Capital & Industrial Strategy

Caitlin Kalinowski, OpenAI's head of robotics, resigned Saturday citing the company's agreement to deploy its AI models within the Pentagon's classified network. In her departure statement, Kalinowski specifically objected that 'surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got,' according to Politico. The resignation follows parallel tensions at Anthropic, which is locked in negotiations with the Department of Defense over safety restrictions on its Claude models, as reported by The Guardian.

The departures signal that Pentagon contracts are becoming a defining fault line in AI talent retention. Companies pursuing defence revenue must now price in the cost of losing senior technical staff who view military applications as incompatible with their values. The dispute also exposes tactical vulnerability: if the US government can compel cooperation through regulatory or procurement leverage, founder intentions about appropriate use cases become subordinate to national security demands. For investors, this creates dual risk — both talent flight and potential regulatory coercion that overrides governance structures meant to constrain AI deployment.

An Iranian Shahed 136 drone struck an Amazon Web Services datacentre in the UAE at 4:30am Sunday, marking what The Guardian describes as the first deliberate targeting of a commercial datacentre by a state military force. The attack caused a devastating fire and forced a power supply shutdown, with further damage during suppression attempts. Iran also targeted facilities in Bahrain, demonstrating its ability to reach critical infrastructure across the Gulf despite air defence systems.

The strikes fundamentally alter the economics of Gulf AI infrastructure investments. Building datacentres in the UAE and Saudi Arabia now requires pricing in missile defence systems and hardened facilities — costs that were not in initial business cases when sovereign wealth funds began competing for AI compute leadership. For hyperscalers, the attacks demonstrate that commercial datacentres are now legitimate military targets in regional conflicts, not just civilian infrastructure with implicit protection. The incident also exposes concentration risk: if state actors can successfully strike facilities housing models and training runs, companies relying on Gulf compute for sovereign AI projects face operational continuity gaps that no SLA can cover.

Google awarded CEO Sundar Pichai a $692 million pay package with most compensation tied to performance metrics for Waymo and Wing, its autonomous vehicle and drone delivery ventures, according to TechCrunch. The structure explicitly links executive compensation to AI-enabled physical world businesses rather than core search advertising or cloud infrastructure, marking a strategic pivot in how the board defines success.

The compensation structure reveals board-level conviction that Google's long-term competitive position depends on translating AI capabilities into new revenue streams outside traditional advertising. By tying the CEO's incentives to Waymo and Wing rather than Search or Cloud margins, the board is effectively declaring that protecting legacy businesses is insufficient — the company must win in emerging categories where AI enables entirely new business models. For investors, this creates clarity on capital allocation priorities: expect continued heavy investment in autonomous systems and physical AI applications even if they cannibalise resources from higher-margin software businesses.

Samsung's device chief stated that future Galaxy devices will host multiple AI models, allowing users to mix and match tools rather than relying on a single proprietary system, according to Financial Times. The strategy represents a direct challenge to Apple's vertically integrated approach and acknowledges that no single device manufacturer can build a sufficiently capable proprietary model to compete alone.

The Samsung announcement clarifies the emerging strategic divide in consumer AI: Apple is betting on tight integration of proprietary models with hardware and services, while Samsung is positioning itself as a neutral platform that aggregates third-party AI capabilities. For model developers, Samsung's approach creates distribution opportunity — if the company executes, it offers a channel to reach hundreds of millions of devices without requiring exclusive partnerships. The strategic risk for Samsung is commoditisation: if the device becomes merely a vessel for third-party models, the company loses pricing power and becomes a low-margin assembler. The approach only works if Samsung can differentiate on integration quality, speed, or user experience in ways that justify premium pricing despite using non-proprietary AI.