There is an obvious question to ask about the satellite spectrum rate India's telecom department has reportedly settled on, and it is the wrong one. The obvious question is whether 5% is a lot. It leads nowhere, because a percentage prices nothing until you know what it is a percentage of, and nobody outside SpaceX — quite possibly nobody inside it — knows what Starlink's Indian revenue will be. There is no published subscriber forecast, no announced price, and no launch date.
The question worth asking is narrower and answerable: what is this rate a rate of, what sits alongside it, and which of the two competing proposals actually costs an operator like Starlink more? That last part matters because the framing everywhere is that the department overrode its regulator upward. On the regulator's own published arithmetic, it did the opposite for most of the customers Starlink would sign — and went up only for the ones Starlink says it is coming to India to serve.
So: the rate, the stack it sits in, and then the honest statement of what nobody can size yet.
What has been reported, and what is a document
Five Indian business outlets reported on 8 and 9 September 2026 that the Digital Communications Commission — the Department of Telecommunications' top decision-making body — has approved a 5% of adjusted gross revenue spectrum charge for satellite operators, with a one-point concession, a five-year assignment extendable by two, and no auction. Every one of those reports cites unnamed officials. The meeting is reported as 2 September by one outlet and 3 September by another.
There is no gazette notification and no departmental press note. A Cabinet note is being drafted. Treat the 5% as a proposed rate throughout; nothing below assumes it is in force.
Two things underneath it are documents, and they change the story:
- The regulator's recommendation, 9 May 2025. TRAI recommended 4% of AGR for satellite fixed services, plus Rs 500 per urban subscriber per year, subject to a Rs 3,500 per MHz annual minimum. The formula it published is
Max{(4% of AGR + 500 × Nu), annual minimum}, where Nu is urban subscribers. The extra charge is additive to the 4%, not an alternative to it. - TRAI's response to a departmental back-reference, 8 December 2025. The department wrote to TRAI on 12 November 2025 proposing 5%, and TRAI refused. Its words: "the Authority does not concur with the DoT's proposal to charge the spectrum @ 5% of AGR with conditional discounts", and it "reiterates its Recommendations no. 4.12".
The 5% is not a September development. It is a ten-month-old departmental position that the regulator declined in writing, and which the department has now reportedly adopted anyway.
The 5% is not the whole rate
Spectrum charges in India sit on top of the licence fee. Indian telecom authorisations carry a licence fee of 8% of AGR, inclusive of the 5% universal service levy, paid quarterly — and satellite operators are not exempt from it. SpaceX ($SPCX) is already licensed in India: Starlink holds the unified licence with GMPCS, VSAT and ISP Category-A authorisations, reported granted in June 2025, and its space-regulator authorisation reported in July 2025. What it does not have is a spectrum assignment. That is what this rate governs.
So the all-in take on Indian adjusted gross revenue, at the reported rate:
8% licence fee + 5% spectrum charge = 13% of AGR.
SpaceX and Amazon's Kuiper asked TRAI for a spectrum charge below 1% of AGR with no other charges. On the spectrum line alone, the reported outcome is at least five times the ask. But the licence fee is not something an operator can ask its way out of, and it is the larger of the two charges. The whole argument everyone is having — 4% against 5% — is over the smaller half of a 13% total, and the one point in dispute is under 8% of what India would collect.
That is the number this piece exists to supply, and it is arithmetic on two disclosed rates rather than anyone's estimate.
The concession is probably not what it is being reported as
Every report describes the one-point discount as a rural or remote rate — 4% outside the cities. The department's own written proposal describes something different, and the difference is large.
From the back-reference, quoted in full in TRAI's December response: "An annual spectrum charge of 5% (instead of 4%) with a discount of 1% if a certain percentage (say 5%) of overall customers enrolled in the year are from the above areas as in (1.) above and notified by Government." The areas named in (1.) are border, hill and island regions — what the department calls "hard-to-connect".
