Spectrum has a standard unit and decades of public auction prices. It still has no market. The binding constraint is transferability rather than valuation.

Compute lacks an agreed unit. Spectrum does not have that problem. The comparator is bandwidth across a geography for a term, conventionally quoted per megahertz per head of population covered. Regulators run public auctions that produce large, observable, legally binding prices, and analysts have built comparable tables from them for decades.

So spectrum has both a unit and prices. What it does not have is a market.

Auction prices are not reference prices

A band is awarded, the process runs for months, a price is struck, and then nothing happens in that band for years. That produces a data set rather than a curve.

Auction prices are also specific to their circumstances. They include the licence obligations attached to the lot, such as coverage commitments and build-out deadlines. They reflect the auction design, which is built to extract surplus rather than to reveal a clearing level. And they reflect strategic bidding among a small number of firms that will compete with each other for another twenty years. What comes out is a price for that lot, under those conditions, among those bidders.

A market needs a price that refreshes, and someone who will stand behind it between the auctions.

The constraint is transferability

Spectrum is licensed rather than owned. A licence is a right to use a public resource under conditions set by a regulator, and in most jurisdictions it cannot change hands simply because two parties agreed a price.

Secondary trading and leasing regimes do exist. The United States and the United Kingdom have both operated them for many years and they are genuinely used. But they run on regulatory timescales. Consent is required, review takes months, competition authorities may take a view, and licence obligations generally transfer with the licence. That is reasonable policy for a public asset. It is also incompatible with market making. You cannot quote two-way in something whose transfer requires third-party permission measured in quarters.

Satellite capacity is different

Buying capacity on a satellite is a commercial service contract with an operator rather than a licence transfer. It does not need regulatory consent to change hands. It is traded bilaterally today, usually on long-term take-or-pay terms negotiated deal by deal.

That is roughly where ocean freight sat before it had indices: a large and economically significant business conducted through brokered bilateral negotiation, with no continuous price and no way to hedge.

Supply is also no longer stable. Capacity used to be geostationary and effectively fixed, adjusting on a decade-long cycle. Large low-orbit constellations change that, and supply can now be added and re-pointed far faster while demand keeps moving. That produces price volatility where there used to be long-term contracts, and volatility is what creates demand for hedging.

Financial settlement is the unlock

If the underlying cannot be delivered freely, the practical route is to write contracts that do not require delivery.

Freight solved the same problem. Ships are lumpy, individually negotiated assets and chartering one is slow. The market became tradable when an independent body began publishing daily assessed rates for defined routes, collected from a panel of brokers, and participants started writing contracts that settled financially against those assessments. Forward freight agreements let an owner hedge without moving a vessel. The physical market stayed bilateral and slow, and a liquid financial market formed on top of it.

The same structure applies here, and it needs three things, none of them technical. A defined product: specified bandwidth, over a defined coverage region, for a defined period, at a stated quality. A published assessment of what that product is worth, collected from firms actually transacting. And standard terms for financial settlement against it. With those, economic exposure transfers without a licence transferring.

Why this is slow

An assessed index is an institution rather than a data product. It is credible only if contributors transact in real size, if the methodology holds up when a print is disputed, and if it survives publishing a number that a large participant dislikes. That takes years.

Spectrum and satellite are early, earlier than compute and considerably earlier than crypto was when we started trading it. The unit exists, the scarcity is real, demand is growing, and supply has become volatile enough to make hedging worth paying for. What is missing is reference pricing between the auctions, standard terms, and enough repeat bilateral flow for an assessment to mean something.

Hidden Lattice Capital · Field Notes