Stay informed with free updates

The writer is deputy director of the Thurman Arnold Project and a resident fellow at Yale’s Information Society Project

When US antitrust enforcers issued new merger guidelines in 2023, there was cautious optimism that lessons from the previous era of Big Tech deals had been absorbed.

The guidelines made clear that mergers would be scrutinised not only when a company acquired an existing rival and exceeded 30 per cent market share, but also when it acquired a potential competitor or formed part of a pattern of serial consolidation.

The recent wave of artificial intelligence partnerships was the first real test of these principles. Just last year the Federal Trade Commission launched a study of AI partnerships and warned of potential harms.

But the mood has shifted since then. Domestic monopolies are now being rebranded as national champions in a geopolitical race with China.

The problem began with Stargate, the $500bn joint venture between Arm, Microsoft, Nvidia, Oracle and OpenAI. A project of this scale that unites so many technology companies should have triggered rigorous merger scrutiny. Instead, it was announced by President Donald Trump in a gesture of tacit government endorsement and was celebrated as a means of promoting US technological supremacy.

In the past year, OpenAI, Nvidia, Oracle, CoreWeave, Microsoft, AMD and SoftBank have all struck a series of partnerships designed to power the great US AI infrastructure build-out.

The circular financing in these deals raises the risk of a bubble bursting in AI. But what has drawn less notice is the competition problem being created. The deals are forming a series of monopoly moats across the entire AI supply chain.

Large-scale mergers, acquisitions, joint ventures and investments should attract antitrust attention. Both federal and state regulators have the authority and the tools to act. Merger reviews should examine how deals reshape competition, whether horizontally (when rivals merge or collaborate) or vertically (when companies at different levels of the supply chain integrate in ways that could raise rivals’ costs). 

The current spree of AI investments and partnerships creates both horizontal and vertical concerns. Nvidia’s $5bn investment in rival Intel, for example, has clear horizontal implications in an already concentrated chip market. Similarly, Microsoft, Oracle and CoreWeave operate in the same concentrated cloud-services market. By joining coalitions such as Stargate, they have linked their economic interests. Such strategic partnerships risk blunting head-to-head competition.

The alliances also raise vertical concerns of input foreclosure, that is, when control over a key input allows one party to disadvantage rivals. Take the Nvidia-OpenAI partnership. The antitrust question is whether Nvidia could prioritise OpenAI by giving it early access to its most advanced chips, or offer them at discounted prices. In a market where chips are in high demand, such behaviour would restrict rival AI companies from accessing an essential input. The problem cuts both ways. Competing chipmakers may also find it harder to win OpenAI’s business.

This expanding web of alliances also opens the door to classic antitrust concerns such as exclusive deals and bundling. Many cloud providers already impose minimum-spend commitments, locking in customers and limiting their ability to spread demand across multiple cloud providers.

Buying nascent rivals is an old page from the tech monopoly playbook. So too are deals that align incentives and neutralise potential competition. Google paid Apple billions of dollars to become the default search option on its devices. This also encouraged Apple to stay out of search. Google also struck an agreement to discourage Facebook from launching competing ad technologies, giving Facebook preferential treatment in its ad auction.

The risk is clear. As the economic fortunes of AI companies become more intertwined through cross-investments and partnerships, the incentive to compete erodes. Competitive discipline could give way to quiet co-ordination, with each company guarding its own monopoly moat while sharing monopoly rents with partners.

The assumption that concentration breeds strength is both economically and historically misguided. Competition, not protectionism or concentration, is what drives innovation. Antitrust regulators should resist the pull of national security rhetoric and see these alliances for what they actually are: efforts to entrench power and extend monopoly control into the next technological era.Â