Qualcomm introduced the Snapdragon 8 Elite Extreme Gen 6, a 2nm processor capable of running 30 billion parameter AI models directly on smartphones, the company announced at its Snapdragon Summit on September 22.
The chip represents the first commercial smartphone processor built on a 2nm process node. The Extreme tier's ability to execute models of 30 billion parameters or larger on-device marks a shift in how flagship phones handle AI workloads previously delegated to cloud servers or smaller local models.
Qualcomm built the chip on TSMC's 2nm process technology. The 30 billion parameter threshold matters because it covers mixture-of-experts models large enough to deliver meaningful reasoning and multi-step reasoning tasks without server connectivity. Most on-device AI in current flagship phones relies on models between 1 billion and 8 billion parameters; the jump to 30 billion represents roughly a 4x increase in model capacity on the same silicon.
The announcement arrives as smartphone makers face competing pressures on component supply. Premium flagships already push memory allocations higher to support generative AI features; adding the silicon needed for 30 billion parameter inference without external connectivity will force hardware designers to decide between allocating more DRAM, prioritizing leading-edge process nodes, or both.
Qualcomm also introduced a standard Snapdragon 8 Elite Gen 6 variant on the same 2nm process, though the company did not disclose model size specifications for that tier. The company positions the Extreme tier for devices launching in late 2026 and 2027.
