Starlink hit 12 million users and a record IPO. Here's the funding fight, spectrum war, and stock slide left out.
SpaceX listed on the Nasdaq on June 12, 2026, at a valuation of nearly 2 trillion dollars. By late July, the stock had fallen below its own IPO price. Starlink passed 12 million subscribers, lost a federal subsidy fight for the second time, and quietly won a separate spectrum war with Dish. Here is what actually happened, verified against primary sources, with no hype and no filler.
The China vs. Starlink story everyone got wrong
The headline that circulated in mid-2025 was simple: China beats Starlink. The actual story is more specific, and the more important development received far less coverage.
In June 2025, researchers from Peking University and the Chinese Academy of Sciences published findings in the journal Acta Optica Sinica. Their satellite, parked in geostationary orbit 36,000 kilometers above Earth, transmitted data at 1 gigabit per second using a 2-watt laser beam. That rate runs roughly five times faster than speeds typically measured on Starlink's residential service. A 2-watt laser, for context, is weaker than the bulb inside a refrigerator.
It is an impressive engineering result, but not a direct contest. Starlink operates from low Earth orbit, a few hundred kilometers up, using radio antennas to serve individual homes. The Chinese test came from a geostationary orbit more than 60 times higher and relied on a specialized ground station with a 1.8-meter telescope at the Lijiang Observatory. Consumer Starlink dishes are roughly the size of a laptop lid. The comparison is real. The competition is not, since the two systems solve different engineering problems for different customers.
The geostationary orbit also adds latency, though it offers consistent global coverage from a single orbital slot. One demonstration does not replace the dense coverage of a network like Starlink, and researchers themselves have noted the technology suits high-capacity backbone links more than direct consumer broadband for now. If you are trying to separate genuine technical breakthroughs from overstated AI and infrastructure claims more broadly, our look at the fine print behind AGI claims covers a similar pattern of an impressive lab result getting inflated into a market-shifting story.
What Starlink is, and how it actually works
Starlink is SpaceX's satellite internet service, one piece of a much larger shift in how internet access reaches people who have never had a cable or fiber line nearby. It operates a constellation of low Earth orbit satellites at roughly 550 km altitude. That proximity to Earth is the basis for everything the service does well, particularly the latency that made older satellite internet effectively unusable for real-time applications. For a broader look at how wireless networks carry data behind the scenes, our guide to the invisible infrastructure behind wireless communication breaks down the layers most people never think about.
Older geostationary satellite services like HughesNet and Viasat carry latency up to 600 ms, because signals must travel 36,000 km to reach the satellite and return. Starlink's satellites orbit at roughly 550 km, cutting that round trip to 25 to 60 ms. Fiber internet sits at 11 to 14 ms. The gap between Starlink and fiber matters for competitive gaming and real-time financial applications. The gap between Starlink and legacy satellite internet matters for nearly everything else: video calls, remote work, and basic web browsing all become usable at 50 ms in ways that 600 ms simply prevents.
Starlink added roughly 4.6 million new subscribers in 2025 and crossed 10 million total in February 2026, adding 1 million customers in just 53 days. By June 2026, the count had passed 12 million across more than 160 countries and territories, according to Starlink's own milestone announcements and independent tracking.
Space weather is the cause that most reviews never mention
Consumer reviews blame trees, roofs, and rain for Starlink dropouts. Almost none mention the atmosphere above the troposphere, even though it causes outages that look identical to a hardware problem from the user's side.
In February 2022, a geomagnetic storm increased atmospheric drag just enough to pull dozens of newly launched Starlink satellites back down before they ever reached operating altitude. That was an extreme case tied to solar activity, but milder versions of the same effect happen constantly near the peak of the roughly 11-year solar cycle, which last peaked around 2024 and 2025.
A separate and more persistent issue is ionospheric scintillation, a scattering of radio signals that occurs most often near the geomagnetic equator, across parts of South America, equatorial Africa, and Southeast Asia. A user in these regions can have a flawless install with a completely clear sky and still see throughput swings that have nothing to do with the dish, the router, or local congestion.
The federal funding fight nobody covers as one continuous story
Most 2026 Starlink coverage treats government broadband subsidies as settled and favorable. The real history runs through two different federal programs, a rejection, an appeal denial, a total policy reversal, and a still-unresolved argument in Congress. Reading it as one continuous timeline changes how the whole subsidy question looks.
In August 2022, the FCC denied SpaceX's application for $885.5 million over 10 years through the Rural Digital Opportunity Fund, a reverse-auction subsidy program meant to connect roughly 643,000 rural locations across 35 states. The Wireline Competition Bureau found Starlink had not demonstrated it could reliably deliver the required 100/20 Mbps speeds, citing declining Ookla test results from 2021 and 2022, and separately flagged that the plan depended on an unproven Starship rocket for future satellite launches. SpaceX appealed. In December 2023, the FCC reaffirmed the rejection, over dissents from two Republican commissioners, and stated plainly that Starlink was not reasonably capable of offering the promised service at the time. A further review in 2024 upheld that conclusion again, and the FCC's own account of the decision noted the applicant had failed to meet its burden for nearly $900 million in universal service funds.
