Samsung and Apple Are Winning the Smartphone Market — But AI Memory Is the Real Reason Why

The AI data-center supercycle is quietly hollowing out the smartphone mid-tier — and handing the on-device AI battlefield to exactly two companies.

What Happened

Counterpoint Research’s preliminary Q2 2026 global smartphone data shows a market splitting cleanly in two. Samsung climbed from roughly 20% to 24% share year-over-year. Apple moved from ~17% to ~20%. Every other major name in the top five — Xiaomi, Oppo, Vivo — slipped between one and two points. The long tail of “Others” compressed from ~28% to ~26%. An independent read from Omdia confirms the direction: Samsung at ~22% and Apple at ~20%, both up roughly two to four points on a market that contracted approximately 4% year-over-year.

This is not a normal share-shift cycle. The headline number that explains the entire chart is a supply input: memory. Counterpoint has flagged 2026 as on track for the worst annual smartphone unit decline on record, driven by a memory crisis that has nothing to do with consumer demand and everything to do with what AI data centers are consuming upstream. The same HBM and high-bandwidth DRAM that Nvidia’s GPU clusters are devouring is made on overlapping process nodes with the LPDRAM that goes into phones. When data-center AI capex spends at historic velocity, it reprices the entire memory market — and the phone supply chain sits downstream of the same wafers.

The result is a bill-of-materials squeeze that falls unevenly. Apple and Samsung — one the world’s most powerful hardware buyer, the other a vertically integrated memory maker — can absorb higher component costs or lock supply through long-term contracts. Xiaomi, Oppo, and Vivo compete primarily on price-to-spec ratio in the mid-tier. When memory gets expensive, their core value proposition erodes: they either shrink margins or ship lower-spec devices. Neither outcome defends market share.

The Structural Read

Zoom out and the Q2 chart is a real-time readout of how AI infrastructure spending propagates downstream through the consumer-electronics supply chain. The data center is not a separate economy from the phone in your pocket — it competes with it for the same wafer starts. When hyperscalers commit hundreds of billions in AI capex, they are not just building clouds; they are repricing every device that requires high-density memory.

Samsung is the only company in the top five that sits on both sides of this trade. Its semiconductor division benefits from the memory supercycle in exactly the proportion that its device division is shielded from it — a structural hedge no Chinese OEM can replicate. Apple’s position is different but equally durable: its supply agreements, negotiating leverage, and ability to price at a premium mean it can absorb input-cost inflation without volume sacrifice. The mid-tier has neither.

The second-order effect is where it gets strategically decisive. On-device AI — the next battleground for the entire app ecosystem — requires high-density, fast memory inside the handset. Apple’s M-series silicon roadmap is being built for exactly this: running large multimodal models locally, without a cloud round-trip. Samsung’s Galaxy AI stack follows the same logic. As memory scarcity concentrates handset volume into precisely these two companies, the distribution surface for on-device AI narrows to the same two players. A supply shock in one layer is quietly deciding who controls the layer above it.

Three Implications