SpaceX Asks the FCC to Clear a 100,000-Satellite Starlink Network Built on Starship

The third-generation, non-geostationary system would spread up to 100,000 satellites across two very-low-orbit shells packed with beamforming and optical inter-satellite links to multiply capacity, but the entire plan hinges on regulators clearing spectrum and Starship becoming routine enough to fly the manifest — even as Tesla stacks parallel infrastructure bets in grid storage, CyberCab, and domestic lithium.

The number that will grab headlines — 100,000 satellites — is the least interesting part of what SpaceX just asked Washington to approve. The real story is a loop: Starship makes the network physically possible, and the network becomes a distribution and data advantage that compounds across every company in Musk's orbit. Read that way, the filing is less a satellite-scale story and more a bet on AI-infrastructure optionality — and it rhymes with a run of Tesla disclosures this week that all point at the same thesis, that the companies building their own physical layer will own the next decade.

Key Takeaways

• SpaceX asked the FCC to approve a third-generation, non-geostationary Starlink system of up to 100,000 satellites across two very-low-Earth-orbit shells.
• Elon Musk said the deployment would require Starship, making launch cadence the central constraint on the entire architecture.
• Separate from this filing, SpaceX already has a prior application for up to 1 million orbital AI and data-center satellites.
• Tesla Energy landed a Megapack order worth up to $3 billion from Eshesoft for more than 15 gigawatt-hours across the UK, Western Europe, the GCC, and India.
• Over a six-week span, the company announced more than $9 billion and 43 gigawatt-hours of new storage orders.
• CyberCab will be Tesla's primary robo-taxi vehicle, using reaction injection molding to cut paint-related cycle times from hours to minutes and trim those parts' emissions by 35%.
• In Robstown, Texas, Tesla's lithium refinery aims to be North America's largest, producing lithium with more than 30% lower emissions via an acid-free alkaline leach process.
• Powerwall virtual power plants prevented 6.5 million outages globally in 2025, delivered more than 20 gigawatt-hours, and saved owners more than $1 billion.
• Giga Berlin used about 0.3 million cubic meters of water in 2025, down from 0.45 million and well under its 1.4 million permit.

---

The Satellite Count Is a Distraction

The instinct with a filing like this is to fixate on the headline number and argue about whether 100,000 satellites is realistic or reckless. That misses the mechanism. What SpaceX is describing is a third-generation, non-geostationary system spread across two very-low-Earth-orbit shells, stacked with phased-array beamforming, electronic steering, optical inter-satellite links, and higher-frequency spectrum. Each of those is a capacity multiplier, not a vanity spec.

The point isn't how many objects orbit; it's how much throughput the constellation can push and how far beyond today's consumer, broadband, mobility, and government markets Starlink can reach. Scale here is a means to capacity, and capacity is the actual product.

Starship Is the Real Gating Factor

Musk's own framing tells you where the bottleneck sits: this deployment requires Starship. That single dependency reorders the whole risk profile. Falcon 9 cadence built the first Starlink; a constellation of this scale can't be lofted the same way.

So the milestone that actually matters isn't FCC ink — it's whether Starship becomes routine enough to carry a manifest of this size on a predictable schedule. Until launch cadence is boring, the 100,000-satellite number is an aspiration with a rocket-shaped asterisk.

The AI-Infrastructure Read Hiding in the Filing

Here's the second-order angle most coverage skips. A denser, higher-capacity, optically linked constellation isn't just faster internet — it's a distribution layer and a data advantage that can compound across Musk's companies. And this filing sits next to a separate SpaceX application for up to 1 million orbital AI and data-center satellites.

Stack those two ideas and the strategy stops looking like a broadband upgrade and starts looking like optionality on putting compute and connectivity in orbit. That's the bet worth watching, and it's why the regulatory questions — orbital debris, collision risk, spectrum interference, and astronomy impact — are the real near-term friction rather than the physics.

Tesla Energy Is Still Priced Like a Side Business

Storage keeps getting treated as a footnote to the car company. The numbers argue otherwise. Tesla just booked a Megapack order worth up to $3 billion from Eshesoft — more than 15 gigawatt-hours across the UK, Western Europe, the GCC, and India, sold under an "Eshesoft Energy Storage powered by Tesla" brand. That lands inside a six-week stretch that produced more than $9 billion and 43 gigawatt-hours of new orders.

