Summary

  • The express open-access duty for Samoa's cable is written to the state, not the operator: the World Bank/IDA Financing Agreement requires Samoa to guarantee cost-based, non-discriminatory access for all access seekers, while the larger ADB grant asks the regulator only to establish wholesale tariffs including price and non-price terms.
  • The operator's binding moment was a disbursement condition — money moved only once the company offered a reference interconnection offer and the regulator accepted it — and when ADB closed the books, 41 of 49 covenants were compliant while both hard failures were reports that never arrived.

The Tui-Samoa cable is a two-bank project with a three-sentence public story: Samoa got international bandwidth, wholesale prices collapsed, the development banks closed their books. Everything interesting sits in the paperwork between those sentences, and the paperwork does not point where a reader would expect.

Start with the address line. The strongest access language in the whole stack is not in the Asian Development Bank's grant. It is in the World Bank's IDA Financing Agreement for the Samoa Connectivity Project, Grant D063-WS of 11 August 2015, and it is phrased as an obligation of the Recipient: Samoa must ensure independent regulation which "guarantees open access to international communications services by all access seekers on a cost-based and nondiscriminatory basis", under Part VI of the Telecommunications Act, within three months after effectiveness and throughout implementation (IDA Financing Agreement).

ADB's own regulatory-support schedule is lighter. Grant 0458-SAM (SF), US$25 million, dated 3 December 2015, lists as an activity to be supported the establishment of "wholesale tariffs including price and non-price terms for access to all international fiber option bandwidth services" (Grant Agreement). The word printed there is "option"; I have left it as extracted rather than quietly correcting it to "optical". The Project Administration Manual goes a step further in prose — "Access to capacity will be offered by SSCC on a fair and equitable basis to any licensed operator" — and hands the Office of the Regulator oversight of sales and services so that pricing stays cost-based (Project Administration Manual).

Look at what those three sentences are. One is a covenant owed by a state. One is an activity to be funded. One is a commitment recorded in a project administration manual. None of them is a clause in a contract signed by the cable company that a wholesale buyer could point at.

The teeth were a payment condition

There is a company-facing instrument, and it is worth reading closely, because it is the only place the operator is directly bound. The IDA agreement requires the subsidiary agreement to oblige the company to offer the regulator a reference interconnection offer, or equivalent, ensuring cost-based and non-discriminatory services on terms acceptable to the bank and the regulator. Then the enforcement: Category (1) withdrawals require satisfactory evidence that the company offered that offer and the regulator accepted it.

That is a gate, not a ratchet. It creates one strong moment, before money moves, and nothing continuous afterwards. The World Bank's completion report describes the access outcome exactly that way in retrospect — open and non-discriminatory international bandwidth secured through a covenant plus a Component 1 disbursement condition requiring a reference interconnection offer (World Bank ICR). It is not scored as a separate open-access compliance finding. That distinction matters, because a disbursement condition is spent once it is satisfied.

The Report and Recommendation of the President supplies the structure the gate sits inside. It describes the company as the sole wholesaler of cable capacity to retail service providers, with government retaining strategic control through subsidiary-agreement covenants on dividend policy, capacity pricing and consent to major decisions affecting capacity availability or affordability. It also records the onlending terms for the US$25 million: 1% interest over 30 years including seven years' grace, subject to ADB endorsement, with a later review of repayment sustainability while preserving low wholesale prices (RRP). Co-financing is US$16 million from IDA and US$1.5 million from Australia's DFAT grant (World Bank ICR).

The second instrument is a return cap. An appraisal-level ADB financial analysis states: "Return on equity is capped at a maximum of 10% based on respective equity contributions and payable only when the SSCC maintains to be profitable" (Financial Analysis). That is an appraisal document, not the company's Articles, which I did not obtain. Read it as the lenders' stated design, not as quoted corporate law.

Forty-one out of forty-nine

ADB's completion report, December 2021, keeps a covenant ledger: 49 covenants, 41 complied, 5 partially complied, 2 not complied, 1 to be complied (Completion Report). Forty-one of 49 is 83.7% by my division on the reported count.

The two hard failures are the interesting part, and neither is about access. Grant Agreement schedule paragraphs 7(a) and 10 are rated "Not complied with": semiannual safeguards reporting and regular progress reporting.

The report says it without decoration: "Regular progress reports were not received by ADB during the project." Sections 4.02(a)-(b), the audit and financial reporting bundle, are rated partially complied, and the record behind that rating is a stack of late paper — audited project financial statements late for FY2016 to FY2018 and timely for FY2019 to FY2020; audited entity statements late for FY2016 to FY2019 with FY2020 outstanding; only the final safeguards report actually delivered.

