Compute & Infrastructure

Iran launched a Shahed 136 drone strike on an Amazon Web Services datacentre in the United Arab Emirates at 4:30am Sunday morning, causing a devastating fire and forcing emergency power shutdown. The Guardian reports this is believed to be the first deliberate targeting of commercial datacentre infrastructure by a nation-state military force. The attack, which also affected facilities in Bahrain, represents Iran's escalation of asymmetric warfare tactics to target critical digital infrastructure rather than purely physical or military assets.

The incident introduces a new category of geopolitical risk for hyperscale infrastructure investment. The Gulf states have positioned themselves as emerging AI superpowers through massive datacentre buildouts — the UAE alone has committed tens of billions to become a regional compute hub — but proximity to active conflict zones now creates vulnerability that insurance models and site selection criteria did not previously account for. Defence analysts quoted in the Guardian piece suggest future datacentre projects in the region may require integrated missile defence systems, fundamentally altering capital expenditure models and operational security requirements.

Caitlin Kalinowski, who led OpenAI's robotics hardware team, resigned Saturday citing the company's agreement to deploy AI models within the Pentagon's classified network. TechCrunch and Bloomberg confirmed the departure. Kalinowski, formerly a Meta hardware executive who joined OpenAI to build its physical AI capabilities, stated per Politico that 'surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got.'

The resignation is significant because Kalinowski was building OpenAI's bridge from software to embodied AI — the robotics work that would enable models to control physical systems. Her departure on ethical grounds regarding military applications directly undermines OpenAI's ability to credibly recruit top hardware talent while simultaneously pursuing Defence Department contracts. It also establishes technical precedent: once OpenAI's models are integrated into DoD classified networks and paired with autonomous systems, the company's stated safety guidelines become subordinate to military command authority.

The Department of Defense has retained a former Uber executive to negotiate with Anthropic over restrictions the AI startup wants to impose on military use of its Claude models, Bloomberg reports. The move signals the Pentagon's determination to prevent Anthropic from establishing precedent that AI companies can dictate terms of use for defence applications. Anthropic has insisted on maintaining safety restrictions that would limit how its models can be deployed in military contexts, creating a legal and contractual standoff over who controls the acceptable use boundaries for foundation models procured by government.

This represents a fundamental test case for sovereign authority over AI capabilities. If Anthropic succeeds in maintaining meaningful restrictions, it establishes that private companies retain veto power over military AI deployment even after government procurement. If the Pentagon prevails in forcing unrestricted access, it sets precedent that defence contracts supersede corporate AI safety frameworks. The Guardian frames this as illuminating ethical fault lines between commercial AI development and warfighting applications, but the practical infrastructure question is whether the US government will tolerate dependency on models where the provider can constrain operational use.

Airbus Defence and Space is coordinating with Rheinmetall and OHB on a joint bid to build a Starlink-equivalent satellite internet service for Germany's Bundeswehr armed forces, Bloomberg reports. The consortium approach combines Airbus's aerospace integration capabilities, Rheinmetall's defence systems expertise, and OHB's satellite manufacturing capacity to create an all-European alternative to reliance on US-based commercial space communications infrastructure.

This reflects broader European anxiety about strategic dependency on American companies for military-critical communications and compute infrastructure. SpaceX's Starlink has become operationally essential for NATO forces, but recent political volatility around Elon Musk's relationships with foreign governments has accelerated European determination to build sovereign alternatives. The military satellite buildout is infrastructure-adjacent to AI compute questions — secure, high-bandwidth communications are prerequisites for distributed military AI systems, and European defence planners are unwilling to route sensitive AI model inference traffic through US-controlled networks even when allied.