Posted inThe Lede

Gulf AI projects keep building while the region’s new subsea cables sit stalled

Work on four new subsea cable systems in the Gulf has stalled with the conflict, even as construction continues on the region’s big AI projects — and Iran has now threatened the cables already in service beneath Hormuz. Few Gulf countries have a substantial alternative to the strait, Doug Madory, head of internet analysis at Infoblox, tells EnterpriseAM. A cable doesn’t need to be cut to cause damage, either: users losing access to their services counts as a disruption, “even if every link is still technically up,” he says.

Paused mid-lay: None of the cables already in service have been damaged so far. Installation is on hold for 2Africa Pearls, the Gulf extension of SeaMeWe-6, Ooredoo’s Fibre in Gulf, and the WorldLink Transit Cable, Madory tells us. With those projects delayed indefinitely, growth in the region’s IT sector could be held back for years, he warns.

IN CONTEXT- Tehran has drawn up plans to target the cables if attacks resume, two senior Iranian officials told Reuters last month, and one said Iran would also try to block repairs. At least four Iranian-linked vessels made unusual movements around submarine cable routes in the Gulf of Oman last month, the Financial Times reported, citing Windward tracking data. The manoeuvers could be coincidental, but might also be intended to demonstrate Iran’s ability to damage undersea infrastructure, naval warfare expert at the Royal United Services Institute told the FT.

Seen from above, not below: Windward confirmed to EnterpriseAM that the activity could be described as consistent with possible cable reconnaissance, while stressing that its purpose remained unconfirmed. “Windward offers visibility into the vessel movements, but not what happens beneath the surface,” the company said.

Why this matters: The Gulf’s digital economy still needs a physical connection. Banking, cloud computing, and online commerce in the Gulf run on networks that cross borders and, in places, contested waters. For governments building their economies into technology hubs, that leaves investors with an uncomfortable question: how reliable is access to the computing capacity going up across the region?

Cloudy, with delays

Some cloud workloads have already moved out of the region: Iranian drone strikes in March knocked two of three availability zones in Amazon Web Service’s (AWS) UAE region, ME-CENTRAL-1, and one zone in Bahrain’s ME-SOUTH-1, Madory adds. Amazon said at the time that affected Bahrain customers were shifting applications elsewhere. Madory expects the affected facilities eventually to return to service, but only when operators consider it safe for staff to return.

Still pouring concrete: Construction on more than USD 100 bn of major Gulf AI projects has generally continued despite the conflict, analysts told us. That makes cable delays consequential: the region is adding computing capacity while some of the connections intended to support its growth remain unfinished.

Standing room only

A rounding error, globally: Less than 1% of global international bandwidth crosses Hormuz, according to the International Cable Protection Committee (ICPC), citing TeleGeography. The industry body has explicitly cautioned against confusing the strait with the much broader Middle Eastern cable network — serious regional exposure doesn’t translate into a worldwide internet shutdown.

Roll call: Islamic Revolutionary Guard Corps (IRGC) spokesperson Hossein Mohebbi referred to seven major internet cables beneath Hormuz in his warning. Windward’s regional inventory — spanning existing infrastructure and new projects — names AAE-1, FALCON, Gulf Bridge International, Tata TGN-Gulf, and OMRAN/EPEG, alongside the 2Africa Gulf extension and the planned Fibre in Gulf and SeaMeWe-6 systems. Iran also has connections at stake, including Falcon and Gulf Bridge International, Windward notes. Damage to shared sections could therefore disrupt Iranian connectivity too, depending on where it occurred and how traffic could be rerouted.

Location, location, landing station: Iraq, Bahrain, Qatar, and Kuwait depend on Hormuz for their direct submarine routes, though not for all their connectivity: Bahrain, Qatar, and Kuwait have terrestrial links through Saudi Arabia, and Kuwait also connects through Iraq. The UAE has cable landings at Fujairah, and most Saudi international bandwidth arrives through the Red Sea coast, according to a TeleGeography regional assessment. Those routes reduce dependence on the strait, but overland capacity may not be able to absorb all the displaced traffic.

Iraq is building its case as a data detour: Baghdad recently signed agreements with UAE’s e& and Bahrain’s Alliance to carry internet capacity across Iraqi territory toward Europe. The agreement follows a similar one with Qatar’s Ooredoo in September, adding to Iraq’s push to connect Gulf networks to Europe overland.

The catch is capacity: Networks can divert traffic through other submarine cables or overland links, the ICPC says. The business question is how much capacity remains available when several connections fail together — and whether that is enough to keep services running normally. It is hard to measure directly how international traffic divides between Hormuz and alternatives, including overland routes through Iraq and Turkey, through Iran to Armenia and Azerbaijan, and satellite connections, Madory says.

Extra cables only count as backup if they avoid the same corridor. “Redundancy strengthens autonomy when the alternatives can survive the same crisis,” Scientific Project Manager at the ICI Institute Daniel Sidiqie tells us. Several cables running through one vulnerable corridor can add capacity without providing a dependable way around disruption. Connection through Oman or terrestrial routes across Saudi Arabia can diversify exposure, while the developing Iraq-Turkiye corridor offers another potential alternative, he says. Their value depends on sufficient spare capacity and tested rerouting.

Satellites can cover selected needs: Satellite links can support emergency communication, Sidiqie says, but can’t replace submarine capacity at scale. Madory similarly cautions that satellites, including Starlink, could replace only a small fraction of the capacity lost if the strait’s cables were damaged.

