Can They Do Consumer Internet In Cities

Can They Do Consumer Internet In Cities?

Bottom line: today, yes — but only as a niche, and the niche is bounded by orbital geometry, not by capital. Starlink already has more subscribers in urban areas than in rural ones in five US states (FL, MA, HI, NJ, CT) per Ookla's H2-2025 broadband report. It is also actively rationing that urban demand: residential waitlists, "sold out" cells across Seattle/Portland/San Diego/Sacramento/Austin/London in late 2024, and a one-time congestion surcharge of $250–$1,500 in specific high-demand cells through early 2026. V3 satellites (1 Tbps downlink each, ~10× V2 Mini) plus 60 Tbps of capacity per Starship launch widen that niche in 12–24 months — perhaps from 5 states to 15–20 — but they do not change the fundamental physics that an urban cell has 100–1,000× the household density of a rural one and shares the same per-cell capacity envelope. Treat "consumer internet in cities" as a real LTV pocket worth a single-digit-percent share of the bull-case sub trajectory, not a regime change that re-rates Connectivity.

Sources: SpaceX S-1 (May 20, 2026); Ookla H2-2025 US Broadband Performance Report (via Light Reading, Apr 9, 2026); Starlink Residential pricing $55/$85/$130 per SatelliteInternet.com 2026.

The Real Question — Which "Urban Consumer" Are We Talking About?

The decision question is not whether a Starlink dish in a New York apartment works — physics says it can. It is whether SpaceX's $11.4B Connectivity engine, valued at $600B–$900B in the bull SOTP, can pull a meaningful slice of the ~140M US/EU urban broadband households away from gigabit fiber and cable. The answer is different for every urban customer type, and conflating them is where most analysis breaks.

Three of these five rows — primary, MDU, urban-edge — get conflated in casual analysis. The bull case rests almost entirely on row four (urban-edge / exurban) plus international urban-edge equivalents, with a sliver of rows two and three. The bear case assumes the company has been quietly trying for row one and will be exposed as having lost to fiber. Neither read is correct: the company is not trying to bid for urban primary, but it is not getting the credit for the urban niche it actually occupies, which is rows 2–4 in real demand and largely capped by physics in row 1.

The Physics Ceiling

A Starlink "service cell" is approximately a 20 km (12 mi) wide ground footprint where one or more satellites' beams share a capacity envelope. Per Wikipedia's Starlink article (citing SpaceX's own waitlist disclosures), sales have historically been capped at "a few hundred fixed users per 20 km service cell area" because the per-cell wireless capacity is finite. A 20 km cell is roughly 300 km². Urban household density runs 100× to 1,000× the cell-capacity ceiling, depending on the metro.

How to read this chart and table. The dark blue bar is the urban household pool that exists. The green bar is the user pool Starlink can serve in that cell after V3 multiplies per-cell capacity 10× off the V2 baseline (assuming a 20 km cell with the historical ~400 user cap, scaled to 4,000 on V3). The structural penetration ceiling in NYC is roughly 0.12% of households even at the optimistic V3 number; in Los Angeles it is 0.41%; in San Francisco 0.18%; in Chicago 0.28%; in Miami 0.26%. Move from a rural cell where Starlink can carve out double-digit penetration on the same hardware, and the density crunch is plain.

This math is approximate but the order of magnitude is the point. Two refinements an analyst can layer on top:

• Per-cell capacity scales not just with per-satellite throughput but with the number of satellites in view at any moment. As the constellation grows from ~9,600 sats (Mar 31, 2026) toward the 30,000+ filed-for limit, more sats blanket each cell, which improves the user ceiling. This is captured in the 10× V3 assumption but could be 20–30× over a 36–48 month window if Starship cadence cooperates and the FCC clears the expanded V3 constellation.
• The user cap is a priced cap, not a hard one. SpaceX's congestion surcharge ($250–$1,500 in select US cities through Jan 2026) is the price mechanism that rations urban cells without explicitly waitlisting. That tool widens urban capacity at the upper end of price discrimination but only modestly grows units sold.

