How Did the Next Billion Users Become Collateral Damage in the AI Race?
The world added 160 million new mobile internet users in 2025, down from 190 million the year before. Network coverage keeps expanding and speeds keep climbing, yet the number of people coming online for the first time keeps falling. That’s the puzzle at the heart of the GSMA’s State of Mobile Internet Connectivity 2026 report: better infrastructure isn’t translating into more users.
Global connected population rose from 58% to 59%, or 4.8 billion people, while smartphone penetration moved from 54% to 56%, or 4.6 billion. The coverage gap has shrunk to just 3%, meaning only 270 million people now live outside mobile broadband range. But the usage gap, people who have coverage and still don’t go online, sits at 3.1 billion, 38% of the planet. Networks are reaching almost everyone. Getting people to actually use them is a different problem.
The primary barrier is cost. For the poorest fifth of the population in low- and middle-income countries, an entry-level smartphone eats up 44% of monthly income. In Sub-Saharan Africa, that climbs to 76%. Handset affordability has always dragged on adoption, but it’s now facing pressure from an unlikely source: the global AI buildout.
Memory prices more than doubled between Q3 2025 and Q1 2026, then jumped another 80 to 90% in Q2. Chipmakers are redirecting production toward the high-bandwidth memory that powers AI data centers, squeezing the supply of the older, cheaper memory that goes into entry-level phones. The devices poorer households rely on to get online are becoming harder to afford, right as billions still need one.
This week’s #TechTalkThursday looks at why network expansion alone can’t close the digital divide, how the AI chip race is reaching into the cheapest phones on the market, and what accountability further up the supply chain might look like.
Infrastructure growth, adoption standstill
Mobile broadband networks cover 97% of the global population, and more than 90% of the 3.4 billion people who still don’t use mobile internet live within that coverage. They’re not offline because the signal doesn’t reach them. Tower deployment used to be the story of digital inclusion. Now it barely moves the needle.
“Nigeria already sits at about 87 percent 4G coverage, yet more than half of MTN’s customers do not use the internet consistently. This is the heart of the usage gap. People are covered by the network but cannot benefit from it because the devices remain out of reach.”
-Tobe Okigbo, Chief Corporate Services and Sustainability Officer, MTN Nigeria
One bright spot: the number of people relying solely on borrowed or shared devices dropped from 690 million to 650 million (8% of the global population) in 2025, a shift the GSMA attributes partly to shared users finally getting their own devices. But that modest gain wasn’t enough to offset the bigger trend. Subscriber growth still slowed sharply, from 190 million new users in 2024 to 160 million in 2025, and that deceleration, not network coverage, is where the real story is.
Cheap phones are getting more expensive
Handset affordability remains the single biggest barrier to mobile internet adoption across low- and middle-income countries, and the report traces that squarely back to silicon.
Data centers pay premium margins for high-bandwidth memory, so component makers are shifting fabrication capacity away from the older, lower-spec memory entry-level phones depend on. It’s already showing up in retail prices. The Xiaomi Redmi A7, launched in April 2026 at around $110, costs 40% more than its predecessor did a year earlier, on nearly identical hardware. In India, the Realme C71 launched in mid-2025 at the same price as its predecessor and has since gone up almost 70%.
Global smartphone shipments are on track for the steepest single-year drop in the market’s history in 2026, down 14% year over year, roughly 174 million fewer units. The sub-$100 segment is taking the worst of it, projected to fall 36%, around 90 million fewer phones. Sub-Saharan Africa alone is expected to see 16 million fewer smartphones shipped this year, a drop of more than a quarter.
It reflects a structural reallocation of chip capacity, not a routine market dip, and the cheapest phones are absorbing most of the cost.
Who’s falling furthest behind
The barriers keeping people offline don’t fall evenly.
On gender, there’s real progress: women in LMICs are now 12% less likely than men to use mobile internet, down from 14% the year before, thanks to targeted digital literacy work. But women still make up 58% of the adult usage gap, so the gap is narrowing slower than the underlying inequality would suggest.
On geography, the picture is worse. Rural adults in LMICs are now 28% less likely to be connected than urban residents, up a full percentage point from 2024. Higher deployment costs and lower average revenue per user make rural rollouts commercially unattractive without direct intervention, and the gap has been widening for years.
Progress on gender hasn’t stopped the rural-urban split from getting worse. Whatever’s working for one divide doesn’t automatically work for the other, and any policy response has to treat them as separate problems.
Beyond the telecom industry
Closing the usage gap could add an estimated $3.5 trillion to global GDP by 2030. But the bottleneck now sits further up the chain than telecom operators can fix alone. Cell towers and spectrum still matter, but so does what happens inside the fabs producing the chips that go into every phone.
“Unless we protect the affordability of entry-level smartphones, billions of people could be excluded from the next generation of digital services before they even have a chance to experience the internet. The current memory price surge makes this risk stark and urgent.”
– Vivek Badrinath, Director General, GSMA
The GSMA’s own push for $40 4G smartphones across six African markets shows how tight the margin has become: a single entry-level memory chip now costs more than that target price on its own.
Protecting affordability at the bottom of the market will take more than operator subsidies. It means holding component manufacturers accountable for keeping base-tier production alive, so the AI boom doesn’t end up pricing the next billion users out of the internet before they get the chance to log on.
