Musk: Each Starlink V3 Satellite Could Hit 10 Tbps

Elon Musk said each next-generation Starlink satellite should deliver around 10 terabits per second in both directions, with a path to more than 100 Tbps, in a post on X reported by Yahoo Finance on 20 September 2026. The post expanded on comments the SpaceX chief executive made on the company’s 4 August 2026 earnings call, where he predicted Starlink could carry the bulk of the world’s internet traffic within 10 years.

The figure has not been demonstrated in orbit, and the constellation it assumes has not been licensed. Both points are made in the report itself, and both are worth holding onto while reading the numbers below.

Key facts

• Musk’s X post says connectivity per satellite will be closer to 10 Tb in both directions, each satellite drawing 250kW and carrying a SpaceX-designed computer based on hardware from Nvidia, the US chip designer, described as a Vera Rubin NVL72, Yahoo Finance reports.
• On the earnings call, Musk said a Starlink V3 satellite is about an order of magnitude more capable than a V2, and that SpaceX expects to launch about an order of magnitude more of them. Musk did not combine the two figures; by implication, multiplying them would give roughly a hundredfold increase in delivered bandwidth, but that arithmetic is ours, not his and not the outlet’s.
• An FCC filing covering up to 100,000 Gen3 satellites was accepted for filing in mid-September 2026 but has not been approved, according to the outlet. Acceptance for filing is a procedural step at the commission — it means an application is complete enough to be put out for public comment, not that regulators have agreed to anything.
• SpaceX reported second-quarter 2026 revenue of $7.81 billion, up 92% year over year, per the same report.
• Starlink added more than 1.7 million net consumer subscribers in the quarter, its best ever, with average revenue per user holding at $66 a month, the outlet says.

Why it matters

What would change if the number held

Yahoo Finance frames the target as a change in kind, not degree: if achieved, 10 Tbps per satellite — and especially a path to 100+ Tbps — would turn Starlink from a high-capacity broadband network into one capable of carrying a meaningful fraction of global internet traffic and supporting AI and industrial applications from orbit. The outlet calls it one of the most aggressive capacity targets publicly stated for any satellite system.

Our own reading is narrower, and we want to label it as a reading rather than a finding: a per-satellite jump of that size, if it held up, would alter the arithmetic of selling capacity in places where terrestrial fibre is absent, because the cost of delivering each additional gigabit falls as more of it rides on the same spacecraft. That is an inference drawn from a single-source claim about hardware nobody outside SpaceX has measured. It is a premise, not a result, and no independent market analysis we could find has tested it.

The outlet is equally clear that this is a forward-looking claim. It says the FCC application for the large Gen3 constellation is still under review, and that actual on-orbit performance, spectrum access, ground infrastructure and Starship launch cadence will decide how much of the vision is delivered.

What it does not tell us about price or availability

Nothing in the report connects the capacity target to what a Starlink subscription costs or to where it can be bought. The only price-adjacent number in the quarter’s disclosure is a blended average revenue per user of $66 a month, which the outlet reports held flat while subscriber numbers grew — on those figures, satellite internet pricing is not where the growth came from. Residential plan prices vary by country and by whether a given service cell is congested, the outlet publishes no per-country breakdown, and we have not independently verified current advertised prices in any market for this report.

Availability, meanwhile, is a licensing question rather than a bandwidth one. A satellite that can move 10 Tbps is worth nothing over a country that has not granted a landing right, which is why our running tally of which countries have licensed Starlink is the more useful reference for readers asking where and at what price they can actually sign up.

What Musk told analysts

Asked on the 4 August call how big Starlink could get, Musk went further than on previous calls, Yahoo Finance reports. He said it is not out of the question that Starlink will at some point deliver a majority of the world’s internet — at least in countries where SpaceX is allowed to operate, which he put at the vast majority of countries — and that this is less than 10 years away rather than in the distant future.

That conditional — “where SpaceX is allowed to operate” — is Musk’s own, and it is the part of the forecast that regulators rather than engineers control.

On demand, Musk tied the forecast to machine traffic rather than people, saying that with the advent of AI, humanoid robotics and vehicle robotics, bandwidth requirements will be far greater than in the past, per Yahoo Finance.

On the money, he said that even if monetisation per bit fell by a factor of 10, Starlink would still see a 10x increase in revenue. Read plainly, that is a statement that the company expects volume to outrun any price erosion — the same pattern the flat $66 ARPU already shows in miniature.