The clock was not vague. Section 4.02(a)(iii)-(v) requires annual independent audits and English-language audited statements, audit report and management letter within six months after each related fiscal year ends; section 4.02(b) requires the bank to publish those statements and the audit opinion within 30 days of receipt. Six months is generous by audit standards and it was still missed for three consecutive years.

Under the Project Agreement as reproduced in the report's Appendix 7, sections 2.01-2.07, 2.08(a)-(b), 2.09(d) and 2.10-2.15 are rated compliant, 2.08(c) is "to be complied", and 2.09(a)-(c) partially complied. Section 2.08(b) credits quarterly financial reports to the project management unit, not to government progress reports — a different return from the one that failed. ADB's catalogue entry dates a Project Agreement between the bank and the company to 2 November 2016, but the PDF did not come down: the catalogue page returned HTTP 403 and the printable alternative HTTP 404. Everything I can say about that agreement comes from the completion report's reproductions (PA catalogue entry).

Two targets, two dates, no single price

Now the number everyone quotes. ADB's monitoring framework set a 2014 baseline and 2021 targets in US$ per Mbps per month: wholesale 1,500 to 300, residential 43 to 19, business 630 to 200 (Project Administration Manual). The completion report's outcome table records 56 for September 2021, citing the Office of the Regulator (Completion Report). That is 18.7% of the target and 3.7% of the baseline — a 96.3% reduction against the baseline, my division on the reported balances, not a published rate of change.

The World Bank's figure is different and the difference is not rounding. Its results annex records a wholesale baseline of US$1,500/Mbps/month dated 19 June 2015, an original target of US$400/Mbps/month, and an achieved US$72/Mbps/month on 30 June 2022 (World Bank ICR). After 2018 that indicator measures a 1-Gbps purchase tier, and larger purchases could obtain lower rates. Two financiers, two targets, two dates, two purchase tiers. Report them side by side; do not average them into "the price of Samoan bandwidth".

Two unit traps sit in the same tables and I will not smooth them. ADB's 2021 retail outcomes are 3.82 and 2.74 US$ per GB at 25 Mbps — a per-gigabyte measure sitting beside per-megabit targets. And the World Bank's paragraph 28 prints the achievement as US$72/Gbps/month while Annex 1 and Annex 7 print US$/Mbps/month. One of those two is wrong; the annexes are the ones carrying the definition.

Purchase size moves the unit rate inside ADB's own narrative: 2021 wholesale at US$160/Mbps/month for a 500-Mbps purchase and US$54/Mbps/month for a 5-Gbps purchase (Completion Report). Across those two points the larger purchase pays 33.75% of the smaller one's unit rate — my division again; the report states no discount policy. A quoted rate without a tier is not a price.

What the record does not contain

Four documents that would settle this were not obtained: the executed subsidiary agreement, the company's Articles, the accepted reference interconnection offer, and the ADB-company Project Agreement as a document rather than an appendix. I did not inspect the accepted offer's terms, its pricing, its dispute clause or its acceptance record. That is a gap in what I could read, not a finding that it was never accepted, never published or never used.

The Grant Agreement, the Project Administration Manual and the Report and Recommendation of the President were read in public mirrors; the exact ADB-hosted PDFs for those three were not recovered. The open-access covenant quoted above is the IDA one, addressed to Samoa as Recipient; it is not an ADB grant clause, and the two instruments should not be merged into a single requirement.

Covenant ratings are the lender's own legend — complied, partially complied, not complied, to be complied. "Not complied with" is a completion-report rating of a covenant, not a legal finding, a penalty or an enforcement action. And no document in this stack connects the missed reports to the access outcome in either direction, so I will not draw the arrow.

Tariff figures are the financiers' reported outcomes citing the Office of the Regulator. They are historical. They are not today's wholesale price in Samoa, and I make no claim about current wholesale, retail or transit pricing.

Earlier BTW reporting stands as context only: the 13 October 2026 piece on the wholesale offer's three-year lease against a 15-year IRU, the 21 October 2026 piece on Order 2026/T01, and the 10 November 2026 directory card (Samoa Submarine Cable Company). This briefing does not re-argue the order, the lease-versus-IRU split or the card's reference count.

Three readers, three things to take away

A household promised a cheaper connection should know what the money actually obligated: a state covenant to keep regulation independent and cost-based, an operator obligation to produce an interconnection offer once, and a price that fell far past its target. The continuous check that would have proved it — quarterly progress reports, annual audited statements — is the part that failed.

A competitor seeking wholesale capacity should know which instrument names a duty to serve it. The answer is uncomfortable. The duty runs to Samoa, through regulation; the operator's direct obligation was an offer delivered to a regulator, gated on a disbursement. What that means for one specific buyer depends on a document I could not read.

A regulator or lender should know what the ratings ledger actually scores. It scores paper, and the paper that slipped was the lender's own information channel. The price fell and the window closed in the same ledger — and only one of the two was ever measured.