The Single Point of Access

On September 1, 2016, at 9:07 a.m. Eastern, a Falcon 9 sitting on SLC-40 at Cape Canaveral came apart during a routine static fire — a test, not a launch, the rocket bolted down and the engines meant to run for a few seconds against a clamp. The failure started in the second stage liquid-oxygen tank during propellant loading. SpaceX traced it to a composite overwrapped pressure vessel: buckles in the carbon fiber let super-chilled oxygen pool where it should not have been, and carbon fiber and solid oxygen are, it turns out, an ignition source. The pad burned. The launch complex was out of service for roughly fifteen months.

Bolted to the top of that rocket was AMOS-6, built by Israel Aerospace Industries, operated by Spacecom, insured, and destroyed. Eutelsat had contracted a multi-year lease on its Ka-band payload covering sub-Saharan Africa, with Facebook as partner, service scheduled to begin in early 2017. Mark Zuckerberg — in Africa that week — posted that he was deeply disappointed. Internet.org had a satellite, and then it did not.

The story got told at the time as bad luck, and it was bad luck. A pressure vessel does not check the ownership structure of the payload above it before it buckles. A publicly funded, openly governed connectivity satellite would have burned in exactly the same fire. Rockets fail. That is a fact about rockets, and it is not an argument about enclosure.

So the fire is not the evidence. It is the exposure. What it showed was everything standing behind the payload, and behind that payload there was nothing: a single satellite carrying a single leased payload under a single agreement, riding a single launch, to deliver one company's framing of the internet to a population with no second option. Not one of those singularities was physics. All of them were contract.

And the singularity that actually mattered was already dead by then, killed six months earlier by a regulator rather than a rocket. In February 2016, India's telecom authority banned differentially priced data services outright, and Free Basics went dark in the largest market it had entered. TRAI did not ban it for being fragile. It banned it for being a door with an owner.

That is the distinction an explosion cannot draw and the regulation draws cleanly. Free Basics was never mainly an argument about access paths. It was an argument about who authors the default shape of the internet for a person meeting it for the first time — which services simply are the internet, which cost extra, whose ranking decides what a billion people learn the medium to be. Redundancy is a resilience claim. That is an ownership claim, and it survives a counterfactual the resilience claim does not.

Because the counterfactual is unkind. The connectivity arrived anyway. Sub-Saharan Africa came online through undersea cable landings, through mobile operators laying towers and selling data at prices set against each other, through regional exchange points — boring, unbranded, incremental, redundant, and no more dependent on that Thursday morning in Florida than on the weather in it. Facebook shut down Express Wi-Fi in 2022. Free Basics has been wound down market by market. The connectivity that stuck is the kind nobody can name after a mission, because it was never one thing and had no single owner to name it after.

That is what a commons offers that an enclosure structurally cannot — not virtue, not fairness, but paths. The inefficiency people complain about in open systems, the duplicated effort and the competing standards and the absence of any single throat to choke, is the same inefficiency that means no one Thursday takes the whole thing out. You cannot sell the position of sole gateway and also hold redundancy in reserve. Redundancy is the competing route you spent the money to eliminate.

AMOS-6 was insured. Spacecom collected. SpaceX rebuilt, flew SLC-40 again fifteen months later, and now puts more mass into orbit than every other launch provider on the planet combined. The only party with no coverage was the one that never had a contract in the first place: the users on the other end of a plan they did not negotiate.

The satellite was the cheap part. The single path was the expensive part, and it was never on the manifest.

It is not on the current one either. A serious share of the new capacity reaching remote and underserved regions now arrives by low-earth-orbit satellite, and the largest of those constellations by an enormous margin is Starlink — built, launched, and operated by the company whose pad burned that morning. One firm. One owner. No operating peer at that scale. The party that had somewhere else to be is becoming the place everyone has to go.