Summary

  • BSCCL’s 2024–25 report states 7,220 Gbps of ultimate achievable capacity, 6,850 Gbps lit, 2,946 Gbps of monthly-average utilisation and a 43.01% utilisation ratio — and 43.01% reproduces exactly against the 6,850 Gbps lit figure, not the 7,220 Gbps design figure, which yields 40.80%.
  • At the cable layer the sums close both ways: SMW4’s 4,650 Gbps plus SMW5’s 2,570 Gbps achievable equals 7,220, and 4,650 plus SMW5’s 2,200 Gbps lit equals 6,850. At the product layer they do not: IPLC sales of 3,123.4 Gbps plus local IP transit of 323.84 Gbps come to 3,447.24 Gbps against a stated June 2025 utilisation of 3,563 Gbps.
  • That 115.76 Gbps difference is 3.25% of the reported June figure. The inspected tables carry no reconciliation note, so the gap is recorded here as a gap and not explained.
  • Wholesale capacity sits inside regulator-set floor and ceiling bands rather than being bid, so the tariff cannot tell a reader whether the pool is tight. Utilisation against a stated denominator is the only scarcity signal the published record supports — and the report does not print that denominator.

Bangladesh Submarine Cable Company Limited — BSCCL on its own filings — closed its fiscal year on 30 June 2025 and published the annual report on 23 October 2025 (BSCCL Annual Report 2024–2025). The document is the single most detailed public account of the country’s international data capacity held by the party that lands the cables. Within it, the same twelve months carry four distinct capacity quantities: 7,220 Gbps of ultimate achievable capacity, 6,850 Gbps of lit-up capacity, 2,946 Gbps of monthly-average utilisation, and 3,563 Gbps of utilisation in June 2025 stated to exclude 863 Gbps of IPLC exports.

Four numbers, one year, one operator. None of them is wrong in itself. The relevant question is which of them a reader can check, and what happens to the story if the checks are run.

One year, four capacities

The report is careful about vocabulary in a way that most operator documents are not. Capacity is described as present, lit-up or ultimate achievable, and never simply as installed. That distinction matters because the gap between what a cable system can carry at full design and what has been lit and equipped is precisely where money has to be spent before capacity becomes sellable. The company maintains 370 Gbps of difference between the two figures — 7,220 minus 6,850 — and the labels keep that difference visible instead of collapsing it into a single headline.

Utilisation is the fourth quantity and the one that carries the most economic meaning. Monthly-average utilisation of 2,946 Gbps against a 43.01% ratio is the company’s own statement about how full the pool was across the year. June 2025 utilisation of 3,563 Gbps is a point-in-time measurement at year-end, taken with 863 Gbps of IPLC exports excluded from the count. The two are not substitutes. One is a year, the other is a day. The distinction is doing real work in the numbers below, and the report itself is explicit about it: the FY2023–24 comparative figure of 2,474 Gbps is labelled a monthly average in the 2024–25 report (BSCCL Annual Report 2023–2024).

The cable layer closes, twice

Start with what works. The cable split is the part of the ledger a reader can verify with a pencil.

SMW4 carries 4,650 Gbps. SMW5 carries 2,570 Gbps of ultimate achievable capacity. Sum them: 4,650 + 2,570 = 7,220 Gbps. That is exactly the ultimate achievable figure the report states for the whole company.

Now take lit capacity. SMW4’s 4,650 Gbps is unchanged, and SMW5 contributes 2,200 Gbps lit. Sum them: 4,650 + 2,200 = 6,850 Gbps. That is exactly the lit-up figure the report states.

The directional split closes a third time, at a smaller scale. SMW5’s 2,200 Gbps lit is composed of 2,100 Gbps toward Singapore and 100 Gbps toward France. 2,100 + 100 = 2,200. Three sums, three exact hits, no rounding slack.

This is not a trivial observation. It establishes that the company reconciles its capacity across the cable layer with enough precision that a reader with a calculator can reproduce every total. The capability to publish a closing ledger exists inside this reporting process. It is worth remembering that when the sales tables come up, because the failure there is not a failure of arithmetic competence.

The denominator nobody printed

Take the utilisation ratio. The report states 43.01% for FY2024–25 alongside monthly-average utilisation of 2,946 Gbps. Divide 2,946 by 6,850 and the answer is 43.01%. Divide the same 2,946 by the larger, more flattering design figure of 7,220 and the answer is 40.80%.