That is a firm-level, all-or-nothing test, not a per-subscriber geographic rate. Under it an operator pays 5% on every rupee, or 4% on every rupee, depending on whether one enrolment threshold is cleared for the year. And the qualifying areas are narrower than "rural": TRAI's December response puts border, hill and island populations at about 5% of India, against about 65% for rural and remote combined.
Whether the adopted version is the department's conditional test or the geographic rate the press describes cannot be established without the notification. Nobody should quote a blended rate until it exists.
Counted the regulator's way, the department went down for urban and up for rural
TRAI did its own arithmetic on the Rs 500 urban charge it recommended, and it is the most useful number in either document. In its December response: if satellite tariffs land near present Indian fixed-wireless base tariffs, "this Rs. 500 per subscriber charge will amount to 4-6% of AGR of urban connections."
Put the two packages on the same subscriber:
- An urban subscriber. TRAI's package: 4% plus another 4–6%, so 8–10% of AGR. The department's: 5%. The department's rate is three to five points cheaper.
- A rural subscriber. TRAI's package: 4%, because rural users are exempt from the Rs 500. The department's: 5%, or 4% only if the enrolment test is met. The department's rate is one point dearer, or level at best.
The reported story is a department overriding its regulator upward. The regulator's own published numbers say it overrode downward for city customers and upward for country ones. Starlink's entire stated case in India is rural connectivity, which puts it on the only side of that swap that got worse.
Add the licence fee back — the bars above stack all four positions the same way — and the comparison holds in the same direction: 12% all-in on TRAI's rural treatment against 13% on the department's, while TRAI's urban treatment is the most expensive package anyone has proposed at 16% on the conservative end of its own estimate.
Nobody can size the revenue this applies to, including us
A rate needs a base. Here is every price anchor that exists for Starlink in India, and they point in opposite directions.
TRAI's implied benchmark: roughly $7 to $11 a month. This is derivable from the sentence quoted above. If Rs 500 a year is 4% of a subscriber's AGR, that AGR is Rs 12,500 a year; if it is 6%, Rs 8,333. At 94.80 rupees to the dollar on 8 September 2026, that is $7.33 to $10.99 a month. It is TRAI's working assumption benchmarked to Indian fixed-wireless tariffs, not a Starlink price — but it is the only per-subscriber revenue figure with any official provenance attached, and the regulator built a levy on it.
Starlink's own placeholder: about $91 a month. On 8 December 2025 an India residential plan appeared on Starlink's website at Rs 8,600 a month with a Rs 34,000 hardware kit — $90.72 and $358.65. Hours later Starlink vice president Lauren Dreyer said the figures were dummy test data exposed by a configuration glitch, that the India site was not live, that pricing had not been finalised and that no orders were being taken. It is not a price. It is worth recording only because it is eight to twelve times the regulator's working benchmark, and 1.37 times the $66 blended global ARPU our own model carries.
The government's own anchor: about $31.65 a month, and a figure of two million. In late July 2025 Pemmasani Chandra Sekhar, Union Minister of State for Communications, said on the sidelines of a state-telco review meeting that "Starlink can have only 20 lakh customers in India and offer up to 200 Mbps speed" — twenty lakh being two million connections — and, in the same remarks, that "the monthly cost may be around ₹3,000." Neither figure is a SpaceX disclosure, the price is explicitly the minister's estimate, and he did not say what the two million rests on: outlets reporting the remarks split between reading it as the constellation's capacity over India and as his judgement of what the market will bear. Treat it as the government's own working figure rather than a binding cap. Rs 3,000 is $31.65 at the rate above, which lands between the regulator's benchmark and the disavowed placeholder.
It is still the only quantity of its kind on the record, and it is worth running. Two million connections at Rs 3,000 a month is Rs 72 billion, about $759M a year; 5% of that is $38.0M a year, or 0.22% of Connectivity's June-quarter run rate annualised at $17,164M. That is the shape of this charge at the largest India anyone in government has described out loud.
