced AI chips to China, citing national security concerns, Jensen Huang lobbied publicly against the controls. His argument, that restrictions would cost Nvidia market share without meaningfully slowing Chinese AI development, was economically coherent. Yet, it didn’t matter: while President Trump partially reversed course and approved H200 sales to China, the implications go beyond market expansion.
Part of the rationale is to flood the Chinese market with mid-range foreign chips and discourage the development of local solutions. NVIDIA’s top-quality products are still subjected to export bans, in the hope that in so doing the Trump administration would achieve its triple goal of lowering Chinese firms’ desire for local solutions, keeping control of the most advanced chips to have an edge on the tech race, and reinforcing NVIDIA’s position in the global supply chain. NVIDIA is now, in effect, a semi-public instrument of American statecraft. AMD and Intel face the same architecture. Indeed, it could be concluded that any firm operating at the frontiers of technology should read this as a signal about what the new firm-state contract looks like.
Advocating against policies of national security is a risky business for firms. When state authority is engaged in “normal” economic regulatory policy, firms’ advice might even be actively sought by the regulator, who regards firms as subject matter experts and privileged sources of information about regulatory design, costs, and implications. However, when policymakers see themselves as tasked with the supraordinate goal of protecting the security of the nation, there is little room for firms’ wisdom about how to maximise the potential of the market.
This paves the way for a structural dilemma. Namely, the markets offering the highest growth are precisely those where operating now carries direct political cost and regulatory contingency. China is the most obvious case. A company that deepens its China exposure risks congressional sanctions, reputational damage in Washington, and the sudden reversal of access that characterised the Biden-to-Trump transition. What the current administration has demonstrated is that access can be restored, but only on the state's terms and at a price, thereby making the firm-state relationship less a background condition of doing business and more an active variable in corporate management.
Firms that resist this dynamic find out to their detriment just what resistance costs: when Anthropic declined to allow the Department of Defense to use its products for mass surveillance and autonomous weapons systems, it was publicly framed as a national security liability and formally deemed a supply-chain risk. OpenAI moved in immediately, striking a deal to supply its services to the Pentagon's classified networks and replacing Anthropic as the only frontier AI provider with access to those contracts.
What these cases show is that the intersection of chip controls, permit regimes, and restricted market access represent a global trend in industrial policy, whereby states set the terms of competition and firms compete partly on the quality of their relationship with governments. The permit architecture being drafted in Washington would give the federal government visibility and leverage at every level of the AI supply chain, transforming the Commerce Department from a sectoral regulator into a gatekeeper for the entire global AI infrastructure buildout. Firms still orienting their political strategy around traditional regulatory lobbying are operating on assumptions that no longer hold. The relevant question for corporate strategy is no longer how to manage regulators but how to manage a state that has decided an industry is an extension of its foreign policy.
