AI Infrastructure Is Becoming Critical National Strategy

A hyperscale data center is surrounded by a certain kind of silence. The large ones in the UAE, which are windowless concrete warehouse-sized boxes surrounded by dust and double fencing, have a stillness that is almost agricultural. Then that silence ended in March. Several reports claim that Iranian drones attacked Amazon Web Services locations in Bahrain and the United Arab Emirates, striking actual structures rather than systems. Hyperscale cloud campuses were considered strategic targets for the first time in contemporary warfare. It’s the kind of event that, like the oil embargo of 1973, will likely be included in policy textbooks ten years from now.

Until then, the term “critical infrastructure” was primarily used in white papers to describe data centers. The category began to feel literal after that. In an article published in early April, the World Economic Forum made the exact argument that compute should not be regulated as commercial real estate but rather in terms of law, planning, and defense. The Forum rarely speaks first, but when it does, it usually indicates that the discussion has already taken place in private. It appears to be believed by investors. Defense ministries do, for sure.

It’s easier to overlook the numbers beneath this shift than it should be. According to research done for the Department of Energy by Lawrence Berkeley, US data centers used 176 terawatt-hours of electricity in 2023, or about 4.4% of the country’s total demand. That range is estimated to be between 325 and 580 TWh in 2028, or six to twelve percent of the nation’s total electricity. This is a structural shift in grid planning, according to the IEA’s Energy and AI report. It is no longer a private business decision when one industry begins to influence transmission decisions in three states simultaneously. It’s now a public one.

This is supported by the economics. According to Cushman & Wakefield, the average cost of building one megawatt of critical-load data center capacity in the US is $11.7 million. If a facility is hit, recovery from physical damage is not something that can be fixed in a quarter because commissioning takes months and supply chains for transformers, chillers, and chips are tight. According to the Uptime Institute survey, the majority of significant outages cost more than $100,000, and about one in five cost more than a million. Before geopolitics enters the picture, those are figures from peacetime.

Once the framing changed, it’s difficult to ignore how swiftly governments have changed. In late April, Liz Kendall, Britain’s tech secretary, unveiled a national AI hardware strategy to lessen reliance on a few foreign chip and compute suppliers. The message was clear despite the cautious language (sovereignty, not isolation). Microsoft has confirmed that a Saudi Arabia East data center will go online in late 2026. Saudi Arabia and the United Arab Emirates are investing heavily in sovereign cloud build-outs. The Gulf, which has long served as a test bed for energy geopolitics, is now a testing ground for digital power. This seems to be one of those infrequent times when a technology becomes a domain rather than a sector.

How all of this is governed is still genuinely unclear. The cloud industry was not designed with audit trails, required continuity planning, foreign investment screening, or even physical hardening in mind. This means that an American utility commissioner, a British minister, and a Saudi regulator are now strangely approaching the same issue from three different perspectives. The question of the coming years will be whether they merge or diverge. The question was urgent because of the drones over the Gulf. They didn’t respond.