2.1 The architectural revolution of smartphones

The hardware configuration of flagship models in 2024 (such as iPhone 16 Pro and Samsung Galaxy S25) shows three major trends:

2.1.1 Sensor Fusion

Upgrade of environmental perception:

• Apple iPhone 16 Pro is equipped with a lidar scanner and dual ambient light sensors, supporting spatial perception and scene recognition, but does not integrate millimeter-wave radar or quantum light sensors.

• Samsung Galaxy S25 uses a multi-spectral sensor array (similar technology to Apple AMS spectral sensor) to achieve zone color temperature adjustment and automatic white balance optimization to enhance color expression.

Breakthrough in computing power on the edge： Apple A18 Bionic Chip: The neural engine has a computing power of 45 TOPS (predicted value, based on TSMC's 3nm process and M4 chip performance), which is about 20% higher than the A17 Pro. With LPDDR6 memory (bandwidth 28.5-38.4 GB/s), it supports Llama 3-7B level model fine-tuning, but is limited by memory capacity (up to 1TB), and complete fine-tuning of Llama 3-70B requires external storage.

Qualcomm Snapdragon 8 Elite: The Snapdragon 8 Elite on the Samsung Galaxy S25 integrates the Adreno 750 GPU and the sixth-generation AI engine, with a computing power of 45 TOPS and supports the generation of Stable Diffusion XL (1 billion parameters) on the edge.

2.1.2 Enhanced security isolation

• Independent security chip (SE)

The Honor Magic4 series pioneered a dual TEE security system (HTEE+QTEE), which, together with the CC EAL5+ certified independent security chip, achieves local data encryption and biometric information isolation. Apple iPhone 16 Pro uses Secure Enclave and data isolation partitions to support full-process encryption of sensitive operations on the end (such as Face ID and health data), but has not disclosed the technical details of fully homomorphic encryption.

• Latency optimization

The local data processing delay of the independent security chip is less than 5ms (Honor official data), which is 60% lower than the delay of software encryption solutions (such as iOS Data Protection).

2.1.3 AI and DePIN integration trend

• Hardware foundation

Sensor fusion and increased computing power on the edge provide high-quality input and real-time reasoning capabilities for AI models. For example, the vivo X200 series uses the Blueprint imaging chip V3+ to implement a 20x telephoto AI algorithm and optimize multimodal data acquisition.

• Ecosystem synergy

Systems such as Huawei's HarmonyOS and Xiaomi's Hyper OS use a distributed architecture to share computing power among devices. For example, the OPPO Pantanal system supports the collaborative training of lightweight AI models between mobile phones and smart home devices.

The current flagship mobile phone hardware emphasizes breakthroughs in side computing power to cater to the potential computing power market, and the "security chip" aims to protect user data sovereignty and reduce the risk of privacy leakage, which is in line with DePIN's "distributed supply of hardware resources" concept and DePIN's "data value chain circulation" goal. These trends have created good conditions for the integration of AI and DePIN, providing support from data collection, processing capabilities to security assurance. At the same time, it also gives us more directions and thinking points, that is, how to better integrate the concept of AI and DePIN integration into mobile phone architecture design under such a trend.