The race to build AI infrastructure is changing how the largest technology companies compete, and antitrust regulators are watching closely. Many of the companies now spending billions on data centers, power, land, and computing hardware are the same companies that faced major antitrust scrutiny over the past two decades. That history is affecting how enforcers and private plaintiffs are starting to look at the current AI buildout, even though no major AI infrastructure case has reached a final resolution yet.
Four theories are emerging as the most likely areas of risk.
Locking Up More Supply Than a Company Needs
Long-term power purchase agreements, land acquisitions, and reserved grid capacity are standard business practices, but the concern arises when a dominant company acquires far more than it needs simply to keep that supply away from competitors. Because land near reliable power grids and interconnection capacity are limited, a company that locks up these inputs through a pattern of agreements could make it difficult for smaller AI companies to build competitive infrastructure.
This theory would extend existing antitrust law and has not yet been tested directly against a hyperscaler.
Acquisitions Designed to Eliminate Future Competition
Regulators are also scrutinizing acquisitions of promising AI hardware or infrastructure companies where the buyer’s real goal may be to prevent that company from ever supplying a rival, rather than to use what it acquires. Under current merger guidelines, a target does not need current revenue or market share for this concern to apply.
Many of these deals fall below the dollar thresholds that require advance notice to regulators, but that does not mean they escape scrutiny. Both the FTC and DOJ have shown a willingness to review these transactions after they close.
Exclusive Agreements That Cut Off Rivals From Essential Inputs
Contracts that lock up power suppliers, landowners, or hardware makers for one company can leave rivals with nothing to buy. These agreements do not need to say “exclusive” to raise concerns. Provisions such as volume commitments that consume a supplier’s full capacity, most-favored-nation clauses, and rights of first refusal can have the same effect. A recent class action against Microsoft over its exclusive cloud computing agreement with OpenAI shows this theory is already being tested in court.
Vertical Integration That Chokes Off Access to a Key Input
When a company buys a supplier that makes something its competitors need, regulators consider whether the combined company has both the ability and the incentive to cut off rivals or offer them worse terms. Courts have historically been reluctant to block these deals, and the government has lost several recent vertical merger challenges, but investigations into these transactions remain common and are costly regardless of the outcome.
What This Means Going Forward
None of these theories are settled law when it comes to AI infrastructure, and enforcers would need to break new ground to succeed on several of them. However, the volume of long-term contracts, acquisitions, and infrastructure deals being signed right now means the businesses, competitors, and consumers affected by this activity should pay attention.
Talk to Keller Grover About a Potential Antitrust Claim
If you believe your company has been shut out of essential inputs, or that you have paid higher prices because of anticompetitive conduct tied to AI infrastructure, contact Keller Grover’s antitrust litigation team to discuss your options. Reach out to our office today.