AI Market Structure Reform Act

Policy Recommendation

The United States should enact the AI Market Structure Reform Act (AMSRA) to regulate the emerging AI industry through a tiered tax incentive rewarding Frontier Labs that supply certified Intermediaries at stable prices; mandatory price transparency filings and a discrimination prohibition preventing Frontier Labs from undercutting the intermediary market they are intended to serve; and a structural separation requirement for labs operating both foundation model infrastructure and consumer-facing products.

Strategic Context

There are a small number of Frontier Labs with foundation models: models that can be utilized for a wide range of downstream applications and are trained with above 10²⁶ compute operations for all training, testing, and tuning [1, 2]. The Intermediaries are companies that build on top of foundation models to create domain-specific products. However, the Intermediaries face a constant threat of assimilation into the foundation model they build their infrastructure upon due to the disparity in resources [3]. This threat for Intermediaries comes from three main directions. First, Frontier Labs have an incentive to bypass Intermediaries and sell directly to consumers, to increase their margins. Second, the oligopolistic structure of the Frontier Lab market makes implicit price coordination likely and difficult to prosecute under existing antitrust law [4]. And third, Frontier Labs who operate foundation model infrastructure will attempt to integrate domain-specific intermediary products to further model ability. Without the AMSRA, Frontier Labs will crowd out Intermediaries and concentrate economic power and technical ability into a handful of firms; thereby reducing market competition, limiting innovation, and hindering development in the foundation model market.

Policy Proposal and Implementation Mechanics

1.) Two Tiered Tax Incentive: First tier utilizes known tax incentives to shift behaviors towards policy preferred activities [5]. Frontier Labs supplying certified Intermediaries receive an enhanced deduction of 130% of qualifying compute expenditure. The second tier provides additional credit for multi-year fixed-price or capped-priced contracts, while lacking precedent, is justified by the absence of any existing mechanisms that reward price stability in the foundation model market.

Implementation and Deployment Considerations

Tier 1 will operate through existing IRS workflows. Intermediary certification is administered by the FTC, and can be replicated through existing workflows such as the Data Privacy Framework [6]. Tier 2 requires economic modeling against current Frontier Lab margin structures. OpenAI's CFO confirmed $20B in 2025 ARR at a 33% gross margin with consumer subscriptions dominating over API revenue [7, 8], suggesting meaningful headroom for a credit to shift behavior, though the precise threshold requires rulemaking.

2.) Antimonopoly Provisions: Three sub-provisions aimed to address market structure directly. A. Institute mandatory public price disclosures for Frontier Labs enabling prosecution of implicit coordination [9]. B. Require labs with consumer products to choose between accounting separation or structural divestiture within 18 months of enactment, modeled on the structural separation framework applied to AT&T in 1984 and theorized by previous chair of FTC Lina Khan [10, 11]. C. Prohibit Frontier Labs from offering consumer products at effective model costs below what they charge certified Intermediaries for equivalent capabilities by utilizing non-discrimination rules.

Implementation and Deployment Considerations

Scholarly sources identified that full public filing of pricing data can function as a coordination mechanism by giving pricing information to competitors, though the consumer benefit of transparency is judged to outweigh this risk [12]. The non-discrimination clause needs a workable and malleable definition of "equivalent capabilities"; NIST will provide technical benchmarks to answer this question but faces challenges with risk estimation due to lack of visibility in model training and the lack of precedent in the area of AI [13]. So, even with a sufficiently narrow definition, Frontier Labs may satisfy requirements while preferencing internal products through other non-price mechanisms.

Trade-offs and Projected Impacts

Revenue Impact

The Tier 1 deduction reduces federal tax receipts on Frontier Lab’s qualifying revenue, monitored by the CBO, will likely position Frontier Labs as net tax recipients [14]. The Tier 2 credit creates an additional cost that is dependent on the existence of multi-year fixed-price contracts, which are currently rare in the API market. However, the expected benefit of a functioning intermediary ecosystem exceeds possible expenses by expanding the total revenue opportunity for the given foundation model capabilities, likely increasing total taxable activity over a 5–10 year horizon.

FTC Impact

The AMSRA requires the FTC to perform three distinct new functions: running an Intermediary certification program, maintaining and overseeing pricing schedules of Frontier Labs, and enforcing the discrimination prohibition. None of these capacities exist in the FTC today. The manpower required for the Intermediary certification process would pose a significant strain on the FTC. This would require a larger budget for the FTC, which would need to be approved by the CBO. But ultimately, this problem is a fiscal one instead of a structural one. AMSRA does not include funding for the new FTC functions requiring new staffing levels and increased budgets. The implementation of the AMSRA immediately would not provide the non-discrimination enforcement authority ample time to fix this issue.

Durability

The AMSRA is durable in two different ways. First, it will benefit current and future Intermediaries. Even if Frontier Labs aggressively develop vertically in the upcoming years, the market will never be exhausted of ideas, generating new Intermediaries who will be benefited. Second, the antimonopoly provision of the policy addresses an unprecedented but pressing issue in the field of AI. Since there is little existing AI regulation, it lacks the basic guardrails in existing fields. From both perspectives, the AMSRA serves as an innovative policy capable of setting the foundation for legal restrictions and regulations of AI for years to come.