Contents
Most plans to sell more through digital channels assume the customer already has a device good enough to use them. For a decade that assumption got safer every year: phones got cheaper, and the floor of the market kept dropping. In 2026 the floor started moving the other way, and the cause is not in African markets at all. It is the global competition for the memory chips that go into AI data centres.
This piece sets out what the evidence shows, what it does to the addressable digital market across African countries, and what a business should change as a result.
What is happening?
On 15 September 2026 the GSMA published its State of Mobile Internet Connectivity 2026 report. Its headline finding is slowing progress: 4.8 billion people were using mobile internet on their own device at the end of 2025, about 59% of the world’s population, but only around 160 million came online during the year, down from roughly 190 million the year before. Some 3.4 billion still do not use mobile internet, and the overwhelming majority are not waiting for a network — 38% of the world’s population lives inside mobile broadband coverage without using it, against 3% with no coverage at all. That difference between what is built and what is used is the usage gap, and it is the number that matters commercially.
The report’s new element this year is a supply-side one. The GSMA links the stalling to component costs: memory prices more than doubled between the third quarter of 2025 and the first quarter of 2026, then rose a further 80–90% during the second quarter of 2026. The cause is demand for AI infrastructure, which absorbs the same DRAM and NAND capacity that goes into consumer devices. TrendForce, which tracks memory contracts, forecast on 3 July 2026 that DRAM contract prices would rise another 13–18% quarter-on-quarter in the third quarter and NAND 10–15% — slower than earlier in the year, but still upward from a record base.
Cheap phones absorb that shock worst, because memory is a bigger share of what they are made of: Omdia’s August 2026 analysis puts it at close to 60% of the bill of materials for phones under $400, and above 64% for devices under $99. A doubling in memory prices does not add a few dollars to a $70 phone; it removes the price point.
The African evidence is already in the shipment data. Omdia reported on 20 August 2026 that Africa’s smartphone market fell 7% year-on-year in the second quarter of 2026 — its first contraction in three years — while the sub-$100 segment fell 34%, roughly three million units, and the average selling price rose $41 to $202. Omdia expects the African market to decline 26% across 2026 as a whole. The quarter varied sharply by country: South Africa grew 17%, while Nigeria fell 11%, Kenya 15% and Egypt 26%.
Affordability was already the binding constraint. The GSMA puts the cost of an entry-level internet-enabled handset at 44% of average monthly income for the poorest fifth of people in low- and middle-income countries, and at 76% in Sub-Saharan Africa. Reporting on the same report’s regional analysis puts Sub-Saharan Africa’s usage gap at around 66% of the population and its coverage gap at about 9% — the widest of any region.
What remains uncertain is how long this lasts and what it does to behaviour. Memory cycles have turned before. Whether 2026’s lost entry-level shipments are purchases deferred by a year, or people who stay on a feature phone for another three, the data does not yet establish.
Why it matters economically
For a business selling into these markets, the relevant consequence is not the price of phones. It is that the digital market stops growing on its own.
For a decade, the reachable audience for a website, an app or an online checkout expanded every year without anyone doing anything, and acquisition costs fell against a rising base. That tailwind has weakened, with a shrinking device pipeline behind it. The GSMA estimates that closing the usage gap would add $3.5 trillion to GDP between 2023 and 2030, about 90% of it in low- and middle-income countries — the scale of what is currently not happening.
The second consequence is a change in the device mix, working in two directions at once. Omdia’s Manish Pravinkumar described “a forced upward shift in the African market. Vendors can no longer profitably manufacture $75 smartphones, while consumers who need connectivity are increasingly having to stretch their budgets towards $200-plus devices.” Those who buy anyway end up with a better phone. Those who cannot keep an older one for longer, or stay on a feature phone. The middle thins out, and the installed base ages.
Our view: the practical planning number is not the shipment forecast but the replacement cycle. Shipments falling 26% in one year does not remove 26% of your customers’ phones. It slows the rate at which those phones are replaced — so the browsers, operating system versions and storage limits you design for in 2027 and 2028 will be closer to today’s than the last few years have trained anyone to expect.
What it means for African businesses
Who is affected depends heavily on which market and which customer. The GSMA’s affordability figures describe the poorest fifth of the population; a business selling professional services in Lagos or Nairobi is not selling to that group, and its customers’ devices are unaffected. Omdia describes Africa as a two-speed market: South Africa, Egypt, Morocco and Tunisia moving towards replacement and 5G, while much of Sub-Saharan Africa still depends on first-time buyers.
