The Uplink Revolution of Urban Networks in the AI Era: Restructuring Architecture from Download to Interaction

From Download to Interaction: The Paradigm Shift in Urban Networks

For a long time, mobile network design has been centered on a core assumption: users need to download content quickly. From streaming video to web browsing, the data flow has almost always been from the network to the terminal. But AI is fundamentally overturning this model. In urban environments, AI assistants, augmented reality navigation, real-time translation, autonomous driving, and intelligent surveillance systems not only consume data—they generate data continuously. Every voice command, every frame of sensor imagery, every environmental reading needs to be uploaded from the device to the network for processing or relay. The design logic of networks is undergoing a fundamental restructuring.

According to the Wireless Infrastructure Association (WIA), AI traffic now accounts for over 4% of total wireless network traffic in the United States and is growing rapidly. GSMA Intelligence predicts that by 2040, uplink data (data sent from user devices to the network) will account for 35% of total network load. For urban network operators, this is no longer a distant hypothesis, but an imminent reality.

Uplink Performance: The Hidden Bottleneck for Urban AI Applications

Urban AI applications impose unique demands on networks. Take autonomous driving as an example: a test vehicle can generate several gigabytes of sensor data per second, which must be uploaded to edge servers with low latency and stability for processing. If the uplink experiences jitter, the vehicle’s real-time decision-making capability is directly compromised. Similarly, AR navigation applications need to compare the user’s camera feed with spatial maps in real time; upload delays can cause misalignment between virtual information and the actual scene.

Key dimensions of uplink performance include:

• Responsiveness: Applications such as real-time translation and AR/VR require millisecond-level uplink latency.
• Consistency: Video stream processing requires a stable uplink rate rather than bursty peaks.
• Cell edge performance: Users often first encounter uplink bottlenecks inside buildings or at the coverage edge, and AI applications’ demand for consistency makes edge performance a critical requirement.
• Edge-cloud synergy: Distributed AI architectures split tasks between edge nodes and the cloud; reliable uplink is the neural pathway for real-time decision-making.

In traditional networks, the uplink is often treated as an accessory to the downlink. In high-density urban scenarios, this asymmetric design has become insufficient.

Software-Defined Networking: A Flexible Lever for Urban Infrastructure

The key to addressing uplink challenges lies in the flexibility of network architecture. Samsung’s virtualized RAN (vRAN) platform demonstrates how software-defined architecture can unlock uplink potential. Unlike hardware-bound traditional base stations, vRAN runs baseband processing functions on general-purpose servers, allowing uplink characteristics (such as carrier aggregation, higher power device classes, and advanced antenna configurations) to be remotely deployed and upgraded via software, without requiring engineers to adjust hardware on site.This flexibility is especially important for urban operators: urban user density and AI application demands change dynamically over time, making it difficult to frequently iterate hardware networks. Software-defined networks, on the other hand, can automatically adjust capacity based on real-time traffic patterns—for example, temporarily boosting uplink resources along commuter routes during morning rush hour, or dynamically allocating more frequency bands for large event areas.

More crucially, the software architecture provides fertile ground for AI-driven network optimization. Uplink scheduling is no longer a fixed static strategy; instead, AI can be used to predict traffic hotspots and allocate resources in advance. Samsung’s intelligent uplink scheduler leverages the Uplink Transmit Switching technology from 3GPP Rel-16/17 to intelligently select transmission paths between FDD and TDD carriers, achieving uplink throughput improvements of up to 70%.

Practical Validation: City-Level Breakthroughs in Uplink Performance

Joint tests by Samsung and chip partners have confirmed the uplink potential of software-defined networks. The 3Tx 5-layer uplink configuration in collaboration with MediaTek achieved a total throughput of 625.99 Mbps, paving the way for urban applications such as cloud gaming and immersive experiences. The Power Class 1 validation with Qualcomm improved cell-edge uplink throughput by ten times and extended coverage by 40%, meaning users inside high-rise buildings in cities can also enjoy a stable AI experience. At the same time, the first simultaneous 2x uplink and 4x downlink carrier aggregation over 5G FDD spectrum between the two parties allows operators to flexibly enhance uplink capabilities using fragmented spectrum resources.

These technological achievements are not isolated numbers in the lab. They are gradually being translated into real-world capabilities for urban networks. In pioneering cities such as Seoul and New York, operators have already begun deploying baseband software that supports such features, laying the connectivity foundation for AI-native applications.

Future Cities: Uplink as Infrastructure

The digital transformation of cities is irreversible. When every streetlight, every bus, and every public camera becomes a data-generating node, the network uplink, like water and electricity supply, will become a core urban infrastructure. Operators need to abandon the "download-first" mindset and redesign networks to support continuous two-way interactions. Software-defined, AI-enhanced RAN platforms provide a practical path. In the next decade, the yardstick for measuring a city’s network quality may no longer be the time needed to download a movie, but the response latency of an AI assistant or the real-time precision of AR navigation. Uplink’s time has come.