IRUs and the Economics of Owning a Path

The instrument at the center of long-haul fiber leasing is the indefeasible right of use, universally abbreviated IRU. Understanding it is essential because it is the legal form through which control over a physical route changes hands without ownership of the underlying company changing at all.

An IRU is a long-term, exclusive, essentially irrevocable right to use a defined amount of capacity or a defined set of fiber strands along a route. Terms are long by the standards of most commercial contracts, commonly fifteen to twenty-five years, with twenty as a familiar anchor. The buyer typically pays a large sum up front, sometimes most of the lifetime value, plus an annual operations-and-maintenance fee that covers the carrier's obligation to keep the route physically intact, splicing cuts and maintaining the conduit. The word indefeasible is the load-bearing part: the right cannot be casually revoked for the duration, so the buyer can treat the capacity as a durable asset on its own books, close to ownership without owning the cable.

Why does this structure exist? Because building long-haul fiber is enormously capital-intensive and almost none of that cost scales with the number of strands. Trenching the route, securing the rights of way, and boring under rivers and highways is where the money goes; adding strands to a bundle you are already pulling is comparatively cheap. So the rational builder over-installs, laying far more strands than it needs, and sells the surplus as IRUs. The up-front payments recover construction cost quickly and de-risk the project, while the buyer gets near-ownership of a route it could never justify building alone. Both sides win, which is why the market for these contracts is deep and old.

Now the economics that matter for AI. When you sign a dark-fiber IRU, you lease specific glass strands and light them with your own optical equipment. The crucial property is that the capacity of a fiber pair is not fixed by the glass; it is set by the transponders on each end. Optical technology has repeatedly multiplied the bits per second a single strand can carry by upgrading only the endpoint equipment. A pair that carried a few hundred gigabits a decade ago can, with modern coherent optics, carry tens of terabits over the identical glass. So the owner of a dark-fiber IRU captures every future capacity gain for the cost of new transponders, while the strand and the twenty-year right sit unchanged. You buy the path once and ride the technology curve for free. For a party whose traffic grows relentlessly, this is a powerful structural advantage over renting lit capacity whose price the carrier resets as your usage climbs.

Contrast the two cash-flow shapes, because the difference is the whole decision. Lit capacity is an operating expense: a recurring monthly charge per wavelength or per circuit, scaling with how much you buy, cancellable and flexible but permanently a cost that rises with consumption and carries the carrier's margin on every incremental bit. A dark-fiber IRU is closer to a capital asset: a large upfront payment that depreciates over the term, plus a modest maintenance fee, after which incremental capacity costs only endpoint electronics. The first suits uncertain or modest traffic; the second suits enormous, predictable, growing traffic, which is precisely the profile of frontier AI operators. The instrument you choose encodes a forecast of your own demand.

My read: this is why the center of gravity for the largest operators is shifting toward the strand. When your traffic is a near-certain up-and-to-the-right line, renting bits means paying a carrier's margin forever on a rising base and leaving future optical gains on the table. Owning a twenty-year path converts that into a depreciating asset you control, and it does something the spreadsheet undervalues: it locks in a specific diverse physical route before a competitor secures it. On a corridor with only two independent rights-of-way, the operator who signs IRUs on both has not merely bought capacity, it has bought scarcity and denied it to rivals. That denial is a strategic asset that never shows up in a per-bit price comparison.

There are constraints, and a serious model prices them. An IRU ties up capital for decades and presumes the route stays strategically relevant, a safe bet on major corridors, a riskier one on secondary ones. Lighting dark fiber requires optical competence, either in-house or contracted, that not every operator possesses. The maintenance obligation means you still depend on the carrier's crews to fix cuts, so the counterparty's operational quality still matters even after you own the glass. And the recurring operations-and-maintenance charge is where the long-tail cost hides; read it carefully. These are the reasons lit capacity persists for many buyers even as the giants move toward dark.

The durable takeaway: an IRU is how control over a physical route is transferred for the long term, and the dark-fiber IRU in particular lets a buyer own a path and ride the optical technology curve while locking in diversity. When you see a hyperscaler signing long-term dark-fiber deals along a corridor, read it as a claim on scarce routing, not a routine procurement. The next lesson maps the carriers who own those corridors and sit on the other side of these deals.