The Filing
SpaceX asked the Federal Communications Commission on July 6, 2026 for authority to build and launch up to 100,000 third-generation Starlink satellites — a system that would sit on top of, not replace, the Gen2 constellation the FCC has authorized in two tranches: 7,500 satellites in December 2022, and a further 7,500 in a January 9, 2026 order that brought the total Gen2 ceiling to 15,000. The Gen3 request is nearly seven times that number. The satellites would fly in two closely stacked very-low-Earth-orbit shells, at roughly 325 km and 475 km — well below the 550 km Gen1 shell — and each would weigh an estimated 2,000 to 2,500 kg, heavy enough that deployment depends on Starship reaching routine flight, since Falcon 9 cannot lift Gen3-class satellites at any useful cadence. SpaceX also asked to expand its spectrum rights beyond the Ku-, Ka-, V- and E-bands Gen2 already holds, deeper into W-band and, for the first time, D-band frequencies spanning 92 to 275 GHz, framing the request around backhaul capacity for AI-scale data demand rather than consumer broadband alone.
Why Regulators Shouldn't Rubber-Stamp This
The skeptics have a real case, and it deserves to be stated plainly before it's answered. A constellation an order of magnitude larger than anything now in orbit raises collision-avoidance math that gets worse, not better, with scale: researchers have estimated the annual probability of at least one collision involving a Starlink-sized constellation already exceeds 10 percent even with active maneuvering, and Starlink satellites already perform tens of thousands of avoidance maneuvers a year as low-orbit traffic multiplies. Radio astronomers, through the International Astronomical Union's Centre for the Protection of the Dark and Quiet Sky, have documented unintended radio emissions from existing Starlink satellites bleeding into frequencies reserved for passive science, and W/D-band operations sit close to spectrum radio telescopes and weather satellites also depend on. The sheer count matters systemically too: roughly 14,000 satellites are active in orbit today, against an estimated 1.23 million proposed satellite projects now moving through regulatory pipelines worldwide — exactly the kind of unprecedented queue that case-by-case licensing was never built to absorb quickly.
Why the FCC's Incremental Model Still Fits
None of that argues for blocking the filing; it argues for the review process the FCC has already been running. The Commission approved Gen2 in exactly the halting, evidence-gated way critics say Gen3 needs: it granted 7,500 satellites in 2022, deferred the remainder of SpaceX's roughly 30,000-satellite request, then granted another 7,500 in January 2026 only after reviewing operational data from the first tranche — explicitly choosing to proceed incrementally rather than approve the full ask at once. There's no reason to expect Gen3 gets different treatment, and a 100,000-satellite ceiling in a filing is a request for eventual authority, not a launch order. The debris-mitigation floor is also already in place independent of how large any single constellation grows: the FCC's five-year post-mission deorbit rule, in effect since September 29, 2024, replaced the old 25-year guideline and binds every new licensee regardless of fleet size, so Gen3 satellites inherit the same disposal obligation Gen2 satellites do. The timeline defuses the urgency further: full deployment requires Starship flying at a cadence it hasn't yet demonstrated — likely well over a thousand individual launches — giving the FCC, NSF and outside commenters years of real operational data before deployment could conceivably approach six figures. Front-loading the paperwork now, while gating the physical buildout to demonstrated launch capacity and continued deorbit compliance, is the proportionate way to regulate a technology moving this fast — better than either waving it through unconditionally or freezing the docket until every hypothetical harm is resolved in advance.
The Global Stakes
This isn't a US-only story. Very-low-Earth-orbit slots and W/D-band spectrum are scarce internationally, and China is racing to claim its share: the Guowang and Qianfan constellations, with a combined roughly 27,000 satellites planned, are explicitly modeled on Starlink and competing for the same orbital shells and the same unconnected markets across Southeast Asia, Africa and Latin America. A US regulator that takes years to process paperwork on a system Starship can't yet build at scale isn't protecting anyone from additional risk — it's ceding the queue to operators whose home regulators move faster and ask fewer questions. The FCC's Gen2 record shows it can gate real deployment to real evidence without freezing the licensing pipeline that lets American systems compete for orbital and spectrum resources that will not sit unclaimed while the review runs.