Micron's gross margins doubled from 22% to 56% in a single year. Samsung's memory division posted operating profits up 250% year over year. SK Hynix beat Samsung in annual profit for the first time in its history. The memory industry is not suffering through a crisis. It is thriving inside one. The question is not whether the shortage is real. It is who chose it, who profits from it, and who absorbs the damage.
Nobody is asking the obvious question: what happens when three companies that control 93% of the world's DRAM supply all decide, simultaneously, that your phone and your laptop are lower priority than an AI data center? The answer, it turns out, is playing out right now in IDC forecasts and in the price tags at electronics stores across Lagos and Delhi and São Paulo.
Follow the Incentives to Their Destination
The shared facts tell you the price story. DRAM doubling quarter over quarter. PC vendors warning of 20% hikes. Smartphone shipments heading for their lowest annual volumes since 2013. What those numbers don't tell you is why nobody in a position of power is trying to stop it.
The answer is margin. HBM stacks for AI accelerators consume roughly three times the wafer capacity per bit as standard DDR5, and they sell at enormous premiums. SK Hynix has sold out its entire HBM capacity through 2026. Micron has locked in every unit. Samsung is ramping HBM production by 50%. These are not companies scrambling to balance competing demands. These are companies that have made a calculated decision about which customers get served first. As Counterpoint's Tarun Pathak put it, memory companies are asking smartphone vendors to "stand in line behind the hyperscalers."
The market is functioning exactly as designed. Scarce resources flow to the highest bidder. Microsoft, Google, Meta, and Amazon are the highest bidders. A school purchasing committee in Nairobi is not.
Consider what this means at the other end of the supply chain. Across Africa in 2025, 81% of smartphones shipped were priced below $200. Sub-$100 devices accounted for 22% of shipments in South Africa alone. Omdia now forecasts a 23% decline in African smartphone shipments in 2026, nearly double the global average, precisely because the continent's market is concentrated in the price tiers most exposed to component inflation. A first-time buyer in Nigeria, where 99% of smartphones are imported, has no domestic manufacturing cushion, no long-term supply contract, no leverage. She simply pays more or goes without.
The GSMA reported in November that buying a smartphone in Africa already costs about 26% of monthly GDP per capita. For the bottom 40% of the population, that figure rises to 67% of monthly income. Now imagine layering a 14% average price increase on top of that baseline, with IDC warning the sub-$100 category may become "permanently uneconomical." The math is not subtle. We are talking about a structural rollback of connectivity gains that took a decade to build.
The Cost Nobody Invoices
There is a familiar shape to this story. A concentrated industry makes a rational business decision. The benefits accrue to shareholders and to well-capitalized customers. The costs are dispersed across populations with no negotiating power and no visibility in the earnings call narrative. When Dell's COO reports that DRAM prices rose 5.5x in six months, he is describing a procurement challenge for a Fortune 50 company. When the same price movement hits Transsion, the company whose Tecno, Infinix, and iTel brands account for nearly half of all phone sales on the African continent, it is a connectivity crisis for hundreds of millions of people.
I am not arguing that Samsung, SK Hynix, and Micron are doing anything illegal. They are responding to incentives. But the incentive structure itself is the problem. When three companies control a critical input to global connectivity and all three simultaneously redirect supply toward the same high-margin customer segment, the result is a de facto rationing system that prices out the world's poorest technology consumers. The question of whether this constitutes a market failure or simply a market outcome depends entirely on whether you think internet access in 2026 is a luxury or infrastructure.
Intel's CEO says relief won't come until 2028. Counterpoint says the earliest inflection point is late 2027. That is not a quarter of adjustment. That is two years during which a generation of would-be first-time smartphone owners in Africa, South Asia, and Latin America might not get online at all, or might get pushed back to feature phones while the rest of the world builds AI agents on top of the connectivity those same populations are being denied.
This is not a tech story. It is a power story. The power to allocate scarce silicon is, right now, the power to decide who participates in the digital economy and who gets locked out. That power sits with three companies, their largest customers, and the capital markets that reward 56% gross margins. Everyone else is standing in line behind the hyperscalers, waiting to find out what is left over.
