Analysis
Xiaomi posted its third consecutive quarterly profit decline, with adjusted net income plunging 43% to 6.22 billion yuan ($922 million) in the June quarter and revenue falling 6.1%, Bloomberg reported. The company attributed the decline to a persistent global memory shortage that has raised component costs across the smartphone industry faster than manufacturers can pass those costs on to consumers.
The AI buildout is now eating consumer electronics' margins
The memory shortage behind Xiaomi's numbers isn't a smartphone-industry-specific supply problem -- it's substantially a demand problem created by AI data centers. Hyperscalers and AI infrastructure operators are buying DRAM and NAND flash at a scale and price sensitivity that smartphone makers, running on much thinner per-unit margins, can't match in the same procurement queue. When a data-center buyer will pay a premium for guaranteed memory supply to keep a GPU cluster fed, that memory doesn't go to a phone factory instead, and the price of what's left rises for everyone still competing for it.
“## Three straight quarters is a trend, not a blip A single bad quarter is noise.”
This is the same underlying commodity-supply dynamic Pulse has tracked on the chip side more broadly -- semiconductor companies collectively backing over $250 billion in startup rounds this year, Nvidia's balance sheet financing its own customer base -- but Xiaomi's earnings show the second-order effect landing somewhere AI-industry coverage rarely looks: the cost structure of a mass-market consumer electronics company with no direct AI data-center exposure of its own.
Three straight quarters is a trend, not a blip
A single bad quarter is noise. Three consecutive quarters of profit decline, tied consistently to the same root cause, is a trend Xiaomi's own management is naming explicitly rather than attributing to one-off factors. That consistency matters for how investors should read it: this isn't a company execution problem that better inventory management or a new product cycle fixes quickly -- it's an external input-cost shock that persists as long as AI infrastructure demand for memory outpaces new fab capacity coming online.
The read for anyone tracking AI infrastructure spending from the venture or public-markets side: the AI buildout's costs aren't contained within the AI industry. Memory supply is a shared, finite input, and its price is now being set substantially by data-center demand rather than by consumer-electronics demand alone. Any hardware startup with a bill of materials that includes DRAM or NAND -- laptops, IoT devices, automotive electronics -- is exposed to the same input-cost pressure Xiaomi just reported, whether or not that startup has anything to do with AI itself.
