The smartphone market suffered its worst decline on record in the September quarter, as rising memory costs pushed shipments down 22%, with worse to come, according to market researcher IDC. Memory chip prices increased by a stunning 350% over the past year, IDC says, because of competition for memory from AI industry players.
That’s cut into smartphone manufacturers’ profits and forced them to raise prices, cut back on specifications, and discard budget-focused models, says Nabila Popal, senior research director for Worldwide Devices at IDC. Chinese brands like Xiaomi Corp. and Oppo, which sell lower-priced phones with smaller margins, took the worst hit among the top mobile brands. Apple and Samsung were more insulated from the memory price increases because their phone portfolios lean toward higher-margin premium devices.
The same handful of manufacturers that make DRAM and NAND memory chips for smartphones—Samsung, SK Hynix, and Micron—also make memory needed for AI processing. Specifically, the manufacturers use layers of DRAM to make high-bandwidth memory (HBM) chips, which provide the high data-transfer rates required for AI workloads.
AI accelerators, such as Nvidia’s graphics processing units (GPUs), require large quantities of HBM to store model parameters and retain contextual information during problem-solving. HBM is physically integrated into the same package as the GPU, sitting right beside its processing silicon.
Memory manufacturers have been pivoting toward HBM because making it for the AI industry is more profitable than making smartphone memory. “The AI infrastructure and cloud service providers like Google, Meta, and Amazon that are doing these buildouts, the majority of them have secured long-term agreements [with the memory manufacturers] all the way up to like 5 years, to 2030 or 2031, pretty much guaranteeing the output will be given to them at certain price points,” Popal explains.
Memory prices are still rising, in part because the AI data center buildout is just getting underway. IDC believes worldwide spending on AI infrastructure, including servers and storage, will reach $1.21 trillion annually by 2030.
Memory manufacturers, Popal says, are building additional capacity but aren’t motivated to move very quickly because they’re benefiting from inflated memory prices.
Smartphone consumers will ultimately shoulder the costs. Already, the average selling price of smartphones has risen dramatically. “Last year it was $450,” Popal told Fast Company. “This year we’re expecting smartphone ASP to be $581.”
