In 1989 a family-owned Mauritian group launched a cellular network in partnership with Millicom-Comvik. The group describes EMTEL as the first mobile telephone operator in the southern hemisphere.
Set aside the ranking, which is contestable, and consider the timing on its own. In 1989 mobile telephony was a frontier technology in the wealthiest markets in the world. Handsets were expensive, coverage was patchy, and the business case rested on assumptions about adoption that nobody could yet verify. An island of under a million people, several thousand kilometres from the nearest major market, stood up a network in that environment.
Thirty-seven years later the same lineage runs through the METISS submarine cable, live since 2021 and connecting Mauritius to Réunion, Madagascar and South Africa; through 5G reaching 60% of the island’s localities by July 2023 and extending to Rodrigues in 2024; through Airbox fibre-to-the-antenna from 2015; and through blink, a mobile banking service launched in 2022.
By any comparative measure this is strong infrastructure for an economy of this size. Which makes the most interesting fact about it a negative one: the country’s first national artificial intelligence plan, “AI for All”, arrived after all of it.
Why infrastructure leads and policy follows
This ordering is not a Mauritian peculiarity; it is the normal shape of technology adoption, and it is worth understanding before criticising it.
Infrastructure is built by firms with a commercial reason to build it. A telecommunications operator lays a cable because it can bill for the capacity. It does not need a national plan, an inter-ministerial committee or a consultation. It needs a licence and a balance sheet.
Policy is built by states, which must consult, legislate, budget and defend the result electorally. Policy is therefore structurally slower, and generally arrives once a technology is established enough that the questions it raises have become concrete.
The consequence is a persistent gap between what an economy can technically do and what it is institutionally organised to do. Mauritius has a fast network and, on the public assessment offered at the close of the Business Mauritius presidency, an insufficiently fast policy cadence.
The Singapore comparison, precisely stated
The comparison drawn publicly was specific rather than rhetorical: Singapore adopted two successive national AI plans within a few years, and Mauritius needs to accelerate — strengthening digital connectivity and the fintech environment alongside the plan itself. Business Mauritius contributed to “AI for All”.
What matters in that comparison is the word two. Publishing a national technology plan is not difficult; most economies have one. Publishing a second plan that supersedes the first, within a few years, requires something harder: an institutional willingness to declare the previous document obsolete while the people who wrote it are still in post.
A single plan held for a decade is not evidence of strategic conviction. It is evidence that the review mechanism does not work.
The measure of a technology policy is not the quality of the document. It is the interval before the document is replaced.Editorial analysis
Why cable diversity is a sovereignty question
The METISS cable deserves more attention than it usually receives, because its significance is not commercial.
An island economy’s services exports — financial services at 11.9% of GDP in 2023, business process outsourcing, technology services, and the digital layer of a tourism industry that recorded 1.44 million arrivals in 2025 — depend entirely on international bandwidth. A country with one submarine cable has a single point of failure for its entire tradeable services sector. Cable cuts are routine events; anchors and seabed movement cause them regularly worldwide.
A second independent route converts a national vulnerability into a manageable operational risk. That is infrastructure in the same sense that a port or an electricity grid is infrastructure, and it happens to have been delivered through a private consortium rather than a state programme.
The three layers, and where the constraint sits
Layer one: physical capacity
Cable, spectrum, towers, fibre, data centres. Mauritius is well provisioned here. This layer is capital-intensive but conceptually simple, and it is the layer private operators build willingly because the revenue is billable.
Layer two: regulatory environment
Data protection, cross-border data flows, financial technology licensing, digital identity, e-signature validity, cloud procurement rules. This is where the fintech environment named as needing strengthening actually lives. It is cheap to build in capital terms and expensive in institutional terms, because it requires several ministries and regulators to agree.
Layer three: human capability
Engineers, data scientists, security specialists, regulators who understand the systems they supervise. This is the binding constraint, and it connects directly to the skills argument made throughout the same period: an economy of 1.26 million people cannot produce, from a standing start, the specialist depth that a frontier technology sector requires.
Layer one without layers two and three produces exactly what an observer would predict: excellent connectivity, and a fintech sector smaller than the connectivity would support.
What a serious acceleration would look like
If the diagnosis is cadence rather than direction, the response is procedural rather than technological. Four things follow.
Shorter review cycles. A national plan with a scheduled replacement date, not an open-ended one. The commitment is to review, not to any particular content.
Regulatory sandboxes with real exit ramps. Sandboxes are common; sandboxes with a defined path to full authorisation are not. Without the exit ramp, a sandbox is a permanent holding pen.
Skills policy tied to sector policy. A national AI plan and a work-permit regime that treats AI engineers as ordinary applicants are not consistent instruments.
Procurement as demand. The state is usually the largest single buyer of technology in a small economy. Public procurement that favours modern architecture creates a domestic market that private demand alone cannot.
The asymmetry worth remembering
The 1989 network was built by a firm that did not wait for a national telecommunications strategy, because none existed and none was required. That is the advantage of private infrastructure: it moves at the speed of a board decision.
It is also its limit. A firm can build a network. It cannot legislate a data protection regime, accredit a university programme, or make a regulator competent. Those require the state, and the state moves at a different speed. The gap between the two is the whole of the current problem, and it is a problem of institutions rather than of technology.
Sources referenced in this essay: Currimjee Group published corporate history (currimjee.com); Le Mauricien, 28 September 2023; Business Mauritius and L’Express, 1 October 2025; Business Magazine cover interview; Defi Media; World Bank; African Development Bank; International Monetary Fund; Mo Ibrahim Foundation, 2024 Ibrahim Index of African Governance. This essay is editorial analysis prepared for anilcurrimjee.com and is not a statement by Anil C. Currimjee except where a sourced quotation appears.