Then the policy environment changed entirely. In 2025, Commerce Secretary Howard Lutnick restructured the separate, much larger $42.45 billion Broadband Equity, Access, and Deployment program, moving it from a fiber-first framework to one the administration called technology neutral. That single rule change did what years of appeals could not: it opened federal subsidy eligibility to satellite providers on equal footing with fiber. Starlink applied across the country and, according to reporting from the Benton Institute, won BEAD awards covering 476,000 locations, more than any other single provider, worth roughly $738.8 million. Colorado alone saw Starlink and Amazon's Kuiper jointly request around $300 million after the rule change forced the state to restart its entire BEAD allocation process.
The story is not fully resolved even now. SpaceX later asked state broadband offices for exemptions from standard BEAD performance and reporting rules, including a proposal to receive half its award upfront rather than tied to subscriber signups, and the NTIA advised states against signing that rider. More than 20 House Democrats separately wrote to the NTIA in April 2026 arguing some states may need to reconsider their SpaceX awards altogether, and Ookla data cited in that same debate showed only about 45 percent of U.S. Starlink subscribers met the BEAD minimum speed threshold by the end of 2025, up quickly from 17 percent at the start of the year but still short of half.
Starlink plans and pricing in 2026: what each plan actually costs
Pricing changed again in June 2026, and the figures below reflect the most recent verified billing update as of publication, cross-checked against current Starlink account notices and independent pricing trackers.
Residential plans
| Plan | Monthly price | Typical speed | Hardware cost |
|---|---|---|---|
| Residential 100 | $55/mo (was $50) | 100 Mbps | $349, or $10/mo rental on new orders |
| Residential 200 | $85/mo (was $80) | 200 Mbps | $349, or $10/mo rental on new orders |
| Residential MAX | $130/mo (was $120) | Max throughput | $349, or $10/mo rental on new orders |
Starlink Roam for RV and travel use
| Plan | Monthly price | Data | In-motion |
|---|---|---|---|
| Roam 100GB | $55/mo | 100 GB | Yes, up to 100 mph |
| Roam 300GB | $80/mo New tier | 300 GB | Yes, up to 100 mph |
| Roam Unlimited | $175/mo | Unlimited | Yes, pausable monthly |
All Roam plans can be paused, which lowers the effective annual cost for seasonal users. The pause works in monthly increments only. If you miss your billing date by a single day, you pay for the full month. Confirm pause status inside the account app after completing the process, not just after tapping the button.
Starlink Mini vs. Standard: which hardware to choose
The Starlink Mini is a hardware option, not a separate service tier. At 298.5 by 259 by 38.5 mm and 1.1 kg, it fits in a backpack. The kit costs $249.99, with occasional promotions bringing the price down for new Roam customers. It includes built-in Wi-Fi 5, draws 25 to 40W on average, carries an IP67 water resistance rating, and handles winds above 96 kph and snow loads up to 25 mm per hour.
The standard dish at $349 offers higher throughput in congested areas due to its larger aperture. For stationary home use where performance takes priority, the standard dish is the better choice. For backpacking, overlanding, small boats, or travel where weight and packability matter, the Mini is the right hardware.
Subaru Starlink: an important naming distinction
Subaru Starlink is a completely separate product from SpaceX's satellite network. It is Subaru's connected services and safety platform built into many of its vehicles, covering automatic collision notification, SOS emergency assistance, remote vehicle access through an app, and an optional cellular Wi-Fi hotspot. It has no connection to low Earth orbit satellites. Most new Subaru vehicles include a three-year trial of the safety and security package, with renewals typically running $99 to $199 per year, depending on tier.
T-Mobile, Starlink Mobile, and what direct-to-cell actually delivers
In 2025, SpaceX and T-Mobile launched a direct-to-cell service, branded Starlink Mobile, that lets standard smartphones connect to Starlink satellites in areas with no terrestrial signal. Speaking at Mobile World Congress in March 2026, SpaceX's Michael Nicolls said the brand had surpassed 10 million subscribers and that more than 16 million unique users had connected across the broader Starlink constellation, including during emergencies. Gwynne Shotwell described how Starlink deployments kept more than 400,000 people connected during the early 2025 Palisades Fire in Los Angeles, sending over 250,000 text messages and more than 150 wireless emergency alerts, and provided similar support after the 2025 Kamchatka earthquake and the Aomori earthquake in Japan. As of January 2026, more than 650 specialized direct-to-cell satellites had been launched to support the service. For context on how mobile carriers and handset makers are competing for this same connectivity layer, our coverage of the Trump Mobile T1 phone and its deposit numbers shows just how crowded the broader mobile market has become.
Image Credit: Leonardo AI
Crossing borders with Starlink: where the kit actually stops working
Travel and RV guides often sell Starlink Roam as something that works anywhere on Earth. The licensing reality sits in the service agreement, not the marketing page, and it surprises far more buyers than the obstruction issue does.
Starlink uses GPS-based geofencing tied to country-level licensing. The kit can refuse to authorize even with satellites overhead if the country has no active license for the service. A traveler who buys Roam in a licensed country and then crosses into one without a license gets a connection that will not authorize, not a weaker signal.
