Can MediaTek and Airoha’s AI Gateway Steady Data Centres?

MediaTek and subsidiary Airoha are debuting an AI-powered fibre gateway at Network X 2025 in Paris that promises more than 30% gains in service efficiency for telecom operators, faster recovery from faults and smoother end-user performance.

It is billed as a Customer Premises Equipment (CPE) platform, but the implications run straight into the data hall.

“The edge-to-cloud fibre gateway model enables operators to meet soaring data demands while maintaining flexibility and performance. It’s redefining how infrastructure investments translate into real-world service innovation,” says Oliver Loveless, Principal at Analysys Mason.​​​​​

Why this matters for data centres

• Backhaul stability and predictable peaks: AI-driven traffic classification and real-time optimisation reduce packet congestion across Fibre-to-the-Home (FTTH) access and Wi-Fi domains. Fewer micro-bursts and retransmits translate into steadier ingress at peering points and data centre edges, improving Service Level Agreements and capacity planning.

• Faster fault isolation, lower MTTR: Automated detection and remediation at the gateway cuts recovery time by up to 40%. For colocation and hyperscale sites, this means fewer ticket spikes, shorter incident windows and less noisy-neighbour impact on shared infrastructure.

• Edge-ready AI that complements core compute: With up to 50 TOPS of Neural Processing Unit (NPU) performance on the gateway, inference can move closer to users. That reduces needless round-trips to central clusters, freeing core resources for high-value workloads while supporting latency-sensitive services at the network edge.

• Energy and sustainability gains: Smarter packet scheduling and Wi-Fi anti-interference technologies lower retransmissions and idle churn. Less wasted work at the edge compounds into lower power draw upstream, aiding data centre efficiency targets.

What MediaTek and Airoha are shipping

The platform merges MediaTek’s Filogic 680 for Wi-Fi 7 with Airoha’s XGS-PON AN7581 for 10G-class fibre access.

Together they deliver high-throughput connectivity over Wi-Fi 7 and XGS-PON, then apply on-device AI to analyse flows and make real-time decisions that optimise Quality of Experience (QoE) and operator efficiency.

Key points for infrastructure teams:

• AI at the CPE: Built-in edge AI computes on live telemetry to classify traffic, prioritise time-critical sessions and decongest queues before they hit the aggregation layer. Operators report more than 30% improvement in user experience, with up to 20% more throughput on congested Wi-Fi links and lower latency across the access domain.

• Autonomous operations: The gateway continuously learns from usage patterns, tackling issues such as video buffering at peak times, game download stalls and interference on video calls. That reduces truck rolls and shortens fault cycles before they become data centre-visible incidents.

• Open software options: Support for major CPE stacks such as RDK-B and OpenWrt enables faster integration with operator backends, observability pipelines and AIOps toolchains used in modern facilities. Open interfaces also make it easier to surface actionable metrics into data centre Network Operations Centres.

• Scale from premises to platform: MediaTek silicon already ships in more than 2 billion connected devices worldwide. Pairing this footprint with an AI gateway creates an edge-to-cloud fabric that operators can standardise on, simplifying interoperability from the last metre to the metro edge and into core sites.

Immediate benefits for colocation providers

For colocation providers, the most immediate benefit is smoother traffic ingress and more predictable utilisation. Rate-limited peaks and fewer retransmits reduce pressure on top-of-rack buffers, load balancers and firewalls, improving performance density per rack.

For hyperscalers and cloud providers, edge inference offload at the gateway can keep content delivery and collaboration services responsive without overprovisioning central clusters.

It also lays foundations for AI microclouds, small, premises-adjacent AI workloads that burst to regional data centres only when needed, enabling premium tiers for latency-critical enterprise use cases.

For enterprise operators, observability improves as the gateway exposes richer real-time metadata.

That allows tighter coupling with capacity planners and AIOps platforms in the data centre, reducing mean time to detect and triage while informing smarter placement of workloads across core and edge.

A strategic lever for data centres

MediaTek and Airoha’s AI Fibre Gateway may live at the customer edge, but it is a strategic lever for data centre performance and efficiency.

By turning CPE into an autonomous, learning node, operators can tame last-mile volatility, protect core resources and accelerate roll-out of premium, latency-sensitive services.

The result is a more resilient edge-to-cloud continuum that benefits carriers, colocation sites and hyperscale data centres alike.