Qualcomm posted $10.6B in Q2 2026 revenue, down 3% year-over-year, with handset-facing revenue under pressure as OEMs draw down channel inventory and trim build plans. The cause is not demand for phones -- it is memory. HBM and LPDDR pricing has reached levels that are making bill-of-materials math difficult for flagship device programs, and some OEMs are explicitly carrying forward last year's chipset rather than absorbing a cost increase. That is a direct consequence of the AI-driven memory crunch Evertiq's Aasholm flagged in a separate analysis this week.
The more significant disclosure on the call was CEO Cristiano Amon confirming Qualcomm will ship data center chips to a large hyperscaler within calendar 2026. He did not name the customer but said details will come at Qualcomm's June investor day. Qualcomm has been developing data center silicon for over a year; the Oryon CPU core that shipped in Snapdragon X showed that Qualcomm can build competitive high-performance cores. The hyperscaler deployment, if it materializes, is proof that the company's server ambitions are not vaporware.
The OpenAI smartphone chip collaboration, reported separately, is the other thread worth tracking. Qualcomm building AI-native silicon for an OpenAI-designed device is a direct play for the post-smartphone category -- agentic AI compute in a handheld form factor that is not a phone. Whether OpenAI ships hardware at scale is an open question, but Qualcomm's positioning makes sense: its power efficiency and on-device inference work are exactly what a tethered-from-cloud AI device needs.
The Q3 guidance miss ($9.2-10B versus $10.19B expected) is the near-term overhang, and it is almost entirely a handset inventory correction story. The data center chip reveal is what changed sentiment and sent shares up 16% after hours. For hardware engineers: memory is expensive, OEMs are being cautious, and the implications run up and down the BOM.