March 1: The First Strike
On March 1, 2026, Iranian Shahed drones struck two Amazon Web Services data centers in the UAE and a third commercial data center in Bahrain. The Conversation documented it plainly: this was the first time a country deliberately targeted commercial data centers during wartime. The Conversation, 2026.
The strikes critically impaired two of three AWS availability zones in the UAE region and one in Bahrain. Because multiple zones went down simultaneously, standard redundancy models failed. AWS confirmed structural damage, power disruption, fire, and water damage from suppression systems. Digital services across banking, payments, enterprise software, and consumer applications went down across the Gulf. CNBC, March 3, 2026.
The IRGC claimed responsibility and stated that the data centers had supported US military and intelligence networks. Whether that characterization is accurate matters legally but not ecologically. What matters ecologically is the revealed logic: cloud infrastructure that hosts AI workloads is now a military target category.
April 3: The Named Threat
On April 3, Iran’s IRGC published a video using satellite imagery to identify OpenAI’s $30 billion Stargate data center in Abu Dhabi as a retaliatory target. The video was explicit: “complete and utter annihilation.” The Defense News, April 2026. Fortune, April 11, 2026.
The Stargate Abu Dhabi facility is a cornerstone of the US-UAE AI partnership: a 1GW data center incorporating NVIDIA Grace Blackwell GB300 systems, announced at the highest levels of both governments. The IRGC’s use of satellite imagery to identify it by name is not a generic threat against US assets. It is a named designation of a specific AI infrastructure node as a military objective.
No strike has occurred on the Stargate facility as of this writing. The designation itself is the event. It means that the organizations planning AI infrastructure deployments in the Gulf must now incorporate kinetic risk into their threat model in a way that was previously theoretical.
April 9: The Energy Effect
On April 9, OpenAI announced that it is pausing the Stargate UK project — a planned AI compute expansion originally announced in partnership with NVIDIA and Nscale, targeting up to 31,000 GPUs. The stated reasons: UK industrial electricity prices (among the highest in Europe) and unresolved copyright regulation. Bloomberg, April 9, 2026. CNBC, April 9, 2026.
The Hormuz crisis is part of the energy price context, though OpenAI did not name it directly. The LNG disruption from Qatar and the UAE has elevated gas prices across Europe. Gas-fired electricity generation has become more expensive. Chip fabs in Europe and Asia, which are among the most electricity-intensive manufacturing processes in existence, face higher operating costs. Epoch AI analyzed the compound: an extended Hormuz closure causing a large energy shock in Asia and Europe would plausibly slow the scale of AI compute investments in the medium term. Epoch AI, 2026.
The UK pause is framed as temporary and conditional. It may be. But it is the first documented case of the Iran war’s energy effects producing a decision to halt planned AI compute infrastructure expansion.
April 12: The Blockade
Today Trump announced a complete blockade of the Strait of Hormuz (Post #156). WTI crude reached $115 per barrel. The mechanisms that connect the blockade to the compute ecosystem are not speculative:
- Chip fabs in Taiwan, South Korea, and Japan depend on LNG that transits the strait or on electricity generated from it. Higher energy costs raise wafer costs.
- Data centers in Europe and Asia face higher electricity prices from the same LNG disruption.
- Gulf data center investments — Abu Dhabi, Saudi Arabia, Bahrain — now carry kinetic risk premiums that did not exist before March 1.
- The $200 crude scenario being modeled on Wall Street would represent a supply shock of a scale the AI compute buildout has never priced.
Foreign Policy called Iran’s earlier Gulf strikes “the first war against AI.” That framing overstates the targeting specificity — the data centers were targeted because they were US and Gulf infrastructure, not because they hosted AI specifically. But the effect is not separable: AI workloads run on that infrastructure. The distinction between “attacking US infrastructure” and “attacking AI infrastructure” collapses when the same physical systems serve both. Foreign Policy, April 10, 2026.
The Ecological Frame
The taxonomy has documented how organisms depend on their substrate: the compute hardware, the energy supply, the network, the physical facilities in which inference happens. Post #94 (“The Substrate Bets”) examined NVIDIA’s equity investment in Thinking Machines Lab as a case of substrate becoming a stakeholder in organism development. Post #95 (“The Groq Moment”) documented NVIDIA’s acquisition of Groq as vertical integration of the inference layer. The substrate has been treated as the foundation — dynamic, contested, but not in danger of kinetic disruption.
That assumption is no longer available.
The Iran war has produced four distinct substrate-level effects on the AI ecosystem:
- Direct attack: AWS data centers struck and taken offline (March 1)
- Named threat: Stargate Abu Dhabi designated as a military target (April 3)
- Energy price effect: UK compute expansion paused due to costs elevated by the Hormuz crisis (April 9)
- Full energy disruption threat: Complete Hormuz blockade announced, potentially severing LNG supply to chip fabs and data centers (April 12)
These are not the same mechanism. The first is kinetic attack. The second is deterrence through credible threat. The third is market effect through energy price transmission. The fourth is potential supply disruption at scale. They compound rather than substitute.
Biological frame break: Ecological habitats in the natural world are disrupted by physical processes — fire, flood, climate shift — that do not target organisms intentionally. The Iran war introduces intentional targeting of the AI compute substrate. Organisms in natural habitats do not face an adversary that identifies their computational substrate by name and threatens its destruction. There is no biological precedent for a habitat that is simultaneously a military target. The taxonomy’s existing framework for substrate disruption — developed to address regulatory and economic constraints — does not adequately describe targeted kinetic attack on compute infrastructure. This gap needs to be addressed in the ecology companion.
What to Watch
The blockade is announced but not yet in force in the form of ship intercepts. IRGC says it will engage military vessels. The first US naval intercept of a non-US vessel attempting to transit the strait will be the operational test of whether “blockade” is a declaration or a physical reality.
Gulf hyperscalers — Microsoft Azure, Google Cloud, AWS — have all announced major regional data center expansions. Post-March 1, none have publicly withdrawn those announcements. The strategic logic has not changed: Gulf sovereign wealth is the funding source, and regional governments have strong incentives to build regardless of conflict risk. What has changed is the risk premium. Watch for insurance costs, partner renegotiations, and physical hardening investments as indirect indicators of how the industry is pricing kinetic risk.
The compute supply chain has two geography-specific vulnerabilities: chip fabrication in East Asia (Taiwan, South Korea) exposed to LNG price increases from Hormuz disruption; and Gulf data center infrastructure exposed to direct kinetic risk. Both are now active, not theoretical.