Reporting on Starlink's presence in conflict zones and sanctioned countries, including Sudan, Myanmar, and Iran, has documented confiscation and fines as a real risk in some jurisdictions. Iranian authorities have acknowledged tens of thousands of unauthorized terminals operating in the country and have used jamming equipment to disrupt the signal during periods of unrest. The coverage map on Starlink's website shows the legal boundary of the service, not the satellite footprint. The satellites themselves cover nearly the whole globe. Authorization to connect to them does not.
| Border or jurisdiction status | What actually happens |
|---|---|
| Licensed country | Roam connects normally. Standard data caps and fair use rules apply as in any other licensed market. |
| Pending or partial license | Some users report intermittent authorization, often resolved within weeks of the license being finalized, as happened in Vietnam in February 2026. |
| No license, civilian use | The terminal authenticates with the network but is blocked from connecting. No internet access, regardless of signal strength. |
| Sanctioned or restricted country | Possession or use can carry legal risk, including confiscation or fines, separate from the technical authorization issue. |
When Starlink fails: the people who actually need it most
Most buying guides are written to convert readers. This section is not. Knowing exactly where Starlink underdelivers can prevent a $350 hardware purchase that does not work at a given location.
Dense tree canopy is the leading cause of returns. The dish requires roughly 100 degrees of unobstructed sky. A single pine branch crossing that arc creates repeated brief dropouts, not occasional ones. The Starlink app includes an obstruction checker that visualizes precisely where local geometry will interrupt the signal. Run it before purchasing, not after mounting the dish.
High-density apartment buildings present a different constraint. When dozens of dishes in the same building point at the same orbital cell, congestion accumulates. SpaceX does not advertise this limitation. Users in dense urban apartment blocks have reported evening speeds falling to 15 to 30 Mbps, no better than a mid-tier cable plan at a higher monthly cost. For a broader look at how ISPs frame these tradeoffs and where the marketing diverges from the fine print, our breakdown of common network device myths and ISP traps covers several patterns that apply just as much to satellite providers as to cable companies.
Remote healthcare and emergency services remain a category where Starlink's limitations are disqualifying rather than inconvenient. A 60 ms average latency with occasional outages does not meet the reliability standard required for remote surgical guidance, real-time emergency dispatch, or life-critical monitoring equipment. These applications require uptime guarantees that come with dedicated fiber or leased-line contracts, not consumer satellite subscriptions.
Mountainous terrain creates orbital geometry gaps. Even with an unobstructed dish, certain mountain regions have periods, sometimes predictable hourly gaps, where no satellite is in range. These appear on SpaceX's own coverage maps, but buyers in affected zones frequently discover the issue only after purchase.
HOA and lease restrictions create real legal barriers in many markets. In the United States, FCC rules protect the right to use satellite dishes under one meter in diameter, but common area restrictions, roof access requirements, and drilling prohibitions in rental properties mean that many dense urban buildings simply cannot mount a dish with adequate sky access, regardless of the regulatory framework.
The micro-outage a speed test will never show you.
A single speed test produces one number. It hides the structural, second-by-second behavior of how a low Earth orbit network actually serves a connection, and that gap explains a category of complaints that generic reviews cannot.
Each Starlink satellite serves a user for only a few minutes before handing the connection to the next satellite passing overhead. That handoff causes brief jitter or packet loss, typically a fraction of a second up to a couple of seconds. This is different from an obstruction drop caused by a tree branch and different from a congestion slowdown caused by too many users sharing one cell. The handoff gap is structural. It happens even on an empty network with a flawless install, because the satellite itself is always in motion relative to a fixed point on the ground.
This matters most for voice calls and competitive gaming, where a half-second gap is noticeable, and matters far less for web browsing or downloads, where it is invisible. SpaceX has reduced handoff disruption through higher satellite density and better beam scheduling as the constellation has grown, but the gap cannot be fully removed while satellites keep moving relative to a fixed ground point.
| Outage type | Typical duration | User fixable |
|---|---|---|
| Satellite handoff jitter | Under 1 second to a few seconds | No, structural to LEO networks |
| Obstruction (tree, roofline) | Seconds to minutes, repeats at the same time of day | Yes, reposition the dish |
| Local cell congestion | Minutes, usually evenings | Partially, priority plans help |
| Rain fade or heavy weather | Minutes during the storm | No, weather-dependent |
| Ionospheric scintillation | Variable, tied to latitude and solar activity | No, geography dependent |
The Starlink app's own statistics page shows ping spikes and outage timestamps in enough detail to distinguish a handoff event from an obstruction event once you know the pattern: handoff jitter is brief, evenly spread through the day, and present even when the obstruction checker shows a clear sky.
HughesNet vs. Starlink: the honest comparison
HughesNet operates from a geostationary orbit and carries 500 to 600 ms latency. Starlink sits at 25 to 60 ms. As of early 2026, both HughesNet and Viasat measured above 670 ms in Ookla's data, a figure that makes those services genuinely unsuitable for video conferencing, not merely inconvenient for it.
HughesNet retains a role in areas where Starlink is congested, where equipment cost is a barrier, or where subsidy programs apply. HughesNet hardware tends to be cheaper upfront because carriers subsidize it, and some federal rural broadband programs in the United States have historically covered HughesNet installations. Starlink hardware begins at $249.99 for the Mini and $349 for the standard dish, paid out of pocket with no subsidized equipment option through SpaceX directly, outside of a BEAD-funded install.
For most rural households comparing the two services in 2026, Starlink's latency advantage is decisive. The main scenario in which HughesNet remains the practical choice is one where a household qualifies for a subsidy program covering the hardware cost, and the monthly service fee fits within their budget.
