Airoha Technology and its parent company MediaTek are exhibiting jointly at Network X 2026, where they are showing a next-generation 50G-PON AI Gateway Processor alongside an agent-based AI system for diagnosing and repairing fixed broadband networks, the company said. The exhibition runs from 13 to 15 October 2026 at Messe Wien in Vienna, at Booth K35.
The processor delivers 50+50 Gbps symmetric bidirectional throughput with low power consumption, which the company puts at up to five times the transmission speed and up to 80% lower latency than XGS-PON. It combines high-speed PON, Ethernet, Wi-Fi 8 and edge AI in a single fixed broadband gateway platform. Airoha said the processor has been shortlisted for the Network X Awards 2026 category of Outstanding Multi-Gigabit Fibre Access Innovation.
The chip uses a system-on-chip architecture with 20 cores and packet-processing capability, and can handle 50 Gbps PON, Ethernet interfaces, Wi-Fi 8 and encrypted packet processing at the same time, for residential, enterprise and industrial applications. MH Shieh, executive vice president and spokesperson, said fixed broadband is moving past high-speed access toward intelligent networking, and that putting PON, Ethernet, Wi-Fi, packet processing and edge AI on one platform raises bandwidth and lowers latency while allowing the gateway to take on network diagnostics and self-healing. He said the company works with MediaTek to integrate high-speed fiber networking, Wi-Fi and 5G CPE, helping telecom operators improve service quality and operational efficiency.
The processor carries built-in TOPS-level compute and can run several mainstream AI models, so operators can place value-added services on the gateway. Those capabilities cover network anomaly detection, automated fault diagnostics and self-healing, and generative AI applications. The platform also supports post-quantum cryptography.
For home broadband problems such as weak wireless signal, misconfiguration, abnormal optical network unit status, unstable connectivity or device overload, the solution runs an LLM-based AI agent with chipset-specific knowledge on the gateway or ONU. Users describe a problem in natural language, and the agent draws on device telemetry, network status and a diagnostic knowledge base to find the root cause and then recommend or carry out safe remediation. Airoha said such diagnostics have been validated for deployment on existing fiber gateways, running on CPU and RISC-V architectures in its fixed broadband chipsets, which reduces reliance on high-TOPS endpoint computing while core routing performance is maintained.