So the exposure is specific. It is highest where growth depends on continuously acquiring new, first-time or price-sensitive digital customers: consumer e-commerce and retail, consumer financial services, agent networks, ticketing, and any model whose unit economics assume the customer will self-serve rather than be served by a person. It is lowest for businesses selling to other businesses, to salaried professionals, or through a sales team, where the device was never the constraint; for them this is a cost question about staff devices, not a demand question.
A third group is worth naming: businesses whose customers are on the network but not on smartphones. That segment is not shrinking this year — which is why USSD and SMS still carry real commercial traffic, and why the channel mix across the markets we work in rarely reduces to “build a website”.
Technology and AI implications
The honest starting point is that AI does not help here. This constraint was created by AI infrastructure demand, and no AI feature will make a customer’s phone cheaper or their data bundle bigger. Adding AI to a web experience usually makes it heavier — more JavaScript, more round trips, more time on a slow connection — which is the wrong direction for the customers most affected. Where AI earns its place is internal: handling the enquiry volume that arrives through cheap channels like WhatsApp and SMS, so a low-bandwidth channel can still be answered quickly.
The technology response that addresses the constraint is unglamorous. Page weight and time to first interaction on a mid-range Android phone over a congested network. Server-rendered pages instead of heavy client-side applications. Testing on a low-end device rather than a development laptop. Progressive web apps rather than an app-store download the customer has no storage for. Offline-first behaviour for field and agent tools, which is a design decision in web and mobile applications, not a plugin. None of this is new advice; what has changed is that the device base will no longer improve underneath you.
Practical business implications
Four questions to answer with your own data this quarter.
What does your traffic actually run on? Pull device model, browser version and connection type from analytics for the last twelve months, split by market. If your conversion rate on cheaper devices sits materially below your average, that gap is a structural cost, not a rounding error.
What does your site weigh? Measure the transfer size of your key pages and the time to interaction on a mid-range device over a throttled connection. A checkout that takes eleven seconds to become usable is a pricing decision you did not make deliberately.
Where does a slow or failed session go? If a customer cannot complete an online order, the alternative should cost them almost nothing — a WhatsApp thread, a call-back request, a USSD step — with the context carried across so nobody starts again. That hand-off is a system design question; the equivalent problem on the payment side is covered in mobile money reconciliation.
Is device financing part of your customer’s path? Omdia reports vendors turning to financing partnerships rather than discounting. Where operators and lenders extend handset financing, recognising a financed-device customer reaches people a price cut would not have.
RevenueStack perspective
Read economically, this is a demand-side constraint arriving through a supply chain nobody in African commerce controls. That distinction matters for where you spend. When a market’s growth comes from new users arriving, the highest return is in acquisition: reach, awareness, top-of-funnel spend. When new arrivals slow, the same money returns more in conversion and retention. That is a reallocation, not a cut, and it is testable: compare your cost per new digital customer this year against the cost of lifting conversion by a point, and let the numbers decide.
The second implication is architectural. If the installed device base ages rather than improves, performance stops being a technical nicety and becomes a commercial specification. Our position when we build a digital commerce system is that the target device and the target connection belong in the brief alongside the revenue target, because they decide whether it is reachable at all.
The third is about channels. A business that depends on a single digital channel is exposed to the device economics of that channel. One that can take an order through the web, a messaging thread or a person — with one customer record behind all three — is not. That redundancy costs something to build; in markets where the cheapest phone just got more expensive, it is worth what it costs.
Conclusion
The cheapest way to get online stopped getting cheaper in 2026, and the evidence from the GSMA and Omdia is already in African shipment data. The implication is not that digital channels matter less. It is that the audience for them will grow more slowly, sit on older devices for longer, and split more sharply by market than continental averages suggest. The businesses that hold their ground will be the ones that measure what their customers actually browse on, build for that, and keep a cheap human or messaging path open for everyone the website loses.
Sources
- GSMA warns of emerging global AI divide as three billion people remain offline and smartphone component costs soar — GSMA, 2026-09-15
- The State of Mobile Internet Connectivity 2026 — GSMA Intelligence, 2026-09
- Omdia: Africa smartphone market expected to decline 26% in 2026 following first contraction in three years — Omdia, 2026-08-20
- Beyond the $100 smartphone: how rising costs are reshaping Africa's budget smartphone market — Omdia, 2026-08
- AI Server Demand Continues to Support Memory Prices in 3Q26, but Gains Moderate as Consumer Demand Weakens and High Base Effects Take Hold — TrendForce, 2026-07-03
- Sub-Saharan Africa faces digital divide as handset costs and usage gap persist — Capital Newspaper (Ethiopia), 2026-09-20