The reported ratio is therefore computed against lit capacity. The 2.21 percentage-point difference between the two possible answers is not a rounding artefact but a choice of denominator, and that choice is made silently: in the passage inspected here, the ratio appears with the Gbps figure beside it and without a statement of which capacity figure sits underneath.

The prior year confirms the same convention. FY2023–24 reported monthly-average utilisation of 2,474 Gbps and a ratio of 36.12%. 2,474 ÷ 6,850 = 36.12%, exactly. Against 7,220 the same numerator gives 34.27%. So the denominator has been stable across two reporting years, which is the reassuring half of the finding. The unsettling half is that a reader who wants to know whether the pool is tight has to reverse-engineer the denominator before the ratio means anything at all.

Where the totals stop meeting

Now the part that does not close.

At 30 June 2025, the report states capacity sales of 3,123.4 Gbps for IPLC domestic service and 323.84 Gbps for local IP transit. Add them: 3,447.24 Gbps. That is the sum of what the company says it had sold into the domestic market at year-end.

The stated June 2025 utilisation is 3,563 Gbps. Subtract the sales subtotal and the difference is 115.76 Gbps. As a share of the reported June figure, that is 3.25%.

The obvious first instinct is to reach for the export rows, which are also published at 30 June 2025: 863 Gbps of IPLC exports and 10 Gbps of IP-transit exports. Add all four rows and the total is 4,320.24 Gbps, which sits 757.24 Gbps above the reported June utilisation. Tempting — and wrong. The 3,563 Gbps figure is stated to exclude the 863 Gbps of IPLC exports. Adding excluded traffic back into a total that explicitly excludes it compares two different boundaries and proves nothing. The domestic test is the only valid one, and it leaves 115.76 Gbps standing.

What the inspected tables do not contain is any line explaining that difference. A gap of this shape can be produced by several ordinary mechanisms: a measurement point one hop away from the billing point, a year-end stock compared against traffic averaged over a different window, a service line that is neither IPLC nor local IP transit, or simply a definition of “capacity sold” that counts committed rather than delivered capacity. Any of those would be unremarkable if stated. None is stated in the pages read for this article, so none is claimed.

The number is recorded as what it is: two published figures that do not reconcile, in a document whose cable-level figures reconcile exactly.

That asymmetry is the substantive finding. An operator that can make 4,650 and 2,570 land precisely on 7,220 has demonstrated the accounting discipline to close a 115.76 Gbps difference in a footnote. At the layer where capacity turns into revenue, the ledger stops being self-checking.

What actually grew

Growth is where the reporting is strongest, provided the boundaries are held constant. Both June figures exclude the same 863 Gbps of IPLC exports, which makes the year-over-year comparison legitimate: 3,563 − 2,696 = 867 Gbps, or 32.16% more traffic in June 2025 than in June 2024.

The year-average measure is more modest and more meaningful: 2,946 − 2,474 = 472 Gbps, a 19.08% increase, with the utilisation ratio moving from 36.12% to 43.01%, or +6.89 percentage points.

Two growth rates, one year, one business: 32.16% at the June point and 19.08% across the average. The difference is not a contradiction. A single month can carry a seasonal peak, a new customer’s ramp, or a one-off event, and an annual average dilutes all three. Readers who quote the June number as the growth rate of Bangladeshi international data demand are quoting the most flattering single observation in the document.

The year-end stock position is worth putting in plain terms. Against 6,850 Gbps of lit capacity, 3,563 Gbps of June utilisation is 52.01%. Against the full 7,220 Gbps the same traffic is 49.35%. On either basis, the company ended its fiscal year with more than half of its usable international capacity unsold for the month — 3,287 Gbps of lit capacity not carrying traffic at the measurement point. That is not a distress signal for a national landing party, whose purpose includes holding capacity ahead of demand. It is a reminder that “utilisation” in this report describes headroom, not congestion.

The price that cannot speak

Now the mechanism that makes the utilisation figure so consequential, and the tariff so uninformative.

Wholesale international data capacity in Bangladesh is sold by IIG licensees — International Internet Gateways, which the regulator defines as carrying international data traffic and connecting ISP and BWA operators to the global internet, a category distinct from the IGW licence that carries international voice traffic (Regulatory and Licensing Guidelines for International Internet Gateway Services in Bangladesh). Under clauses 14.01–14.03 of that guideline, a licensee needs the Commission’s written approval before changing an approved tariff, and must file interoperator agreements with tariff breakdowns. Price is not discovered in this market. It is approved.