The demand signal is unambiguous: renewable penetration, grid congestion, and AI data-center load are turning grid buildout into a battery-deployment problem, and fast capacity beats a decade-long thermal plant. The open question is execution, not demand. Backlog is a promise; margins, delivery dates, named projects, and factory output are the proof. Convert the backlog into repeatable, high-trust deployment and the market's "side business" framing breaks.

CyberCab Is a Cost-Per-Mile Machine

CyberCab is best understood not as a car but as a cost-per-mile engineering exercise designed from the vehicle up. Tesla has named it the primary vehicle for its robo-taxi fleet and stripped out the steering wheel, pedals, and traditional controls to cut weight and operating cost in a compact two-seat layout aimed at the one-or-two-passenger trips that dominate ride-hailing. The manufacturing story is where the leverage lives: 4680 cells, steer-by-wire, a 48-volt architecture, an unboxed assembly process, and body panels that skip the traditional paint shop. The detail I'd flag is reaction injection molding, which collapses paint-related cycle times from hours to minutes and trims emissions on those parts by 35%.

None of that is a rollout date. There's still no production start, volume plan, or regulatory clearance attached. But cycle-time and capex reductions are exactly the ingredients that make the next phase of robo-taxi economics viable once regulators sign off — the vehicle, the line, and the fleet model engineered as one system.

The Quiet Vertical Integration in Lithium

The least glamorous move this week may end up mattering most. Tesla's refinery in Robstown, Texas — outside Corpus Christi — is positioned to be the largest in North America, producing lithium with more than 30% lower emissions than traditional hard-rock methods through an acid-free alkaline leach process, and it topped the Lead the Charge ranking for a second straight year. Strategically, this is Tesla pulling battery-materials processing in-house so its EVs and Megapacks lean less on third-party processors and foreign supply chains.

The size claim is the easy part. The real test is yield and commissioning, plus the friction that comes with siting a chemical process in drought-prone South Texas: water use and wastewater discharge are live community issues. Nameplate capacity, actual output, battery-grade qualification, and third-party verification of the emissions benchmark are what turn this from a press line into a genuine cost and supply advantage.

Powerwall Turns Homes Into a Power Plant

Powerwall is quietly demonstrating why distributed batteries can add grid capacity without waiting a decade for new generation. Aggregated as virtual power plants, Tesla's units prevented 6.5 million outages globally in 2025, delivered more than 20 gigawatt-hours, and helped owners save more than $1 billion through solar, storage, and grid exports — with individual accounts seeing grid-participation credits approaching $300. That reframes residential storage as capacity, resilience, and software-enabled grid services rather than hardware sold once.

The valuation gap is the model. The market still prices Tesla Energy as boxes shipped; the more accurate frame is hardware plus grid services plus software economics. The methodology deserves scrutiny — how an "outage prevented" is defined, how regional events are counted, how much margin Tesla actually captures — but the strategic direction is clear. The unlock is whether utilities graduate from scattered pilots to large, recurring contracted capacity.

The Fine Print: Water, Price, and Proof

Two smaller disclosures round out the picture, and both feed durability arguments. Giga Berlin used about 0.3 million cubic meters of water in 2025, down from 0.45 million the year before and well under its 1.4 million permit, backed by a recovery plant that can recycle up to 100% of processed wastewater — figures that will resurface in every future fight over factory expansion. On affordability, Tesla pegs Model 3 pricing below the average US new-car price and puts Model Y rear-wheel drive at 77 cents per mile against 81 cents for a Toyota RAV4, 85 cents for a Hyundai Ioniq 5, and $1.15 for a BMW X3.

The through-line across all of it — satellites, storage, robo-taxi hardware, lithium, and cost-per-mile math — is the same first-principles move: own the physical layer, drive the unit cost down, and let the infrastructure compound. The next set of proof points is unglamorous and decisive: regulatory clearance, production ramp, and whether reported demand actually converts into deployed capacity.