e& to expand UAE international connectivity from 20 Tbps to over 500 Tbps by 2030

The headline multiple is not the story. The admission behind it is: 20 Tbps of international capacity is the choke point holding back the UAE's AI and cloud ambitions. If you run GPU capacity here, or plan to, you already know that serving inference into Europe, India and East Africa comes down to egress capacity, latency and route diversity. Sizing 500+ Tbps explicitly for AI campuses and hyperscalers moves the UAE from a market where AI workloads happen to sit into a credible serving platform for three continents. Abu Dhabi's AI build-outs and the wider data-centre expansion only pay off if this pipe exists. This project is the foundation underneath all of them.

The delivery path is already visible in etisalat's engineering trail. In February 2026, e& Carrier & Wholesale activated the 2Africa submarine cable at its SmartHub data centre in Abu Dhabi: roughly 45,000 km linking Europe, Asia, Africa and the Middle East. On land, etisalat completed a 400G Super C+L transport trial with Huawei in September 2023, stretching usable optical spectrum from 4.8 THz on the extended C-band to 12 THz, a claimed 2.5x efficiency gain that multiplies capacity on fibre already in the ground. Group-wide, e& also pledged $6 billion of infrastructure investment for 2024–2026 under the ITU Partner2Connect coalition. Others are laying cable too: du is landing PEACE through a Gulf Extension, with ready-for-service expected in H2 2026, and Africa-1, part-owned by G42, lands at Kalba. Genuine multi-carrier route diversity is arriving.

For regulated UAE entities the sovereignty framing matters as much as the bandwidth. Alzaabi's emphasis on full domestic ownership, combined with diverse routing, means data residency and cross-border data flows stop being a trade-off. Keep training data and sensitive storage in-country, serve global inference traffic over low-latency direct paths. Banks, government entities and healthcare groups gain real room to balance local compliance against regional and global reach.

Most teams will trip over one thing: treating national capacity as a procurement checkbox. Capacity is not latency, and announced capacity is not contracted capacity. A 500 Tbps national backbone does not fix single-carrier egress dependency, does not cut cloud egress billing, and does nothing on the day a subsea cut takes out your primary path. Architecture is what delivers: dual-carrier egress, health-checked failover, and a model gateway that turns a route change into a configuration edit instead of a code release. Azrty runs model and traffic routing through a single control plane with policy-based failover across distinct carriers, hosted on Kubernetes you control. A minimal dual-route egress configuration:

# dual-carrier international egress with health-checked failover
upstream intl_e { server 10.10.0.1:443; }  # e& route (2Africa / SMW5 path)
upstream intl_du { server 10.20.0.1:443; } # du route (PEACE Gulf Extension)

server { listen 443; location / { proxy_pass https://intl_e; proxy_next_upstream error timeout http_502; proxy_connect_timeout 2s; } }

In practice we put an OpenAI-compatible gateway such as FastLLM Proxy in front of self-hosted model servers and hosted providers, so routing, budgets and access control sit in one place. Wire the carrier egress paths into that layer and failover happens without touching application clients. Teams that build this before the capacity lands will be the ones selling inference out of the UAE when it does.