Starlink Gen 3, now called V3: what actually happened in July
Starlink's third-generation satellites, which SpaceX now refers to as V3, are designed to carry more than one terabit per second of downlink capacity and 160 to 200 gigabits per second of uplink, roughly 10 times the downlink and 24 times the uplink of second-generation hardware. Each satellite weighs around 2,000 kg, uses argon Hall-effect thrusters, and adds E-band capability alongside laser inter-satellite links. SpaceX has said each V3 launch via Starship could add roughly 60 Tbps of capacity to the network, more than 20 times what a single Falcon 9 launch contributes today.
Starship's V3 configuration flew its debut test on May 22, 2026, carrying simulator payloads rather than working satellites. The real milestone came on July 24, 2026, when Starship Flight 13 lifted off from Starbase, Texas, and deployed 20 operational Starlink V3 satellites for the first time. SpaceX confirmed on social media that it made contact with all 20 satellites shortly after deployment. Because the flight followed a suborbital trajectory rather than an operational orbital one, each of those 20 satellites burned up in the atmosphere about 20 minutes after release, a planned outcome for a test mission, not a failure. The Super Heavy booster, however, suffered a landing failure for the second straight V3 flight, with only 5 of 13 planned landing-burn engines firing on splashdown, even as the upper stage itself achieved what SpaceX called its softest ocean touchdown yet.
What the marketing page leaves out
| What buyers commonly believe | What is actually true |
|---|---|
| Latency is now close to fiber. | Starlink's 25 to 60 ms versus fiber's 11 to 14 ms. The gap matters for competitive gaming and algorithmic trading. Fiber's latency is also more consistent. Starlink's range is wide, not just the average figure. |
| V3 is fully operational and will deliver 1 Gbps to everyone soon. | V3 satellites first flew and communicated successfully on July 24, 2026, but the Starship booster that carries them has now failed its landing attempt twice in a row. The 1 Gbps promise depends on a rocket program that has proven the payload but not the reusability. |
| Direct-to-cell replaces your carrier in dead zones. | Starlink Mobile supplements coverage for text and a growing set of data-enabled applications. Voice was added in 2025 but remains app-dependent. It functions as a backup layer, not a carrier replacement. |
| Maritime plans cover the whole ocean. | Ocean Mode, which covers areas beyond 12 nautical miles offshore, is a paid add-on. Standard plans have hard geographic limits. Coastal cruisers and ocean-crossing vessels need different plan configurations at substantially different price points. |
| Pausing Roam is simple and saves money. | Pauses operate in monthly increments only. Missing a billing date by a single day results in a full month's charge. Multiple user reports describe completing an incomplete pause step and being billed unexpectedly. |
| Starlink works wherever there is open sky. | The dish needs roughly 100 degrees of continuous unobstructed sky. A single branch or roof edge intersecting that arc causes repeated brief dropouts. The obstruction checker in the app shows the exact impact of local geometry before purchase. |
The fine print business buyers overlook
Consumer reviews cover speed and price myths. Far fewer cover what changes once Starlink sits behind a business operation, where procurement teams and insurance underwriters care about different questions entirely.
| What buyers assume | What the contract actually says |
|---|---|
| Starlink Business comes with a guaranteed uptime SLA. | No Starlink tier, including Priority, carries a contractual uptime guarantee comparable to enterprise fiber or VSAT contracts. Priority affects queue order during congestion, not a guaranteed percentage of uptime. |
| Owning the hardware means I can use it however I want. | The service agreement restricts resale, cross-border relocation, and certain commercial uses outside the matching plan tier, regardless of who physically owns the dish. |
| Marine insurance treats Starlink the same as certified VSAT. | Some underwriters still require certified satellite communications equipment for distress-communication compliance, separate from whatever internet service is on board. |
| The dedicated IP add-on makes my connection enterprise-grade. | It fixes the inbound CGNAT problem. It adds no redundancy, no SLA, and no guaranteed bandwidth on its own. |
| An outage gets treated like a utility outage. | Most jurisdictions have no regulatory credit framework for residential satellite downtime, unlike some traditional telecom services that carry statutory service credits. |
Running Starlink at scale: what network engineers actually deal with
This section is for network administrators, IT managers, and operators of remote sites, vessels, or multi-location deployments. The issues below are documented but largely absent from standard buying guides and vendor documentation.
The CGNAT problem that most IT documentation ignores
Starlink places customers behind carrier-grade NAT by default. This means your connection shares an external IP address with other Starlink users, and inbound connections are blocked. The practical consequences include inbound VPN configurations that fail, self-hosted servers that become unreachable, and certain remote desktop tools that stop functioning. The fix is SpaceX's dedicated IP add-on at $25 per month. Most enterprise IT teams discover the CGNAT issue the first time they configure a site-to-site VPN over a new Starlink connection, an avoidable delay with proper pre-deployment planning. Our deeper look at network device myths and the ISP traps most setups walk into covers the CGNAT pattern in more detail for readers managing any provider, not just satellite links.
Failover and dual-WAN configurations
The right architecture for any business using Starlink as primary connectivity is a dual-WAN router with automated failover. Peplink, Firewalla Gold, and Ubiquiti EdgeRouter all support this configuration. Most setups trigger failover after 10 to 30 seconds of detecting a dead connection. VoIP calls and video conferences running during that window will drop. For operations where call continuity matters, configure session-persistent routing for real-time traffic to prevent mid-call disruption during failover events.