The blunt take: even with a generous V3 step-up, no major US metro is structurally serviceable to more than ~1% of households via Starlink. Compare that to fiber and cable, which serve double-digit-to-near-100% of urban households as their normal addressable market.

The Evidence — Today's Urban Niche Is Real but Small

The single cleanest piece of evidence on urban demand is Ookla's H2-2025 US Broadband Performance Report, summarized by Light Reading on Apr 9, 2026: five states have more Starlink subscribers in urban areas than in rural areas — Florida, Massachusetts, Hawaii, New Jersey, and Connecticut. The framing from Ookla's editorial director (paraphrased in Light Reading): "Starlink's appeal as a broadband provider goes beyond just rural areas where fixed broadband options are non-existent. Starlink appears to be carving out a niche market for urban dwellers too." That word "niche" carries the weight.

Pricing rationing through congestion surcharges is the second piece of evidence. In Nov 2024 the waitlist was reintroduced in Seattle, Spokane, Portland, San Diego, Sacramento, Austin, and London (Citizen Watch Report citing Ars Technica). By early 2026 the explicit waitlist had been replaced by a "demand surcharge" of $250 to $1,500 for new residential customers in congested cells, with the highest charges in Alaska and the Pacific Northwest (satelliteinternet.com, Jan 27, 2026). The signal is the same: urban demand is exceeding cell-level capacity, and SpaceX is rationing by price.

Two reads from this. First, the demand for residential Starlink is real even in markets where fiber and cable exist — otherwise the surcharge would not exist. Second, the cells where SpaceX is sold out or surcharged are mid-tier metros (Seattle, Portland, Austin), not the densest metros (no NYC, no LA on the 2024 sold-out list). That pattern is exactly what physics predicts: the marginal urban household most likely to buy Starlink lives somewhere that mixes higher-than-rural density with weaker-than-fiber alternatives. Genuine NYC or LA households mostly have fiber or cable choices Starlink cannot match on price or latency.

How Starlink Compares to the Urban Stack

Even where Starlink can serve a city household on capacity grounds, it is competing against gigabit fiber at $60–$80/month with sub-10ms latency and against fixed-wireless 5G at $50–$70/month. The pricing table makes the gap explicit.

Sources: Starlink — SatelliteInternet.com 2026 plan disclosures; S-1 confirms 225 Mbps peak-hour median and ~25ms latency for the constellation as of Mar 31, 2026. Verizon Fios, Comcast Xfinity, ATT Fiber — retail published rates. T-Mobile 5G Home Internet — T-Mobile.com plan page 2026; T-Mobile markets itself as "America's fastest growing home internet provider," with the All-In plan at $65/month and a separate $20/month Backup tier.

The chart tells the substitution story in one frame. For the same ~$80/month an urban household pays for 1 Gbps fiber, Starlink offers ~200 Mbps with 3–4× the latency. Starlink wins on availability (anywhere in 99% of the US per FCC National Broadband Map) and resilience (one-link backup, no last-mile cut risk), and loses on speed-for-the-dollar against any fiber footprint. The 5G fixed-wireless tier — T-Mobile 5G Home Internet — is the more direct urban peer than fiber, and Starlink loses to it on price-per-Mbps in the markets where T-Mobile's 5G is dense.

What V3 Changes — and What It Doesn't

V3 is the next 12–24 months' lever. The headline numbers, from the S-1 and basenor.com's May 22, 2026 spec sheet, are real and material.

Source: SpaceX S-1 ("we expect to commence deploying our next-generation V3 satellites, designed to offer one Tbps of downlink capacity per satellite"); basenor.com May 22, 2026.

