TradingKey - Qualcomm (QCOM) unveiled its next-generation Snapdragon 8 Elite Gen 6 mobile platform at the 2026 Snapdragon Summit and introduced a higher-performance Extreme version. Built on TSMC's 2-nanometer process, the new chip will power future flagship smartphones from brands such as Xiaomi, Motorola, and ZTE, competing with Apple's A20 Pro featured in iPhones.

The focus of this upgrade goes beyond CPU and GPU performance, aiming to equip smartphones with stronger local AI processing capabilities. Qualcomm CEO Cristiano Amon stated that the mobile experience is shifting from "phone-centric" to "agent-centric," meaning users will no longer need to switch between apps frequently in the future, but can instead let AI complete tasks directly.

During the live demonstration, Qualcomm collaborated with StepFun, Wuliang, and Longsys to adapt the StepEdge-Omni 30B-MoE model for on-device execution. The model features 30 billion parameters; however, through a Mixture-of-Experts (MoE) architecture, it activates only a fraction of its modules during task execution, thereby reducing computational load and memory bandwidth requirements. Following optimizations in the inference engine, heterogeneous scheduling, and storage solutions, its operational memory requirement was reduced by over 50% compared with conventional solutions, enabling stable operation on smartphones.

The AI assistant in the demonstration handled emails, planned itineraries, synced calendars, recommended flights and hotels, and drafted replies without connecting to the cloud. Test data showed that the model achieved a prefill speed exceeding 330 tokens per second and a decoding speed exceeding 28 tokens per second. This demonstrates that large-parameter models are becoming viable for execution on smartphones, opening up new potential for low-latency, privacy-preserving, and offline AI applications.

Rick Osterloh, head of Google's Platforms & Devices division, also noted that the new chip's computing power and sensor capabilities will help Gemini understand a user's context, such as determining whether the user is inside a vehicle. Qualcomm has also added new security mechanisms to restrict AI agents from accessing contacts, emails, and other sensitive information.

On-device AI is expected to become a new growth engine for Qualcomm. Compared with relying entirely on cloud inference, running models locally reduces network latency and cloud computing costs while enhancing privacy protection. If smartphone manufacturers begin using AI agents as a key selling point for flagship products, Qualcomm has an opportunity to increase its content value per device leveraging its processors, NPUs, and connectivity technologies.

Source: TradingView

From a technical standpoint, Qualcomm's stock price closed at $198.27, breaking above its previous downward trendline and establishing support above the 20-day moving average of $175.93 and the 60-day moving average of $171.62, signaling a clear improvement in the short- to medium-term trend. The 20-day moving average rising back above the 60-day moving average further reinforces the potential for a continued rebound.

Currently, the stock price is approaching the 50% Fibonacci retracement level at $201.40, a key short-term hurdle. If it breaks out above this level on heavy volume at the close, the next target could be $215.28, followed by $235.05.

The RSI has risen to 67.57, standing above the signal line, indicating strong bullish momentum; however, it is approaching the 70 overbought threshold, suggesting potential short-term volatility. To the downside, the first level to watch is $190, with major support concentrated between $175.93 and $170.35. As long as this region holds, the bullish structure remains intact.

Find out more