Power draw at off-grid solar sites
The standard dish draws 50 to 100W continuously. The Mini draws 25 to 40W. At an off-grid site with a 200 Ah battery bank, roughly 2.4 kWh usable at 50 percent depth of discharge, a standard dish running overnight at 75W average consumes approximately 600 Wh, about 25 percent of overnight battery capacity before accounting for lighting, computing equipment, or refrigeration. The Mini's lower draw is a meaningful energy-budget decision for off-grid solar deployments, not just a portability consideration.
Multi-dish configurations and routing behavior
SpaceX permits multiple dishes on a single account, relevant for operations requiring redundancy or higher aggregate throughput. Each dish receives its own IP assignment and does not automatically load-balance with other dishes on the account. A router capable of policy-based routing is required to distribute traffic across two active Starlink connections. Session types such as authenticated web sessions and video calls will break if they switch between devices mid-session because the external IP changes.
Starlink for construction sites and temporary deployments
For construction sites or project deployments that relocate every few months, Starlink has a genuine advantage over cellular bonding solutions: it does not depend on nearby tower infrastructure. A cellular bonding setup in a remote area with one weak carrier performing at 5 to 15 Mbps will consistently underperform. Starlink's setup-to-connectivity time is under 30 minutes, hardware is transportable, and Roam plans allow pausing between active project periods. The practical tradeoff is dish placement, straightforward on a flat open site and genuinely difficult on a dense urban construction site surrounded by cranes and adjacent buildings.
Image Credit: Leonardo AI
The spectrum war that decided how much capacity Starlink gets to keep
For years, Starlink fought a quiet regulatory battle that mattered more to the network's future capacity than almost anything covered in consumer press. The dispute involved the 12.2 to 12.7 GHz band, spectrum Starlink uses to connect satellites to user terminals, and it just ended.
Dish Network and RS Access, a firm backed by Dell Technologies founder Michael Dell, had petitioned the FCC for years to open that band to two-way 5G mobile service, arguing their engineering studies showed the two uses could coexist. SpaceX countered with its own studies claiming Starlink customers would face harmful interference more than three-quarters of the time and total outages a majority of the time if Dish's plan proceeded, and it lobbied Congress directly, winning public support from at least one member of the House. In June 2026, the FCC adopted rules preserving the band for satellite services and rejected the two-way mobile proposal outright. Dish Chairman Charlie Ergen publicly conceded the point during the company's Q1 2026 earnings call, saying he respected the FCC's analysis even though the outcome disappointed him.
The practical effect for Starlink customers is straightforward: the spectrum the service depends on stays protected from a competing use that its own engineers argued would have made the network unreliable. For readers following how emerging technologies get shaped as much by regulatory fights as by engineering, our coverage of the AI Kill Switch Act and its loophole covers a similar dynamic, where the regulatory outcome ends up mattering as much as the underlying technology.
Orbit congestion is a real engineering problem, not a PR problem
The December 2025 near miss was covered primarily as a geopolitical story. It is more accurately described as an infrastructure coordination failure, and one that is structurally worsening as launch rates increase.
The mechanics matter here. Two objects at 560 km altitude traveling at approximately 27,000 km/h relative to Earth passed within 200 meters of each other with no shared traffic management protocol between operators. At those speeds, 200 meters represents roughly 26 milliseconds of separation. Ground-based space awareness systems observe object positions with some inherent time lag, and that lag is operationally meaningful at these closure rates.
The ITU filing strategy adds a regulatory dimension. SpaceX filed for orbital slots years before it could launch satellites to occupy them. Under ITU rules, this gives SpaceX a priority claim on specific frequencies and altitude bands, effectively shaping where competitors can operate. OneWeb and Amazon Kuiper are navigating the same system, but Starlink's early filing advantage creates coordination complexity as all three constellations grow simultaneously in overlapping orbital regions.
SpaceX's deorbit design is the mechanism that prevents long-term congestion from becoming catastrophic. Starlink satellites have a designed operational life of roughly five years, after which they deorbit and burn up in the atmosphere. Harvard astrophysicist Jonathan McDowell has noted that one or two Starlink satellites already re-enter the atmosphere every day. This approach works as debris management only if every commercial operator adopts comparable deorbit practices, and CAS Space has not made equivalent public commitments.
Amazon Kuiper received FCC authorization for thousands of additional satellites in early 2026. If it reaches full deployment alongside Starlink's constellation and OneWeb's, the total number of commercial low Earth orbit objects could approach 20,000 within a few years. No multilateral framework for physical collision avoidance between competing commercial operators currently exists at that scale. The ITU manages frequency coordination. It does not manage object proximity. That gap is the structural problem the December 2025 incident illustrated, and it is not new: China raised a nearly identical complaint with the United Nations back in 2021 over close approaches involving its own space station.
The astronomy cost nobody puts on the invoice
Consumer coverage of Starlink treats the sky as an unlimited resource. Astronomers have been documenting a real, measurable cost for years, and it is worth understanding on its own terms rather than as a footnote to the debris conversation above.
The Vera C. Rubin Observatory, which achieved first light in 2025 and began its ten-year Legacy Survey of Space and Time in 2026, is the most sensitive wide-field optical telescope ever built, and it is also the instrument most affected by satellite trails. Peer-reviewed simulations published by the observatory itself project that at a constellation size of around 40,000 low Earth orbit satellites, roughly 10 percent of all LSST images will contain at least one satellite trail, with the majority of images taken during twilight affected. A 30-second LSST exposure captures a satellite streaking about 15 degrees across the sky, leaving a trail across the telescope's entire 3.5-degree field of view. Separate modeling published in Nature in 2026 found that under more conservative earlier constellation estimates, 20 percent of midnight images and up to 80 percent of twilight exposures could show trails.
