The usual suspects have recently discovered a deep concern about circular flows of capital in the artificial intelligence sector. When hardware producers provide financing to customers who then purchase their chips, critics warn of speculative bubbles and artificial demand. Yet this skepticism is remarkably selective. The same observers who denounce such arrangements in the private sector often celebrate similar mechanisms when the state performs them through taxation and redistribution.
This double standard deserves attention.
In the AI infrastructure market, financing arrangements between chip manufacturers and developers have attracted growing scrutiny. Critics describe these deals as closed loops in which suppliers effectively fund demand for their own products. Recent headlines have focused on agreements in which major cloud providers and chipmakers invest billions into AI labs and then record long-term commitments for computing services or hardware usage. Such arrangements are often accused of “revenue round-tripping” or of being “circular deals” that artificially inflate growth metrics.
But this interpretation fails to distinguish between accounting manipulation and the productive activation of resources.
When a manufacturer extends credit, it converts idle inventory or capital into active use. The manufacturer risks its own balance sheet because it expects the buyer to generate real value. If the investment fails, the correction is localized and severe. The loss falls directly on the participating agents.
Even in the widely discussed cases where a supplier takes an equity stake while securing multi-year commitments for its own cloud infrastructure or chips, the logic remains the same. The parties are binding themselves to a shared expectation that future productivity will more than justify the upfront capital.
This is precisely the strength of markets. Private credit is a mechanism of discipline rather than just a motor for technological acceleration. Since the Renaissance, credit networks have allowed innovators to mobilize resources beyond their immediate reach. While the risk of failure exists, the market ensures that errors lead to liquidation. Investors learn and capital moves to better uses.
Something similar is happening in AI today: capital‑intensive infrastructure such as specialised chips and data centres is pulled forward in time through supplier credit, venture finance and long‑term usage contracts, allowing developers to experiment today with capacities that their current cash flows alone could not justify.
The worry that “the system feeds on itself” only becomes economically meaningful if one can show that the relevant losses will be imposed on unwilling third parties rather than on the investors and firms that structured the deals.
The important point is that those who take the risks also bear the consequences.
Now consider the mechanism that lies at the heart of government spending.
Governments collect resources through forceful taxation and compulsory contributions, then redistribute them through subsidies and publicly funded programs. This process is frequently described as economic stimulus or public investment, as if the state were creating value simply by circulating money through its institutions.
Yet the critique of circularity applies far more accurately here.
The state taxes economic activity to fund programs that subtract resources from the productive economy and then redistribute and then redistribute them inefficiently, thereby justifying the expansion of the state. Unlike private financial arrangements, this cycle is institutionalized and never faces the discipline of market correction.
The difference becomes clearer when examining the nature of the resources involved.
In AI infrastructure markets, suppliers possess tangible assets that may remain idle or be deployed productively. Financing helps move those assets into applications that generate innovation and growth. The transaction activates resources that were previously underutilized. When a supplier effectively “pre‑finances” demand, what is being advanced are specific goods and services like chips or data‑centre capacity that did not yet have a buyer at market price. The risk is that future demand may not justify the advance.
With the government, by contrast, the resources being redistributed were never idle. They were already part of the productive economy, generated through the work of individuals and firms who had their own plans for how to use them. Taxation diverts those resources from their original purposes and reallocates them through political decision-making. When the money is redistributed, no new production appears. It simply changes who spends it.
Yet this process is routinely described as if it were itself a generator of wealth.
The language of “public investment” illustrates the problem. Investment normally implies an expectation of return, measurable performance, and the possibility of withdrawing capital when projects fail. In the public sector these conditions rarely exist. Programs often continue regardless of results, and when they prove inefficient the costs are absorbed by taxpayers through higher taxes or additional public debt.
The contrast with private risk taking could hardly be sharper.
When a semiconductor company extends credit or acquires equity in a customer and misjudges the opportunity, it loses money. Shareholders bear the consequences and managers may lose their positions. The feedback is immediate and unavoidable.
When governments misallocate billions through poorly designed subsidies or ineffective programs, decision-makers rarely face comparable consequences. Losses are dispersed across millions of taxpayers, often hidden within complex budgets or financed through borrowing that postpones the burden. Failed public “investments” frequently become arguments for expanding the same programs rather than terminating them.
In this sense, the mechanism critics claim to fear in the AI economy actually operates in a far more devastating form within the welfare state.
Commentators worry when producers help finance demand for their own products. Yet governments routinely tax economic activity, destroying it, to fund programs that justify further taxation to “stimulate the economy”. What in the private sector is a contingent network of contracts becomes, in the public sector, a permanent political structure that converts private income into claims on that income.
If circular financial relationships raise concerns about sustainability and discipline, those concerns should apply even more strongly where decision-makers are insulated from the consequences of failure.
Cooperation among private agents, even when supported by credit, is often portrayed as artificial or dangerous. Redistribution carried out by the state is portrayed as virtuous and stabilizing. Yet from the perspective of incentives, accountability, and productive efficiency, the comparison points in the opposite direction.
* Tomás Lucena Barreiro is a writer and Local Coordinator for Students For Liberty. He holds a bachelor’s degree in Law and a master’s degree in Law and Management. He has written for outlets such as Observador, organised public events on behalf of the liberty movement, and was elected by Iniciativa Liberal to a parish assembly in Lisbon. His work focuses on classical liberalism, economics, public policy, and culture.
Source: We Are Innovation