Who is making the claim, and why that matters

The incentive structure behind the number

The capacity numbers were delivered in two settings that reward optimism: a post on a platform Musk owns, and an earnings call held for investors in a private company that raises money at negotiated valuations rather than on a public exchange. SpaceX is not listed, so there is no quarterly filing with the US Securities and Exchange Commission against which an outside reader can check the revenue or subscriber figures, and no regulator reviewing the forward-looking statements for the disclaimers a listed company would be obliged to attach.

SpaceX has also not run a conventional media-relations desk for several years. We have no on-record company response beyond the post and the call remarks quoted by the outlet, and nothing in this article should be read as SpaceX answering questions about the figure.

A revision upward without a test result

None of that makes the claim false. It does mean the incentive runs one way. Musk has a documented history of announcing capability targets ahead of schedule, and the report itself notes that the previously stated V3 figure was roughly a tenth of the one he gave last week — a tenfold revision upward that arrived without an accompanying test, flight demonstration or published measurement. Engineering numbers usually move in the other direction as hardware meets reality.

We flag the conflict because the rest of this story is, unavoidably, a relay of what one man said about his own company, filtered through one outlet.

The quarter behind the forecast

Where the growth came from

The Q2 2026 numbers give the projection some grounding, Yahoo Finance writes. The connectivity segment contributed $4.3 billion, up 66% year over year, while enterprise and government revenue grew 108%.

That enterprise growth came from new airline deals with American Airlines, Southwest, Virgin Atlantic, Iberia and Aer Lingus, plus more than $6 billion in multi-year Space Force Starshield contracts, according to the outlet. Starshield is SpaceX’s government-facing satellite line; the Space Force is the US military branch responsible for space operations. The run of carrier signings is the clearest published evidence of how central Starlink’s aviation business has become to its revenue mix, though the airline agreements are commercial contracts whose terms the carriers have not published.

What $66 a month does and does not tell you

The $66 monthly average revenue per user is a blended figure across consumer plans and markets, not a price list, and the report does not break it down by country or tier. It is the closest thing to a pricing signal in the quarter’s disclosure, and it is flat rather than rising — growth in the consumer segment is coming from subscriber count, not from charging existing customers more. A blended ARPU can stay flat while individual market prices move in opposite directions, so it should not be read as evidence that prices held steady anywhere in particular.

Musk credited execution for the enterprise wins, saying that securing large enterprise revenue requires demonstrating very high reliability and uptime, and acknowledging that Starlink did not have high uptime in its early days. SpaceX president Gwynne Shotwell said Starlink “has never lost an enterprise customer,” Yahoo Finance reports — a claim about a private customer list that cannot be checked from outside the company.

The one claim that sits in a public record

Of everything in this story, the FCC filing is the only element that exists as a public document rather than a company statement, so it is the one item we can put independent procedural context around.

Acceptance for filing is a defined stage in the commission’s satellite licensing process, not a decision. Under the FCC’s rules for space station applications, an application accepted for filing has been found complete enough to be placed on public notice, which opens a window for comments and petitions to deny from rival operators, spectrum incumbents and other interested parties. Approval, if it comes, is a separate act by the commission or its Space Bureau — the successor to the International Bureau on satellite matters — and it can attach conditions, cut the authorised number of satellites, or defer part of the request for later consideration.

There is precedent for that last outcome in Starlink’s own regulatory history. When SpaceX asked to fly roughly 30,000 second-generation satellites, the commission’s December 2022 order granted a partial authorisation for 7,500 and deferred the remainder — a quarter of what was requested, with conditions attached. That background comes from the FCC’s published record rather than from the outlet, and it is the reason “accepted for filing” and “100,000 satellites” should not be collapsed into “licensed for 100,000 satellites” in a headline or a summary.

We did not pull the Gen3 docket ourselves before publication, and we are not going to imply otherwise. What we can tell readers is where it lives: applications of this class are filed in the commission’s International Bureau Filing System and tracked by docket number in its Electronic Comment Filing System, both searchable by the public at no cost. That record, not a press summary, is the authority on the application’s current stage.