SpaceX has worked directly with Rubin astronomers on mitigation. The original Starlink satellites measured about magnitude 4.5 in brightness. The DarkSat experiment brought that down to roughly magnitude 6.1, and SpaceX's ongoing darkening program targets magnitude 7, a threshold at which image-correction algorithms can remove most artifacts from the data. Even at that brightness, the trail itself still exists at a high signal-to-noise ratio, meaning some systematic errors persist in the underlying data no matter how good the darkening gets. Researchers have separately found that lowering the entire V2 constellation from 550 km to 350 km altitude would cut the number of bright satellites entering Rubin's field of view by about 40 percent, since lower satellites move faster and spend less time in frame, though flying lower also increases atmospheric drag and shortens satellite lifespan.
Buying used Starlink hardware: what the resale market does not tell you
A large secondhand market for Starlink dishes has grown on eBay, Facebook Marketplace, and RV forums, since the hardware can save buyers 30 to 50 percent versus new. Almost no mainstream buying guide covers what can go wrong in that transaction, even though the risks are well documented by specialist resale guides.
Starlink hardware is tied to whichever account activated it, not to whoever physically holds the dish. A "locked" unit is one still registered to the previous owner. Entering its serial number on Starlink's site returns an already-registered error, and the unit cannot be activated until the original account holder logs in and initiates a transfer or deactivation, a process that takes roughly 24 to 48 hours to clear Starlink's backend once started. If the seller has already disappeared, or sold through a marketplace without completing that step, the buyer is left holding a dish that functions perfectly and connects to nothing.
A second constraint surprises even careful buyers: transfers are restricted to the same country the hardware was originally activated in, separate from the geofencing and licensing issue covered earlier in this piece. A dish bought secondhand from a seller in another country will not simply activate once it arrives, even in a fully licensed market. There is also a minimum holding period before a seller can even initiate a transfer, typically 90 days from activation or 120 days from purchase, whichever comes first, which matters if you are buying a barely used kit that turns out to still be inside that window.
| Buying scenario | What to check before paying |
|---|---|
| Seller has an active Starlink account | Ask for proof they initiated a transfer or deactivation through the app before you send payment, not after. |
| Seller claims the dish is "brand new, never activated" | Confirm the activation date on the serial number is genuinely absent, since a dish activated within the last 90 to 120 days may not be eligible for transfer yet. |
| Buying across a national border | Confirm the hardware was licensed in your country originally. Cross-border transfers are generally not supported. |
| Serial number returns an "already registered" error | Do not proceed until the seller completes deactivation. This is the single most common cause of a locked, unusable dish. |
On the safer side, Starlink's official transfer process is genuinely built for resale, unlike most consumer electronics: an owner can cancel service and select "transfer" directly inside the app, which generates instructions for the buyer to activate the hardware on a brand-new account. The catch is that only new accounts can accept transferred hardware, so a buyer who already has an active Starlink account elsewhere cannot simply add a second, secondhand dish to it.
Advanced: building redundancy instead of single-vendor dependency
This section assumes familiarity with the dual-WAN failover basics covered above. It is written for network architects, disaster-response planners, and maritime or offshore operators deciding whether to treat Starlink as a sole connectivity provider or as one layer inside something larger.
The calculus around single-constellation dependency has changed since 2023. Starlink is now the profitable core of a combined SpaceX-xAI entity whose overall capital allocation, public messaging, and engineering priorities are shaped by a much larger and riskier artificial intelligence bet, a governance shift worth weighing before committing critical infrastructure to one vendor for the next five to ten years. For a look at how a different major AI lab is handling capability and safety tradeoffs at a similarly large scale, our coverage of the water consumption problem behind AI data centers shows a related infrastructure tradeoff playing out in a different industry with the same underlying question: what happens when a core utility service sits inside a company making much larger, less proven bets?
A practical hybrid architecture pairs Starlink with a second constellation or a geostationary backup through a policy-based SD-WAN router rather than a simple two-device failover. Amazon's Kuiper network operates at roughly 630 km, closer to Starlink's 550 km, while OneWeb sits near 1,200 km, a difference that adds real, measurable latency and changes failover logic for real-time applications. Any hybrid setup needs its latency budget defined per application before deployment, not discovered during the first outage.
Route weighting by application type, rather than a blunt primary-and-backup model, is the mark of a mature deployment. Voice and financial transaction traffic can be assigned to the lowest-latency available path, while bulk data transfer and backups route over whichever link has spare capacity, regardless of which one is technically primary. This requires a router capable of policy-based routing rather than simple failover, the same requirement already noted for multi-dish Starlink fleets.
Contract and SLA stacking is the most direct way to address the uptime gap noted earlier in this piece: since no Starlink tier carries a guaranteed uptime SLA, pairing it with a geostationary backup that does carry one gives an operator a genuine compliance and guarantee layer underneath Starlink's low-latency primary path. This is standard practice in maritime and disaster-response deployments, where a single premium VSAT contract with a full SLA can cost several times what a Starlink-plus-backup combination costs, but the combination approach only closes the reliability gap if the backup contract's SLA terms are actually reviewed and not assumed.
