Huawei Reframes Optical Transport Around AI Data Centres and Puts an 800G to 2T Floor Under It

At NetworkX 2026 in Vienna, Huawei moved its optical transport story away from the consumer broadband narrative and anchored it to AI data centre interconnection, private lines for enterprise computing, and integrated computing-plus-network services. Gavin Gu, president of Huawei’s Optical Transmission Network Domain, presented what the company calls the Next Generation OTN solution, built around a six-capability target network branded LIGHTS, covering Low Latency, Intelligent Operation, Green Foundation, High Reliability, Tbps-level Capacity and Security Assurance.

The reframe matters because Huawei is asking operators to stop planning optical networks around administrative divisions and start planning them around AI data centres, or AIDCs. That is a topology change, a sales-model change and a billing-change all at once. Huawei argues operators can move from selling bandwidth to selling differentiated services priced on latency, reliability and security dimensions, dropping the single-megabit-per-second line. Computing-power private lines, for instance, would be quoted on bandwidth, latency, reliability and security, moving past the single bandwidth dimension.

The DCI side of Next Generation OTN delivers 800G to 2T ultra-high-speed capacity and what Huawei calls an OTN capacity pool that pre-deploys capacity in multiple directions using minimal resources. The DCA side covers end-to-end OXC for meshed one-hop-direct architecture, millisecond-level WSON for availability, fgOTN for flexible bandwidth from 10 Mbps to 1 Gbps, and an integrated QKD solution for quantum encryption. Huawei claims the hardware and software upgrades together deliver an end-to-end zero-outage network.

Huawei pointed to a commercial reference in Hubei, China, where operators are combining network and computing advantages to launch integrated services across education, healthcare and finance. It did not name the operators, disclose contract values or give site counts. The Next Generation OTN solution itself is a product architecture disclosure, not a commercial deployment announcement, and Huawei did not commit to a general availability date beyond indicating it is being positioned for the AI era build-out.

The AI-driven optical transport market Huawei is addressing has measurable weight behind it. Dell’Oro Group reported worldwide telecom capex rose about 5 percent in the first half of 2026, with optical transport demand tied to data centre interconnect. Separately, the firm reported the data centre physical infrastructure market grew 27 percent year-over-year to $14 billion in the second quarter of 2026, and flagged permitting delays as the principal near-term risk to the AI data centre buildout. Those two figures put Huawei’s AIDC-centric repositioning into context: the spending is happening, and the optical layer is where it lands first when AI clusters need to talk to each other.

Supplier’s note: We carry Huawei OptiX optical transport platforms, including the OSN 9800 and OSN 1800 series, as regular stock items for carrier and enterprise transport rebuilds. The 800G and 2T line cards Huawei positions for DCI map to the same chassis families we already quote against legacy 100G and 200G coherent deployments, and we can cross-reference drop-in replacements across Nokia 1830 PSS and Ciena 6500 footprints when operators are being asked to mix vendors on a shared fibre plant. Lead times on the newer coherent modules stretch to 14 to 18 weeks from Huawei’s European distribution hub, and we have seen allocation tighten on the 800G modules specifically since the NetworkX announcements. For DCA scenarios, the fgOTN flexible-bandwidth elements and the integrated QKD option are not yet stocked at distribution level, and any project quoting them should expect a special-order window and a longer lead time. Our customers in the Middle East and Africa running mixed Huawei and Nokia optical estates typically ask us to hold buffer stock on the 100G and 200G modules so that a single failed line card does not take down a route while a replacement ships from Shenzhen.