This is where most analysis quietly picks a number. We are not going to. The three anchors that exist are a regulator's benchmark for a different technology, a minister's off-the-cuff estimate and a company's disavowed test data, and the outer two differ by an order of magnitude. SpaceX has published no India price and no India subscriber forecast; the only India subscriber figure on the public record is the government's own ceiling.
What can be stated is the rate applied to a range, with the range labelled as ours:
| India ARPU, monthly | AGR per 1M subscribers | Charge at 5% | The extra point |
|---|---|---|---|
| $7.33 — TRAI's low benchmark | $88.0M | $4.40M | $0.88M |
| $10.99 — TRAI's high benchmark | $131.9M | $6.59M | $1.32M |
| $31.65 — the minister's Rs 3,000 estimate | $379.7M | $18.99M | $3.80M |
| $45.00 — our model's ARPU floor | $540.0M | $27.00M | $5.40M |
| $66.00 — our model's global ARPU | $792.0M | $39.60M | $7.92M |
Every ARPU in the left column is a benchmark or an assumption, none is an India price, and the arithmetic is ours. To scale it: at the top of TRAI's benchmark band, India would need 15.2 million subscribers to generate $100 million a year of spectrum charge — more than the 12.0 million Starlink has worldwide.
One thing the charge does not do is fall only on consumers. AGR is a revenue measure, and Starlink's June quarter shows why that matters: Connectivity revenue of $4,291M split into consumer $2,485M and enterprise and government $1,806M, the latter 42.1% of the segment and growing 108% year on year. A revenue-share charge lands on both halves of whatever India turns out to be.
What AGR excludes is only partly knowable. Under India's general framework it is applicable gross revenue less permissible deductions, with non-telecom revenue, broadcast-licensed activity, universal-service receipts and other income such as interest, dividends, property and foreign-exchange gains already stripped out. The specific deduction schedule attaching to a satellite authorisation is set in the authorisation terms, which are not published. The base is narrower than gross revenue; how much narrower is not establishable from the public record.
What this does to our model: 51 cents, and the reason is not that India is small
Our SPCX model carries the Starlink vertical bottom-up — 12.0 million subscribers, $66.00 ARPU drifting down 2.5% a quarter toward a $45.00 floor, against a 130 million household TAM, plus $1,914M a quarter of non-subscriber revenue growing 9% a quarter. It carries no India assumption and no country split at all. It cannot be revised for this, and it is not going to be.
What it can do is bound the effect. Run the absurd maximal case: pretend every Starlink dollar on Earth is Indian, and charge 5% of it as an operating cost, permanently. Against a base fair value of $187.12:
| Change to the Starlink vertical | Base fair value |
|---|---|
| Published base | $187.12 |
| 5-point EBITDA margin hit, whole segment, forever | $186.61 (−$0.51) |
| Starting ARPU cut 5%, $66.00 → $62.70 | $187.07 (−$0.05) |
| ARPU drift −2.5% → −1.5% a quarter | $189.55 (+$2.43) |
Fifty-one cents. Not because India is small — the run above assumes India is the entire company — but because of how this model values SpaceX. The terminal value is 20x EV/revenue applied to the final year, and a spectrum charge is a cost. It never touches the revenue line, so it is invisible in the exit multiple and shows up only in five years of discounted cash flow. A valuation anchored on a revenue multiple is structurally close to blind to a top-line revenue share.
The second row is the more uncomfortable one. Cutting starting ARPU by 5% moves fair value by five cents, because the model already assumes ARPU falls to its $45 floor regardless of where it starts. The obvious worry about India — that adding the largest emerging market yet drags blended ARPU down — is one this model has already conceded in full. The contested number is the floor, not the level, and the third row shows the asymmetry: a shallower drift is worth $2.43, nearly five times what the maximal charge costs.
That gives the one place this rate genuinely bites, and it is a price question rather than a valuation one. Our $45.00 floor is the level below which the consumer business stops covering terminal and bandwidth cost — a figure net to SpaceX. Skim 5% off the top and the gross price needed to clear the same floor becomes $47.37. At 4% it is $46.88; the fifth point is worth 49 cents a month of gross price. At the sub-1% SpaceX asked for, $45.45.