What we could not independently confirm

This report is built on a single published account. We were not on the 4 August call, and we have not matched the quoted figures against a second outlet, an analyst note, or a document filed by SpaceX. We searched for a second independent account of the 10 Tbps figure before publishing and did not find one we were prepared to cite. Specifically, the following remain uncorroborated at the time of writing:

• The 10 Tbps and 250kW per-satellite figures. Sourced to Musk’s own post on X as summarised by Yahoo Finance. No third-party measurement exists because no V3 satellite is yet operating at scale.
• The revenue, ARPU and subscriber numbers. SpaceX is private and does not publish audited quarterly accounts, so these are company-supplied figures relayed through one outlet.
• The airline and Starshield contract values. Neither the carriers nor the Space Force are cited directly in the report, and the contract terms have not been published by the counterparties.
• The FCC filing’s current stage. The procedural meaning of acceptance for filing is fixed by the commission’s own rules, as set out above, but we have not verified the docket’s present status in IBFS ourselves. That check — not a press summary — settles the question.

The Yahoo Finance piece lists its corrections contact as editorial@247wallst.com, which indicates the article is syndicated from 24/7 Wall St. rather than reported in-house — relevant context when weighing how many independent sets of eyes the numbers have passed through. We will update this story if a second account, a regulatory document or a SpaceX statement confirms or contradicts any of the figures above.

What has to happen next

The capacity story depends on getting V3 hardware into orbit in volume, which is why when the first V3 satellites actually fly is being watched closely. Musk described V3 as a step change over V2, which he called an already very capable satellite, according to the outlet.

Three checkpoints will tell readers whether the target is converging on reality: an FCC decision on the Gen3 application, and whether any grant covers the full 100,000 or a fraction of it as in 2022; a sustained Starship launch cadence anywhere near the one-flight-a-day figure Musk gave analysts; and the first independent throughput measurements from V3 satellites once a meaningful number are operating. None of the three has happened.

Until they do, 10 Tbps is a stated target rather than a measured result — and the difference between those two things is the whole story.

Frequently Asked Questions

When will Starlink customers actually notice the V3 upgrade?

SpaceX management told analysts that a critical mass of roughly 1,000 V3 satellites is needed before service improvements become significant, according to Yahoo Finance. The company expects to reach that point by approximately the second quarter of next year. That timeline is the company’s own forecast, not an independently verified schedule, and it assumes a launch rate SpaceX has not yet demonstrated.

How often does SpaceX plan to launch once V3 deployment ramps up?

Musk told analysts on the earnings call that a year from now SpaceX will be doing at least one flight a day, and possibly more, Yahoo Finance reported. The same report lists Starship launch cadence as one of the factors that will determine how much of the capacity target is realised. Readers can check that claim against SpaceX’s own launch manifest and public FAA licensing records as the year progresses.

How does the 10 Tbps figure compare with earlier V3 numbers?

Yahoo Finance reports that earlier V3 figures were closer to 1 Tbps downlink, and describes the jump to 10 Tbps bidirectional as ambitious. On those numbers the new claim is roughly an order of magnitude higher than what SpaceX had previously put in public — that comparison is our arithmetic on the two figures the outlet cites, not a statement by Musk. The outlet lists on-orbit performance, spectrum access and ground infrastructure among the unknowns.

Will more capacity make Starlink cheaper, and where can you buy it?

Nothing in the reporting links the capacity target to consumer pricing. The only price-adjacent figure disclosed for the quarter is a blended average revenue per user of $66 a month, which the outlet says held flat while subscriber numbers grew — on those figures, growth came from volume rather than from raising prices. The outlet publishes no per-country price list, and we have not independently verified advertised residential prices in any market for this report. Availability is decided by national licensing rather than by satellite bandwidth, so the list of countries that have granted Starlink a licence is the more useful guide to where service can actually be bought.

Is SpaceX publicly traded, and does it have a share price?

No. SpaceX is a privately held company. It has no listed common stock and therefore no public share price or official ticker symbol. Any quoted price attached to a SpaceX-sounding ticker refers to a different, unrelated entity or to a private-market or derivative product, not to shares in SpaceX itself. An earlier version of this article cited a closing share price for SpaceX; that figure was wrong and has been removed.

How well corroborated are the capacity and subscriber figures in this story?

Thinly. Every number in this report traces back to a single published account — a Yahoo Finance piece whose corrections address points to 24/7 Wall St. — summarising Musk’s post on X and SpaceX’s 4 August 2026 earnings call. We looked for a second independent account of the 10 Tbps figure before publishing and did not find one we could stand behind. SpaceX is private and does not file quarterly reports with the US Securities and Exchange Commission, so the revenue and subscriber numbers cannot be checked against an audited filing. The FCC application is the one element that sits in a public record, and the meaning of its current status — accepted for filing — is fixed by the commission’s own procedural rules rather than by the outlet’s characterisation. Until a decision lands, treat 10 Tbps as a stated target.