Buffering…

“Still connected” doesn’t necessarily mean usable: Applications tolerate different levels of latency — the time data takes to travel — and packet loss, when some of the data fails to arrive. Beyond a certain point, either can make an application unusable, Madory says, with user complaints often providing the first indication.

For banks, that can mean delayed transfers: If damaged connections prevent a bank from reaching overseas payment networks or outsourced technology providers, customers could face delayed transfers or unavailable services. Financial-sector resilience body FS-ISAC urges banks to map these dependencies, including international data transfers and outsourced operations. The impact would depend on each bank’s connections and backup arrangements.

The domino problem: Banks use incoming payments to fund outgoing ones, so an institution unable to send money can leave others waiting for funds they expected to receive. The Bank of England identifies this as a channel through which operational disruption can create liquidity pressure elsewhere in the system. Applied to a prolonged connectivity outage, the concern is a growing queue of unsettled obligations — with businesses further down the chain still needing to pay suppliers and staff.

For AI facilities, the workload matters: International customers offer Gulf computing facilities a commercial opportunity, but serving them requires reliable external connections, Sidiqie says. That does not mean every task stops when connectivity deteriorates. “The consequences of disruption vary by workload: an intact cluster with locally stored data may continue training even as overseas customers lose access to its services,” he says.

Backups need bandwidth too: If international connections deteriorate, the exposure extends to overseas users, external services, and the links that keep remote backup copies up to date. AWS’s disaster-recovery guidance says replication needs enough bandwidth to keep pace with changes in the original data. A prolonged bandwidth squeeze could leave a backup further behind, even while the primary servers remain operational.

Moving a workload takes preparation: AWS’s recovery options range from restoring backups to maintaining a second, running version of an application in another region — with different costs and recovery times. Some workloads also face data-residency requirements that constrain where they move. Critical services may need recovery capacity outside the immediate conflict zone, supported by data rules permitting replication in approved jurisdiction, Sidiqie says.

Your servers, whose control?

Sovereign, with an asterisk: Domestic data centers support local AI development and strengthen governments’ jurisdiction over data, but their operation can still rely on foreign cloud providers, and access to advanced GPUs remains subject to external supply chains and US export rules, Sidiqie says. “The strongest test of Gulf digital sovereignty is whether essential services can keep running when access to foreign technology or international connectivity is disrupted,” he adds.

Keeping the hardware local also keeps it within range. Domestic investment places valuable assets inside a regional conflict, and Gulf cloud infrastructure has already been hit, Sidiqie argues. “Sovereignty therefore depends on how effectively states can sustain these capabilities under pressure.”

Iran could dent the Gulf’s pitch without destroying its infrastructure. “Creating uncertainty over how quickly damaged systems could be restored could already undermine confidence in the region as a reliable computing hub,” Sidiqie says. Threats against the cables mean the networks linking those facilities to their markets are exposed too.

The recovery test

A cable doesn’t have to be the target to become a casualty: Around 200 submarine cable faults occur worldwide each year, mostly because of routine accidents involving anchors or fishing, according to ICPC data cited by TeleGeography. Conflict adds another route to damage: in 2024, a vessel hit by the Houthis dragged its anchor before sinking, damaging three Red Sea cables.

A repair ship has to get there and stay put: Specialist vessels need permission to enter the relevant waters and must stay stationary while fixing a fault, leaving crews exposed in a conflict zone. UAE-based e-Marine handles Gulf repairs as the region’s maintenance authority, according to TeleGeography, so safe access for its vessels is a practical limit on how fast service comes back.

The Red Sea benchmark: Three cables damaged there in 2024 took six months to repair, according to TeleGeography. After another incident damaged four cables in 2025, repairing three took five months.

Prep before the break: “Resilience also means restoring service quickly when prevention fails, which makes advance repair permissions and access to suitable vessels essential,” Sidiqie says.

Paying for Plan B

Money is already moving into connections: Microsoft says it plans to invest more than USD 400 mn in submarine and terrestrial connectivity across the Middle East by 2030, building on investments including the SeaMeWe-6 cable system. But SeaMeWe-6’s planned route through the Gulf to Bahrain then overland across Saudi has also faced delays amid Gulf hostilities, according to TeleGeography. Finding a different path is only part of the job, and they also need to be able to build and maintain it.

Connectivity is part of the commercial proposition: Security concerns are already influencing plans, with G42 considering spreading its planned data center footprint across the UAE to reduce the risk of concentrating infrastructure in one location. Our reading is that developers may need to spend more on independent connections and reserve capacity elsewhere, while customers weigh those safeguards against price and performance. Cheap power and available chips remain advantages, but the commercial proposition also depends on customers being able to reach them.

The signal

The measure of resilience is how much service survives a disruption, and how quickly full capacity can return. If several cables remained down for an extended period, Madory expects severe disruption to Gulf internet service. Look first for traffic volumes falling below expected levels, alongside longer delays as remaining routes become congested or traffic travels farther. Depending on the connectivity lost, monitoring could also show withdrawals of BGP routes — the routing announcements that tell other networks how to reach a destination — for networks that become unreachable.

Sounds familiar? We’ve been following the same problem in other industries. A route becomes unavailable, flows shift elsewhere, and alternatives have to absorb the load. Tankers and data move differently, but the resilience question is familiar: how much can the alternative or the remaining route carry, and for how long? Hormuz concentrates several kinds of dependencies in the same geography, that is why a chokepoint is called a chokepoint.