So in India, at the reported rate, SpaceX either charges $47.37 to net $45.00 and loses every subscriber who would have paid $45.00 but not $47.37, or it charges $45.00, nets $42.75, and is below its own cost-recovery line on our assumptions. In a market the regulator itself benchmarks at $7 to $11 a month, that spread is not a rounding difference — it is the whole argument about whether the rural case works at all. The model has no mechanism to charge that back to subscriber additions, which is precisely why the honest output here is a rate and an arithmetic, not a revised fair value.
This is an industry charge, not a Starlink one
Both TRAI documents write the recommendation for NGSO-based fixed satellite services as a class, not for a named operator. Amazon's Kuiper, Eutelsat OneWeb and Jio Satellite face the same rate on the same terms. Kuiper joined SpaceX in asking for under 1%, and it will pay whatever SpaceX pays.
That matters for how to read the competitive picture. A charge every satellite operator pays equally is a tax on the category against terrestrial fibre and fixed wireless, not a handicap on one company. Where it does discriminate is on business mix: the department's flat rate is cheaper for an operator with urban subscribers and dearer for one without, and among the four, Starlink's stated positioning is the most rural.
We priced the spectrum SpaceX bought in August — $19.6 billion for 65 MHz from EchoStar, against a vertical our model still scores at zero. This is the other side of the same ledger: spectrum SpaceX would rent, priced as a share of revenue rather than as a capital outlay, in what would be its largest unaddressed market. The mechanics of why satellite operators need these bands at all are in our guide to spectrum and path loss.
What to watch
- The gazette notification, and whether the concession is geographic or conditional. This is the single unresolved mechanism. A per-subscriber rural rate and a firm-wide enrolment test produce different blended rates for the same operator, and TRAI's own figures put the qualifying population at 5% versus 65%.
- Whether the Rs 500 urban charge survives. Reporting contradicts itself — the Financial Express item says the levy remains part of the framework; others say it was dropped in favour of the extra point. TRAI's arithmetic makes this worth 4–6% of urban AGR, so it is larger than the 5%-versus-4% argument everyone is having.
- The Rs 3,500 per MHz annual minimum and the rural terminal subsidy. TRAI reiterated both in December, including a modification allowing the government to fund user terminals through third-party implementers. Reporting on whether the department accepted either is conflicting.
- The first India price Starlink publishes. One number replaces every band in this piece. Until then the only two anchors differ by a factor of eight to twelve, and the December website figures were disavowed by name by a company officer.
- The remaining approvals. Cabinet, the inter-ministerial committee, and gateway security clearance. Starlink holds its licence and its space authorisation; a cleared spectrum rate is not an issued assignment, and an issued assignment is not a launch date.
The 5% spectrum charge, the one-point concession, the five-year term and the 2 or 3 September commission meeting are as reported on 8-9 September 2026 by Moneycontrol, the Economic Times, Business Standard, Mint, Hindu Business Line and the Financial Express, all citing unnamed officials. No gazette notification or departmental press note exists, and the rate is proposed rather than in force. Everything else attributed to the regulator or the department - the 4% recommendation and its formula, the 12 November 2025 back-reference and its conditional-discount wording, TRAI's refusal, its 4-6%-of-urban-AGR estimate and the 5%-versus-65% population figures - is from TRAI's published recommendations of 9 May 2025 and its response of 8 December 2025. The 8% licence fee, Starlink's 2025 licence and space authorisation, the under-1% ask and the disavowed December website pricing are press-reported. Every total, per-subscriber figure and grossed-up floor price is R40 arithmetic on those rates, converted at 94.80 rupees on 8 September 2026. June-quarter segment revenue is as SpaceX reported. The model section is our model as of 19 August 2026 and is assumption, not fact; it carries no India subscribers, revenue or country split, and nothing here revises it.
