Connectivity

The Invisible Infrastructure of Saudi AI

Before an AI model can answer a question, the question must travel from the user’s device to a data center, be processed by the inference hardware, and return an answer — all within the latency window that human perception accepts as “instant.” For Saudi Arabia to operate as a serious AI compute hub, not just a repository of GPU hardware, the connectivity infrastructure carrying that traffic must be world-class: high-bandwidth, low-latency, redundant, and reaching the markets Saudi AI is meant to serve.

Connectivity is the sector that determines whether Saudi Arabia’s announced AI compute ambitions translate into a genuinely usable global AI platform, or remain capacity that is technically impressive but practically constrained. The $77 billion Humain commitment makes no sense without the fiber, submarine cables, and last-mile networks to move data at the volumes AI workloads require.

Saudi Arabia’s Connectivity Geography

Saudi Arabia occupies a significant position in global telecommunications geography — not by accident, but by virtue of sitting at the intersection of three major intercontinental cable corridors. The Red Sea corridor connects Europe and the Mediterranean to East Africa and the Indian subcontinent. The Gulf corridor connects the Arabian Peninsula to South Asia and Southeast Asia. Terrestrial routes through Jordan and Egypt connect Saudi Arabia to Levant and European networks.

This geographic position has made Saudi Arabia a natural landing point for submarine cables serving the Europe-India-Asia route — one of the highest-capacity international fiber corridors in the world, driven by traffic between European cloud consumers and Asian manufacturing and technology ecosystems. Saudi Arabia’s Red Sea coastline (Jeddah is the primary cable landing location) and its Gulf coastline (Dammam) give it options for diverse cable landings and route redundancy.

The strategic implication for AI: Saudi Arabia does not need to build from scratch. It needs to upgrade and expand connectivity infrastructure that already exists in nascent form. The cables are arriving; the landing stations need expansion; the terrestrial networks connecting Riyadh to the coast need capacity increases; the data centers need direct connection to international exchange points.

2Africa: The Cable That Changes the Equation

The 2Africa submarine cable system is the single most significant connectivity development for Saudi Arabia’s AI ambitions. Led by Meta (Facebook’s parent company) and built by a consortium that includes stc, Telecom Egypt, Orange, and several other carriers, 2Africa spans approximately 45,000 kilometers and connects 33 countries across Africa, the Middle East, and Europe — making it one of the longest submarine cable systems ever deployed.

For Saudi Arabia specifically, 2Africa has multiple landing points, including on the Red Sea coast. The cable’s design capacity is substantial — early specifications indicated a system capacity in the hundreds of terabits per second range, and lit capacity is expandable as traffic demand grows. In the context of AI, where a single training run can involve moving petabytes of data and where inference traffic from tens of millions of users generates sustained high-bandwidth loads, having a cable system of 2Africa’s scale landing in Saudi Arabia is foundational.

The Meta involvement is meaningful beyond the technical specifications. Meta is one of the world’s largest consumers of AI compute, operating its own massive GPU clusters for recommendation models, generative AI features, and computer vision. Meta’s participation in landing 2Africa in Saudi Arabia signals that the company views Saudi connectivity as useful for its own global network operations — an implicit endorsement of Saudi Arabia’s position in the global data flow map.

2Africa also represents a hedge against cable concentration risk. Historically, Saudi Arabia was served primarily by cables on the Europe-Asia route that also served India and had limited redundancy in the Red Sea segment. A major cable cut in the Red Sea — an event that has occurred due to shipping anchor drags and other incidents — could disrupt a significant fraction of Saudi international bandwidth. 2Africa, with its African coastal route option and multiple diverse landing points, reduces this single-point-of-failure risk.

Mobily Cable Landing Station and Infrastructure

Mobily (Etihad Etisalat) operates one of Saudi Arabia’s primary cable landing stations, and its international infrastructure portfolio is a less-discussed but operationally critical part of Saudi connectivity. Cable landing stations are the physical facilities where submarine cables come ashore, connect to terrestrial transmission networks, and interconnect with each other. They are the chokepoints of international bandwidth — a cable is only as useful as the landing station infrastructure connected to it.

Mobily’s international operations extend beyond its domestic Saudi mobile and broadband business. It has capacity agreements and ownership stakes in multiple cable systems, and its landing station operations give it direct control over a significant portion of Saudi Arabia’s international bandwidth. For AI workloads specifically, Mobily’s ability to offer low-latency, high-bandwidth connectivity between its landing stations and data center campuses in Riyadh and Jeddah is a competitive differentiator for attracting workloads that require Saudi-based AI with direct international connectivity.

stc: The National Connectivity Champion

Saudi Telecom Company (stc) is the Kingdom’s largest telecommunications operator and a significant international connectivity player. Beyond its domestic mobile, fixed, and enterprise businesses, stc has invested in international submarine cable systems, international fiber capacity agreements, and a global connectivity infrastructure that spans multiple continents.

stc’s international subsidiary — stc Solutions for enterprise and government, stc International for wholesale — operates a portfolio of cable capacity that gives Saudi Arabia a degree of domestic control over its international bandwidth infrastructure. This is strategically significant: a country that relies entirely on foreign-owned cable systems for its international connectivity is exposed to commercial and geopolitical leverage that a country with domestic cable ownership and operation can resist.

For the AI compute buildout, stc’s dual role — as a connectivity provider to data center operators and as a potential co-investor in AI infrastructure — gives it unusual strategic importance. stc’s network, touching every city and enterprise in Saudi Arabia, is the distribution layer for AI inference to Saudi end users. An AI model running in a Riyadh data center reaches a Saudi business or consumer through stc’s (or Mobily’s, or Zain’s) last-mile network.

stc has been explicit about its cloud and AI ambitions, investing in its stc Cloud platform and positioning itself as a Saudi-national alternative to hyperscale cloud providers for government and regulated-industry workloads. Its connectivity assets are both the infrastructure supporting that ambition and a revenue stream generated by selling bandwidth to competing cloud providers operating in Saudi Arabia.

