The Spectrum Auction That Turned Air Into Treasury

A commons that regulators gave away

The Federal Communications Commission was created by Congress in 1934 with authority to manage and allocate the country's nonfederal spectrum and issue licenses in the public interest, according to a 2025 history of the agency's auction authority prepared by the Congressional Research Service. For nearly sixty years after that, the FCC did not sell the licenses it granted. It awarded them.

Two methods dominated. A comparative hearing had applicants argue their case before the commission, which then chose a winner by judgment, a process that was slow and open to influence by whichever applicant could afford the longest legal fight. A lottery instead handed the license to whoever's name was drawn, regardless of whether that person or company had any plan to use it. Both methods were widely criticized as inefficient, the Congressional Research Service history notes, and the cellular license lotteries the FCC ran through the 1980s became a well known example of the second problem: winners with no intention of building a network could simply resell what they had been handed for a profit.

What both methods shared was a refusal to let price do the sorting. A hearing asked a regulator to guess who deserved a license. A lottery asked luck to decide. Neither asked the market what the spectrum was worth, because for most of the twentieth century nobody thought to ask.

The idea that spectrum should be treated as a commons in the first place goes back further than the FCC itself, to the Federal Radio Commission Congress set up in 1927 to bring order to a crowded, interference-plagued broadcast band. The premise carried forward when the FCC absorbed that role in 1934: the airwaves belonged to the public, and a license was a temporary, revocable permission to use them, not a piece of property a company owned outright. That premise is still the legal basis for every spectrum license issued today. What changed in 1993 was only the method of choosing who received the permission, not the underlying claim that the public, through its government, was the one granting it.

The auction that had to be invented

Congress changed that in 1993, when it authorized the FCC to allocate spectrum licenses through competitive auction for the first time, ending six decades of hearings and lotteries, per the same Congressional Research Service history. But an ordinary auction would not work. Spectrum licenses are not independent goods. A company bidding for a frequency in one city often needs an adjacent frequency in the next city over, or a complementary band in the same city, to build a network that actually functions, so a simple sealed-bid sale risked leaving the most efficient combinations of licenses unassembled.

The FCC turned to auction-theory economists, among them Paul Milgrom and Robert Wilson, whose work on the problem later contributed to the 2020 Nobel Memorial Prize in Economic Sciences, to design something new: the simultaneous ascending auction. Every license in the sale opens for bidding at once, across repeated rounds, and bidders can shift their money between packages of licenses as prices climb. The auction closes only when a round passes with no new bids anywhere in the sale, which lets the market, not a regulator's guess, settle both the price and the shape of who ends up holding what.

The design worked well enough to become what one account of its history, published by Priceonomics, called the international "workhorse" of spectrum sales: the same simultaneous ascending format the FCC pioneered has since been adopted by regulators worldwide and has generated more than $200 billion in cumulative revenue for governments outside the United States. A single piece of market design, built to solve an American licensing problem, became an export in its own right.

Its durability shows up in results decades later. The FCC's AWS-3 sale, Auction 113, drew 17 qualified bidders across 72 rounds and raised more than $3.5 billion, comfortably exceeding the agency's own projections, according to reporting on the auction's results in In Compliance Magazine. A mechanism invented in the mid-1990s was still outperforming expectations twenty years on.

The rounds themselves are the visible part of what the design does. In an early round, bidders place modest opening bids across the licenses they want, largely to signal interest. As rounds proceed, a bidder who is outbid on one license can shift its planned spending toward a different license or a different combination of licenses, building toward whatever set of frequencies actually lets it run a coherent network, rather than being stuck holding a single license it cannot use well on its own. That flexibility is what a one-shot sealed-bid auction cannot offer, and it is the specific problem the simultaneous ascending format was built to solve.

Turning air into treasury

The headline number is straightforward. FCC spectrum auctions have generated more than $52 billion in cumulative revenue since the program began in 1994, the Congressional Research Service reports, all of it flowing directly to the US Treasury as non-tax revenue. That distinction matters more than it sounds. Most large one-time asset sales happen once. Spectrum keeps generating new licenses to sell, because every advance in wireless technology, from 2G to 5G, opens bands that were previously unusable or reserved for other purposes, which turns what looks like a single windfall into a recurring public-finance mechanism that resets with each new generation of network technology.

It is also a form of revenue with an unusually captive buyer. A company cannot legally operate a mobile network without holding a license for the frequencies it uses, so the auction is not competing against a decision to simply do without. That inelastic demand is part of why spectrum sales have raised sums on the scale they have, and part of why the technique proved so exportable once other regulators saw the US results.

The global figures make the point plainly. Beyond the FCC's own $52 billion, the simultaneous ascending format has produced more than $200 billion for governments elsewhere, according to the Priceonomics account of the design's spread. A mechanism built to solve one country's licensing backlog turned invisible, unowned bandwidth into one of the larger recurring sources of public revenue that exists anywhere in modern government finance, without a single new tax being levied to collect it.