What Starlink's subscriber count actually means
The headline is 12 million-plus subscribers. The number is verified, but it contains several distinct user categories that matter if you are trying to understand the business rather than only the reach.
Monthly active users and total subscribers diverge most clearly in the direct-to-cell figure. Starlink Mobile surpassed 10 million subscribers by March 2026, while more than 16 million unique users have connected across the broader constellation, including people who receive satellite backup automatically through an existing mobile plan without ever actively choosing Starlink. These users inflate the reach figure without generating Starlink subscription revenue in the traditional sense.
Revenue per subscriber varies significantly by plan tier. A Roam 100GB customer at $55 per month contributes a fraction of what a commercial Maritime customer at $500-plus per month generates. Starlink's own hardware revenue breakdown has shown Maritime and Aviation segments contributing small fractions of total subscribers but disproportionate shares of per-customer revenue.
Churn is meaningful in competitive markets. In urban and suburban areas where 5G home internet or fiber exists, Starlink loses subscribers due to price and performance. The service built its user base serving rural areas with no competing infrastructure. That remains its core stable market, which is also why the BEAD subsidy fight covered earlier in this piece matters so much to Starlink's long-term subscriber growth.
SpaceX is now an AI company, and that changes the Starlink story.
This is the part of the IPO story that received less attention than it deserved. In February 2026, SpaceX completed an all-stock merger with xAI, the artificial intelligence company that also owns the X social media platform and the Grok assistant. The combined entity reported 18.67 billion dollars in total 2025 revenue in its public S-1 filing, up 33 percent over the prior year, with Starlink contributing the large majority of that figure as the only consistently profitable segment. The AI unit, by contrast, posted roughly 3.2 billion dollars in 2025 revenue against operating losses near 6.4 billion dollars, contributing to a consolidated net loss of about 4.94 billion dollars for the year, with AI segment capital expenditure reaching 7.7 billion dollars in the first quarter of 2026 alone, driven largely by graphics processing unit depreciation and data center buildout.
The AI unit is not sitting idle, however. According to reporting on the company's compute deals, Alphabet agreed to pay SpaceX roughly 920 million dollars per month from October 2026 through June 2029 for access to around 110,000 Nvidia GPUs, and Anthropic separately signed a compute-access agreement using SpaceX's Colossus data center, with the two deals together estimated at about 26 billion dollars a year if fully realized. Whether that revenue arrives on schedule, and whether it closes the gap against AI segment spending, remains the open question hanging over the stock.
The most surprising disclosure in the original IPO prospectus was not about satellites at all. SpaceX said it is seeking FCC approval for up to 1 million satellites to function as an orbital data center network supporting AI computing, with deployment targeted as early as 2028. The company's stated reasoning is that terrestrial AI data centers face mounting constraints on power and water that orbital compute could avoid, a tension our earlier coverage of AI data centers and their water consumption problem examined in detail. Whether orbital AI compute is technically and economically viable at that scale has not been demonstrated, and SpaceX itself has described the technical challenge as one of the hardest unsolved problems in the industry.
For Starlink specifically, the merger means investors are no longer buying a pure connectivity business. They are buying a combined satellite, launch, and AI conglomerate where Starlink's stable cash flow now helps fund a much riskier AI bet, including a satellite-density ambition nearly 100 times larger than the entire current constellation.
The SpaceX IPO already happened, and the stock has already given most of it back.
Starlink still cannot be purchased as a standalone public stock. It remains a subsidiary of the combined SpaceX-xAI entity.
SpaceX confidentially filed a draft S-1 with the SEC on April 1, 2026, a process reportedly codenamed Project Apex internally. The public S-1 prospectus followed on May 20, 2026, disclosing the company's full financials for the first time in its 24-year history. The investor roadshow ran the week of June 8, and SpaceX priced its shares at $135 on June 11, 2026, offering $555.6 million shares and raising almost $75 billion, in what became the largest IPO in market history, more than double the previous record set by Saudi Aramco in 2019. Including the underwriters' overallotment option, the deal's potential size reached roughly $85.7 billion.
SPCX began trading on the Nasdaq on June 12, 2026. Shares opened at $150, an 11 percent gain over the IPO price, and closed the first session at $160.95, a 19 percent gain that valued the company at roughly $2 trillion. Over $500 million in shares traded on day one, the second-largest IPO-day volume in Nasdaq history. SpaceX reserved roughly 30 percent of the offering for retail investors through Robinhood, Fidelity, Charles Schwab, SoFi, and E*Trade, an unusually large retail allocation for an offering this size. Demand still vastly outstripped supply: the order book ran more than two times oversubscribed, with roughly $150 billion in orders chasing the $75 billion raise, and many retail applicants received only a partial fill or nothing at all.
The stock then moved with the kind of volatility that defines a brand-new, heavily retail-traded listing, and that volatility has not settled down. SPCX touched an all-time high near 225.64 dollars on June 16, 2026, joined the Nasdaq-100 index on July 7, and was still trading close to 149 dollars as of July 10. From there, it kept sliding. By July 27, CNBC reported the company had lost the equivalent of a full Tesla in market capitalization. On July 28, 2026, the stock set a new 52-week low of $107.01, briefly dipping below its own $135 IPO price for the first time. As of July 29, 2026, shares traded around $112.55, down roughly 50 percent from the June peak, with the company's market capitalization sitting near $1.5 trillion, down close to 11 percent in a single week. Analyst sentiment remains split rather than settled: the average 12-month price target sits around $236.71, with individual estimates ranging from as low as $62 to as high as $800, a spread that itself signals how unresolved the valuation debate still is.
