Qualcomm is angling for a seat at the AI data-center table, framing its ambitions around a $40 billion addressable-market opportunity and citing Meta as a paying customer for its server-class silicon. The company's FY2025 revenue came in at $44.3B, up 13.7% year-over-year, a respectable clip but one that still reflects heavy dependence on mobile/handset revenue rather than the high-margin data-center business Qualcomm is chasing.
The contrast with Nvidia is stark. NVDA posted $215.9B in revenue — nearly 5x QCOM's top line — with 71.1% gross margins and 55.6% net margins. Qualcomm's reported net margin sits at just 12.5%, which tells the story of how far it has to travel structurally, not just competitively. The Meta design win is real signal, but one customer does not a platform make.
The bull case rests on Qualcomm's ARM-native architecture advantages in power efficiency, its existing hyperscaler relationships, and the thesis that the AI chip market is large enough to sustain multiple winners — especially if inference workloads (where QCOM's efficiency shines) grow faster than training. Meta's procurement validation gives that thesis at least one concrete anchor.
The bear case is heavy: Nvidia's CUDA moat, software ecosystem lock-in, and margin profile are nearly impossible to replicate in a single product cycle. Qualcomm has attempted data-center pivots before without sustained traction, and its current net margin of 12.5% versus Nvidia's 55.6% signals how different the businesses are today. Execution risk is high and the timeline to meaningful data-center revenue contribution is measured in years, not quarters.
The key things to watch: whether additional hyperscalers beyond Meta announce QCOM data-center deployments, how Qualcomm's data-center revenue line grows in the next two earnings prints, and whether inference-era AI spending shifts the competitive calculus away from CUDA-centric training clusters.