Washington has developed a reliable tell for when an industry wants something it cannot justify on the merits: It puts “AI” in the title.
That is bad for serious debates over artificial intelligence, and it cheapens policy discussions by turning “AI” into a buzzword or sales gimmick.
The instinct is understandable. AI is the defining technology story of the moment, and policymakers are eager to be seen as responsive. That creates an opening, and savvy advocates have not been slow to exploit it. The result is a growing list of proposals that invoke AI while advancing objectives that long predate the current AI debate.
Spectrum policy — the rules governing the allocation, licensing and use of radiofrequency spectrum — is now squarely in this territory. It deserves more careful treatment.
The core facts are clear, even if they are routinely obscured. Somewhere between 80% and 90% of mobile data traffic already travels over Wi-Fi — unlicensed local wireless networks, not licensed cellular networks. Wi-Fi carries roughly 10 times more data than all licensed cellular networks combined. And the architecture of the AI services consumers and businesses are adopting — cloud-dependent, fixed-connection-friendly, built around short prompts and returned responses rather than continuous high-bandwidth streams — does not obviously change that calculus. These facts should be the starting point for any serious analysis.
Given that reality, proposals to reallocate 6 GHz spectrum away from unlicensed use toward exclusive licensed access deserve scrutiny. The 6 GHz band supports the Wi-Fi networks doing much of the heavy lifting in the American wireless data ecosystem. Shrinking that resource in the name of preparing for an AI-driven traffic surge would be an odd response to demand that, based on today’s evidence, is more likely to flow over Wi-Fi than cellular networks.
If anything, forward-looking spectrum policy should look upward, to the Lower 7 GHz band, which sits directly above the current Wi-Fi allocation and could provide meaningful additional unlicensed capacity without displacing what already works.
The enterprise AI story is similarly complex. Private 5G networks, largely running on CBRS spectrum, are increasingly used by manufacturers and industrial operators building AI-powered facilities. Satellite connectivity is attracting serious capital for applications such as drone logistics, where terrestrial infrastructure may be unavailable or unreliable. The AI future will not be dominated by a single spectrum technology. It will rely on heterogeneous networks, unlicensed bands and competitive alternatives.
Licensed spectrum remains important to American AI leadership. But importance is not exclusivity, and proximity to AI is not necessity. Policymakers allocating scarce spectrum resources should ground decisions in data: where AI traffic flows, which connectivity models markets are building and what consumers and businesses need.
Saying “AI” loudly is no substitute for answering those questions.