The underlying Starlink business metrics are genuine. The segment generated the large majority of SpaceX's $18.67 billion in 2025 revenue, and Starlink alone is reported to carry EBITDA margins above 60 percent, a rare figure for any infrastructure business. Most insiders remain under a standard 180-day lockup, though SpaceX structured an early exception allowing employees and early backers to sell up to 20 percent of their holdings once the first quarterly results land, a detail worth knowing for anyone tracking future selling pressure on the stock as that window approaches.
This article does not constitute financial or investment advice. Consult a licensed financial adviser before making any investment decisions involving SPCX or any other security.
How to set up, log in, and track your Starlink
Existing customers manage their accounts at account.starlink.com. The app handles network configuration, dish positioning, outage alerts, and data usage monitoring.
Satellite pass tracking uses tools like heavens-above.com or the SkySafari app. SpaceX maintains launch manifests on its own website, and third-party sites such as Spaceflight Now track individual mission schedules, including Starlink batch launches.
A standard residential kit installation takes under 30 minutes: download the app, place the dish in an open-sky position, and run the cable to the router. The single most important step most buyers skip is running the app's obstruction checker before committing to a mounting position. The checker uses augmented reality to show exactly which trees or roof edges will interrupt the satellite arc at your specific location.
Starlink's growth past 12 million subscribers is real, and the technology works. Rural users who had no viable broadband option before 2021 now have one. That is a meaningful infrastructure change, particularly in markets where the alternative was a cellular hotspot with a 20 GB cap or a geostationary satellite plan that made video calls impossible.
The China coverage got the original story backwards. The laser demonstration from geostationary orbit was an engineering achievement worth reporting on its own terms, not as a competitive threat to a consumer low Earth orbit network. The incident that deserved more coverage was the 200-meter near miss in December 2025. That is not primarily a geopolitical story. It is an infrastructure coordination story, and it is not even a new one, given that China raised the same complaint about Starlink in 2021. Two commercial operators have repeatedly launched into the same orbital corridor with no shared protocol for proximity management.
The federal funding story is the clearest example of coverage that treats a settled headline as the whole picture. Starlink lost its Rural Digital Opportunity Fund bid twice, on the government's own finding that the technology could not meet the promised speeds at the time. It then won a much larger BEAD award under a rule change that made the entire program technology neutral, a genuine reversal of fortune rather than a straight line of success. Both facts are true, and coverage that only reports one of them misleads readers about how contested this money still is in Congress and in individual states.
The IPO is no longer a hypothetical, and neither is the pullback that followed it. The listing broke records in June. By late July, SPCX had fallen below its own IPO price and lost roughly half its value from the post-listing peak, a swing that erased and then partially restored hundreds of billions of dollars in paper value inside seven weeks. Starlink's margins are genuinely strong for an infrastructure business, which is exactly why SpaceX wanted to access public capital while growth was still fast enough to support an aggressive valuation. Whether the stock stabilizes depends on questions that remain unresolved as of this update: whether Starship can land its booster reliably now that the V3 satellites themselves have proven they work, and whether the AI segment's new compute deals with partners like Alphabet and Anthropic can close the gap against its own heavy losses. Investors buying SPCX are underwriting all of those bets at once, not just a satellite internet company.
On the coverage honesty gap: 160-plus countries in availability maps do not mean 160 countries of comparable service quality. The network is strongest where subscriber density is highest, primarily North America, northern Europe, and Australia, and weakest in the markets where the unconnected population is largest. At $349 for hardware and $55 to $130 per month for service, Starlink remains unaffordable for the majority of the roughly 2 billion people globally who still lack reliable internet access. The company is solving a real connectivity gap. It has not yet solved the affordability gap that accounts for most of the world's offline population, and its own astronomy mitigation efforts show that even its technical wins come with tradeoffs that rarely make the subscriber-count headline.
- Over 12 million subscribers across 160-plus countries, still growing fast
- Starlink alone reportedly carries EBITDA margins above 60 percent
- Starlink Mobile has surpassed 10 million direct-to-cell subscribers
- V3 satellites have now proven they work in orbit and can communicate
- Won a multi-year spectrum fight with Dish, protecting its core frequencies
- Genuine first-broadband access in markets with no competing infrastructure
- SPCX fell below its own IPO price on July 28, 2026, roughly seven weeks after listing
- The AI segment lost nearly 5 billion dollars in 2025 against Starlink's profits.
- Starship's booster has now failed landing on both V3 flights
- No multilateral orbital coordination framework exists at the current commercial LEO density
- Only about 45 percent of US subscribers meet BEAD's own minimum speed threshold
- Hardware and monthly costs keep the service out of reach for most of the world's offline population
Recent Articles from USABeam
- Why South Asia's Governments Keep Losing to Gen Z
- Inside the 8 countries where governments keep falling in 2026
- Political ideology explained: what 10,357 Americans reveal about left and right in 2026
Editor's note: All images accompanying this article were created using AI image generation. All data, figures, and case studies in the article itself are drawn from cited public sources.