The approved bands show what that approval has done over four years. The schedule effective 1 September 2021 set IIG tariffs in BDT per month per Mbps across commitment bands, with a Dhaka price of 365 BDT/Mbps at the smallest band falling to 330 at the largest, and outside-Dhaka prices running from 399 down to 355 (BTRC IIG tariff approval, effective 1 September 2021).

The schedule effective 1 July 2025 restates the market as floor and ceiling pairs: 265–299 BDT/Mbps for 1–1,000 Mbps, 260–295 for 1,001–5,000, 255–290 for 5,001–10,000, 250–280 for 10,001–25,000, 245–275 for 25,001–50,000, 240–265 for 50,001–100,000 and 230–255 above 100,001 Mbps. The floor applies inside Dhaka; outside Dhaka a price anywhere between floor and ceiling is permitted (BTRC IIG tariff approval, effective 1 July 2025).

At the ceiling, the smallest band falls from 365 to 299 BDT/Mbps — 66 lower, or 18.08%. The largest band falls from 330 to 255 — 75 lower, or 22.73%. Both schedules state a five-year term subject to regulatory change.

Two details in that comparison deserve more attention than the headline cuts. First, the 2021 document’s heading states BDT per month; the 2025 heading states BDT per Mbps and omits a billing interval altogether. That is a difference between two documents, not proof that the 2025 schedule is not monthly — but it means a reader comparing the two is comparing a monthly price against a price whose period is unstated on the page.

Second, the ceilings are not the prices. Nowhere in the documents examined for this article did BSCCL publish the per-Mbps rate it actually charges a gateway operator. The bands are the envelope the regulator permits; the operator’s position inside the envelope is not in the record read here. A 2025 band that permits anywhere between 230 and 255 BDT/Mbps at the top commitment level contains a 9.8% spread between its own floor and ceiling. What a buyer pays within that spread is a negotiation, and negotiations are not published.

The retrieval itself produced one caution worth recording precisely because it affects the tariff timeline. A copy of a BTRC approval hosted by bdHUB Limited returned two different documents from the same URL: a rendered page showing an approval dated 19 April 2026 that cites a 17 February 2026 circular effective 18 February 2026, with bands from 160–210 down to 120–170 BDT/Mbps; and extracted text from that same URL returning a document dated 12 September 2021 with the 365–399 pricing (BTRC tariff copy hosted by bdHUB Limited). Neither version is used as a price fact in this article. The underlying February 2026 circular was not independently retrieved, and the regulator’s own website timed out during the research for this piece, so the authoritative tariff timeline could not be confirmed from BTRC directly. That limits the tariff claim. It does not establish that any later schedule is absent.

What survives all of that is the structural point. In a market where the regulator sets a floor and a ceiling and the operator chooses a point between them, price carries no information about tightness. A price that cannot move with demand cannot report demand. So the only scarcity signal a reader can check is utilisation — and utilisation, as shown above, depends on a denominator the report does not print.

What a reader can actually check

Put the pieces in order and the practical consequence is specific.

For a gateway operator or an enterprise buying international capacity in Bangladesh, the useful question is not “what is the tariff” but “what is the measurement point and the denominator”. A contract written on committed capacity at the handover point and a report written on a year-end stock excluding IPLC exports are measuring different things; the 3,447.24 Gbps against 3,563 Gbps gap is what that difference looks like when it appears in public.

For a reader tracking national capacity, the cheap check is the reverse computation: divide the stated utilisation in Gbps by 6,850 and see whether the published ratio matches. It has matched for two consecutive years. If a future report publishes a ratio that only reproduces against 7,220, the denominator has changed and every year-over-year comparison in that report is compromised.

For the company itself, the fix costs a sentence. A reconciliation line stating that product sales are measured at one point and utilisation at another would convert a discrepancy into documentation. That the cable layer already achieves this precision makes its absence at the sales layer a choice about what to explain, not a limit on what can be explained.

The larger reading is the one that matters beyond Bangladesh. A state landing party with administered prices and a physical monopoly on international data capacity does not need to convince a market of anything — there is no market signal to send. What it needs is a record its citizens, its regulator and its customers can audit. BSCCL’s record audits beautifully at the cable layer and not at all at the commercial layer. The first shows what the company can do. The second shows where it stopped.

A company that can make 4,650 and 2,570 land exactly on 7,220 has the arithmetic discipline to close a 115.76 Gbps gap in a footnote. It has not. Until it does, the most auditable number in Bangladesh’s international data capacity is the one the report still does not print: the denominator. The company’s own entry in the BTW directory records five earlier reports on this operator; none of them had this ledger to check.