Artificial intelligence (AI) tools are rapidly becoming general-purpose technologies with the potential to deliver productivity enhancements to a broad range of task types and industries. We previously analyzed the potential for AI tools to deliver significant federal fiscal benefits provided that AI rules are harmonized nationwide and do not discriminate against AI tools relative to other tools. Now we estimate the benefits of widespread AI adoption for cost-effective broadband deployment, and the results show the importance of avoiding conflicting AI regulations.
AI’s quantifiable capex dividend for broadband
Oliver Wyman’s recent network-operations analysis shows that generative-AI-driven design, permitting, and construction workflows drive a 5% to 10% reduction in baseline capex for broadband builds, while also compressing build schedules by up to 60%. The savings come from productivity improvements across a wide range of tasks, from automated bill-of-materials generation and predictive radio frequency modeling to drone-based progress verification, among many other use cases. These figures are not outliers: McKinsey finds that operators deploying AI “digital twins” of their fixed and mobile networks have already optimized capex plans by 10% to 15% in the field, while Infosys estimates AI can reduce costs by 7% to 15%.
Because AI productivity enhancement applies across every step of the network life-cycle-strategy, engineering, build-out, and maintenance—AI’s returns look more like those from a general-purpose technology than a point solution. In plain terms, that productivity improvement from AI means more broadband coverage can be achieved with the same dollar amount of capex–so more households can be connected to broadband internet service without requiring more spending.
What the savings mean in dollars
U.S. broadband providers reportedly invested $94.7 billion in capex in 2023. If AI adoption proceeds unhindered, a defensible 5% to 10% savings translates into $4.7 billion to $9.5 billion of annual capex headroom—funds that can be redeployed to reach more homes.
However, conflicting state AI statutes could cut adoption rates roughly in half, as legal and compliance teams struggle to reconcile 50 different compliance regimes and some states make compliance prohibitively difficult. Halving adoption would slash the attainable capex savings to 2.5% to 5%—still material, but only half the achievable efficiency.
Applied to the 2023 capex base line, that reduced saving is $2.4 billion to $4.7 billion per year. Keeping capex spending constant while increasing its effectiveness by that margin is economically identical to injecting the same dollar amount of fresh investment, assuming no significant decline in marginal households connected to broadband per dollar spent.
Translating dollars to connected households
America’s $42.5 billion Broadband Equity, Access, and Deployment (BEAD) program, which was established by the Infrastructure Investment and Jobs Act, was designed with a primary measurable objective: connect as many un-served and underserved households as possible to high-speed internet infrastructure. Yet the effectiveness of every federal BEAD dollar hinges on how efficiently U.S. telecommunications carriers can translate capital expenditure into new route-miles of fiber and wireless-based coverage. Artificial intelligence is already proving to be the next great productivity lever for network deployment—but only if adoption is not strangled by a tangle of conflicting state AI rules.
BEAD’s rules effectively fix the public contribution: states receive a formula-based allocation that they must stretch across unserved and underserved addresses. If we assume that households per route-mile are uniform and that provider capex stays constant, then a 2.5% to 5% bump in program efficiency yields a commensurate 2.5% to 5% jump in the number of households a provider can hook up for the same money.
Why preemption is the only workable tool
The existence of state and local AI laws, combined with any future adopted AI safeguards, create novel problems which negatively impact broadband deployment to the detriment of policy objectives and both taxpayers’ and consumers’ bottom lines. Limiting these frameworks through programs like BEAD leads to demonstrable benefits. For example, if one state requires that AI models consider an objective outside of efficiency for all real-world use cases and another state requires that AI models not consider any factors other than efficiency for real-world use cases, compliance forces broadband providers to choose between delaying roll-outs until lowest-common-denominator compliance is engineered, or investing in duplicative AI tools for each distinct state, or to abandon AI tools entirely in certain regions. Any path imposes extra costs and deadweight loss that ultimately shows up in higher deployment costs and fewer homes connected to broadband, including fewer BEAD-funded connections.
Federal preemption through a state-level moratorium is therefore not deregulation as much as it is a temporary coordination mechanism that buys Congress and the relevant expert agencies time to craft a harmonized national framework. During that window, providers can deploy AI at scale, harvest the documented cost savings, and pass those gains to taxpayers and consumers in the form of faster, cheaper broadband builds.
From a taxpayer-value perspective, the calculus is stark. A 5% efficiency penalty on a ~$42.5 billion federal program would effectively squander more than $2 billion in broadband coverage—dollars that could otherwise close the gap for hundreds of thousands of rural Americans. Pre-emption costs the Treasury nothing; the absence of preemption carries a multibillion-dollar opportunity cost.
Conclusion
Economic policy should be judged by outcomes. If broadband deployment to un-served and under-served households is the objective of U.S. policies like the BEAD program, then we must wring every possible bit of productivity out of private and public capital. AI is already demonstrably capable of delivering 5% to 10% capex efficiency gains, but only if providers can deploy the technology uniformly across their network footprints. A limitation on onerous state-level AI regulations is the simplest, most cost-effective way to secure those gains. Absent that step, Congress and the states risk watching billions in taxpayer funds—and a once-in-a-generation chance to achieve broadband ubiquity—evaporate into the friction of 50 different rulebooks.