The auction model is not free of trade-offs, and the debate around it runs both ways. Selling licenses at auction, rather than granting them for a flat administrative fee, revealed spectrum's true market value for the first time, which is precisely what let the technique raise the sums it has. Critics of the model have argued the same market discovery can work against the public it is meant to serve: a company that pays billions for a license has less capital left to build the network coverage the license was meant to enable, and some of that cost is likely passed on to subscribers in what they pay for service. The auction, in other words, liberated spectrum's value from bureaucratic guesswork and concentrated a large share of that value in the treasuries collecting it, a trade-off with no single settled answer.

But the money is only one half of what a government buys itself when it controls a spectrum license. The other half is the power to decide who is allowed to hold one, and that second power is what turns spectrum policy from a revenue tool into a geopolitical one.

Spectrum allocation is a sovereign act

Radio spectrum is coordinated internationally through the International Telecommunication Union, a United Nations specialized agency founded in 1865 as the International Telegraph Union, whose World Radiocommunication Conferences set a global band plan describing which frequency ranges are broadly reserved for which purposes worldwide, whether aviation, maritime navigation, broadcast, satellite, or mobile service. National regulators such as the FCC then license spectrum domestically within that broader plan.

That international layer sets the outline, not the decision. Inside it, each country's own regulator decides who may actually transmit within its borders, a decision no international body and no neighboring country can make on its behalf. A road, a port, or a fiber cable can, in principle, be built and operated by whichever company wins the contract. A wireless network cannot exist at all until a national regulator grants the specific right to transmit, which makes spectrum licensing a genuinely sovereign act in a way few other pieces of infrastructure are.

That authority is also the underlying economic reason spectrum commands a price at all rather than being freely allocated like most communications infrastructure inputs. Spectrum is physically finite and cannot be stored: a given frequency band can carry only one non-interfering signal in a given geographic area at any one time, a basic fact of spectrum economics referenced throughout FCC and Congressional Research Service policy documentation. Scarcity is what makes the license valuable, and sovereignty over that scarce resource is what makes the license a government's to sell, or to withhold.

There is no supranational body that can overrule a national regulator's licensing decision the way, for instance, a trade tribunal can rule on a tariff dispute. The ITU's conferences produce agreement on the broad band plan, and member states have practical reasons to honor it, since a country that ignored the plan would cause interference for its own neighbors as much as for anyone else. But enforcement of who actually holds a license inside one country's territory stops at that country's own regulator. That is a narrower form of sovereignty than deciding tax policy or immigration policy, but it is sovereignty of the same kind: a decision only one government in the world is entitled to make.

Licensing as an instrument of exclusion

Because a national regulator alone decides who may hold a spectrum license, that same authority can be used to decide who may not. Multiple governments, among them the United States, the United Kingdom, Australia, and India, restricted or banned equipment made by the Chinese manufacturers Huawei and ZTE from 5G networks on national security grounds, in a wave of policy that swept through roughly 2019 and 2020. The mechanism was rarely a blanket import ban. It ran through the licensing conditions attached to spectrum access itself, specifying which equipment vendors a network operator was permitted to install in order to keep its license in good standing.

That is a quieter and more durable form of exclusion than a trade tariff. A tariff can be renegotiated in a single trade round. A licensing condition sits inside the ongoing relationship between an operator and its regulator, renewed and enforced continuously for as long as the network runs, which gives a government far more lasting control over which company's hardware ends up carrying a nation's mobile data.

The pattern spectrum set

Spectrum's physical scarcity, the simple fact that one frequency can only carry one non-interfering signal in one place at a time, is the reason it could be priced at all. Once regulators recognized that scarcity could be sold rather than merely rationed, the license itself became a chokepoint controlling two things at once: a government's revenue, and a government's choice of who is permitted to operate inside its own communications infrastructure. Neither function existed before 1993. Both now run through the same signature on the same license.

That double function is a pattern this history keeps repeating around every dominant medium: whoever controls the gate over a scarce channel gains an income from that control and a form of power over who is allowed to pass through it. It was true of the licensed printing press, true of the state-chartered telegraph concession, and it is true of the FCC-licensed frequency.

The newest version of that same gate is not a frequency band. It is a much smaller kind of shelf space: the handful of businesses an AI answer engine names when a person asks it a question. No regulator licenses that shelf space the way the FCC licenses a megahertz, and no treasury yet collects an auction fee for it. But the underlying shape is familiar, a scarce, mediated channel controlled by a small number of gatekeepers, where being left out is functionally close to never having existed for the person asking. Spectrum's history is a reminder that once a channel like that becomes valuable enough, someone eventually decides who gets to use it, and on what terms. That decision has not been made for the AI answer the way it was made for the airwaves in 1993. It is being made now, informally, one citation at a time.