The Saudi Internet Exchange (SAIX)

Internet Exchange Points (IXPs) are facilities where multiple networks interconnect and exchange traffic — routing packets between network operators without requiring them to traverse the public internet or pay transit fees to a third-party carrier. IXPs reduce latency (traffic between networks exchanging at an IXP takes a shorter path), reduce cost (peering is typically cheaper than transit), and improve redundancy.

The Saudi Internet Exchange (SAIX) in Riyadh is Saudi Arabia’s primary internet exchange facility. Its capacity and the number of networks peering at it directly affect the performance of internet services — including AI inference services — for Saudi users. A well-developed, high-capacity internet exchange with many networks connected allows a data center operator in Riyadh to reach most Saudi internet users via a short local path rather than routing traffic through Dubai or London.

SAIX’s development is consequently a meaningful indicator of Saudi Arabia’s readiness to serve AI inference at scale to domestic users. The exchange’s growth in connected networks, total capacity, and traffic volumes over the 2024–2026 period is a leading indicator for the AI inference opportunity in Saudi Arabia. A country whose internet exchange can support terabit-scale traffic interconnection between many networks is structurally better positioned to host world-class AI inference than one relying on international routing for most traffic.

5G: The Access Layer for AI Delivery

AI inference reaches end users through the last-mile network — the connection from the network edge to the user’s device. For mobile users (which in Saudi Arabia means the vast majority of internet users, given the country’s high smartphone penetration and historically lower fixed broadband adoption), that last-mile connection is a mobile network. In 2025 and beyond, that means 5G.

Saudi Arabia has been among the fastest 5G deployment countries globally, with stc, Mobily, and Zain all completing nationwide 5G rollouts ahead of many Western European markets. The combination of high 5G penetration and a youthful, tech-forward population creates genuine demand conditions for AI inference services — voice assistants, image generation, real-time translation, and enterprise AI tools — delivered to mobile devices.

The 5G-AI inference connection is not abstract. AI models like large language models have inference latency in the 100–500 millisecond range for typical queries. For that latency to be acceptable to users, the network round-trip from device to data center and back must add minimal overhead. 5G’s lower latency compared to 4G LTE, combined with edge computing deployments at 5G base stations, enables AI inference that is genuinely responsive in real-world conditions. Saudi Arabia’s 5G coverage is therefore a direct enabler of the AI inference market that data centers in the Kingdom need to justify their capital investment.

The Data Center Connectivity Requirement

A modern hyperscale AI data center is not simply a building full of GPUs. It requires dedicated, high-capacity connectivity infrastructure: diverse fiber paths from the facility to international cable landing stations, connections to the national internet exchange, and direct peering arrangements with major content networks and cloud providers.

The standard connectivity specification for a hyperscale data center includes 100 Gigabit Ethernet (100GbE) uplinks as a minimum, with 400GbE increasingly the standard for new builds, and the ability to aggregate multiple such uplinks into terabit-scale total bandwidth. A 100 MW campus with dense AI workloads might generate 10–50 terabits per second of internal traffic and require multiple terabits of external connectivity to serve its users.

Building this connectivity from scratch in a Saudi location requires: a fiber route from the data center campus to the nearest major transmission node, which may be 10–100 kilometers depending on site location; capacity agreements or dark fiber leases on that route; connections to SAIX for domestic traffic exchange; and capacity on one or more submarine cable systems for international traffic. The total capital cost for the connectivity layer of a large data center campus can run into the tens of millions of dollars — substantial, but secondary to the cost of the facility itself.

The Bandwidth Bottleneck Risk

The central strategic risk in Saudi Arabia’s AI connectivity picture is not domestic — Saudi domestic network capacity is expanding rapidly and the infrastructure for AI delivery to Saudi users is being built. The risk is international bandwidth.

Saudi Arabia’s population is approximately 36 million people, of whom perhaps 30 million are active internet users. This is a substantial domestic AI market, but it is not by itself the justification for $77 billion in AI compute. The justification requires Saudi Arabia to serve regional markets — the broader Middle East and North Africa region, South Asia, East Africa — where the connectivity flows through Saudi cables and Saudi exchange points.

Serving the MENA and South Asia AI market from Saudi Arabia requires international bandwidth at a scale that does not yet exist in the Kingdom’s infrastructure. The current trajectory — 2Africa online, stc and Mobily expanding cable capacity, SAIX growing — is the right direction. But the bandwidth that $77 billion in AI compute would generate, if fully utilized, would dwarf current Saudi international bandwidth capacity. This is the connectivity bottleneck risk: Saudi Arabia may have enough domestic power to run the GPUs, but not yet enough international bandwidth to deliver the AI outputs globally.

Closing this gap is a 3–5 year infrastructure program in its own right, running in parallel with the data center buildout. The cable systems being deployed today — 2Africa, the Peace Cable connecting Pakistan to France via Saudi Arabia, other systems under development — are sized for the demand environment of 2025–2030. Whether they are sized for a Saudi Arabia that is running 600,000+ NVIDIA Blackwell GPUs at scale is a genuine open question.

---

Tracked entities in this sector include 2Africa Cable, Mobily Cable Landing Station operations, stc International connectivity assets, Saudi Internet Exchange (SAIX), and the 5G network operators building the access layer for AI inference delivery to Saudi end users.