The honest read on V3 for urban consumer broadband:

• What it changes. Per-cell user ceiling rises roughly an order of magnitude on the V2-to-V3 transition, assuming roughly the same beam geometry. That expands the urban niche from ~5 states' worth of urban-tilted demand toward maybe 15–20 states, and lowers the price-rationed congestion surcharge in mid-tier metros. It also makes the apartment / MDU use case (single dish serving a whole building) viable on bandwidth grounds where it is constrained today. Existing terminals will need a hardware upgrade to consume the higher V3 speeds (per SpaceX disclosures), creating a one-time CapEx step for any user wanting the gigabit ceiling.
• What it does not change. Cell area is still ~300 km² and urban household density still runs 1,500–10,500/km². A 10× capacity step at the cell level still leaves urban penetration capped well under 1% of households in any major US metro. The bottom-line geometric argument — that LEO physics distributes capacity across area, while urban demand concentrates capacity per area — is unchanged.
• What gates the V3 timing. Two non-trivial dependencies: (a) FCC approval of the expanded V3 constellation (ongoing), and (b) Starship cadence (V3 satellites are launched on Starship, not Falcon 9, and a single Starship V3 launch is designed to add ~60 Tbps vs ~3 Tbps for a Falcon 9 with V2 Minis). Starship V3 first orbital flight was Flight 12 on May 22, 2026; commercial reuse on customer payloads is the gating event the IPO mark already implicitly assumes, per the variant tab. A V3 slip into 2027 with no commercial payload defers this urban-niche expansion by a full year.

Direct-to-Device Is Not Home Broadband in Cities

The EchoStar AWS-4 / H-block spectrum transfer cleared the FCC on May 12, 2026 (Order DA-26-471, ~$17B). That unlocks Starlink Mobile / Direct-to-Device at commercial scale, with first phone-direct tests in late 2026 (Teslarati, Feb 5, 2026 citing Reuters and Musk on X). The temptation is to count D2D as a stealthy backdoor into urban consumer broadband. It is not.

D2D is a parallel product, not a flanking move into urban primary broadband. Wikipedia's Starlink page summarizes the cellular service capacity as "2 to 4 megabits per second total, split across a very large cell coverage area" — that is shared bandwidth for the whole cell, before any per-user fairness. People in cities will use D2D as cellular fallback (dead-zone coverage, disaster), not as home internet replacement. Conflating D2D with home broadband inflates the urban TAM by an order of magnitude on top of the Connectivity bull case and should be discounted in any urban-internet read.

What This Means for the $600B–$900B SOTP

The Connectivity bull case carries the segment from $11.4B FY25 revenue and 10.3M subs toward roughly $22B and ~25M subs at FY28, then ~$35B and 40M+ subs by FY31 (per the long-term-thesis tab and the segment compound path in the business tab). That trajectory does not explicitly say where the next 30M subs come from. If the implicit answer is "urban substitution in developed markets," the physics layer above caps it. If the implicit answer is "international + emerging-market urban-edge + rural," the trajectory is achievable but its terminal value lives in lower-ARPU geographies that compress blended ARPU further (consistent with Q1-26's $66/mo, down from $91/mo a year earlier).

The sensitivity that matters: what fraction of the bull case's incremental subs come from urban consumer broadband in mature markets?

The bull case as written in the long-term-thesis tab effectively sits between the first and second rows above — modest urban exposure, blended ARPU stabilizing around $55–$60 once mix-shift settles. The stretch row is what a "Starlink is unlocking urban consumer broadband" thesis would require, and it depends on simultaneously (a) V3 hardware delivering its full per-cell capacity step, (b) Kuiper failing to subsidize a sub-$300 terminal, and (c) Starlink not being preempted by MDU exclusivity contracts. The bear row is what happens if urban demand stays niche AND Kuiper actively breaks ARPU.

The non-obvious read for a PM: the bull SOTP does not need urban consumer to "work" to clear $600B. It needs international urban-edge + rural to compound and enterprise/aviation/maritime/Starshield to stay wide-moat. Treat urban consumer in cities as 5%–10% of the net-add story, not the swing variable. The variant tab is right that Kuiper terminal price and Starshield disaggregation are the higher-decision-value signals than urban consumer share.

Two Signals to Watch

If you want to know which side of this bet is winning over the next 12 months, two leading indicators dominate.

The asymmetry between the two signals is intentional. The break-out signal requires a quantitative and a qualitative condition to hit simultaneously — that is a high bar and a meaningful regime change if it does. The break-down signal merely requires the status quo to continue, which is the base case. The urban-cities question is therefore a low-probability, high-impact source of upside relative to the bull SOTP, not a structural threat to the base case.

Where We Could Be Wrong

A few honest limitations on the analysis above.