For decades, Silicon Valley believed war existed at a safe distance. Operation Epic Fury shattered that illusion. Launched against Iran on February 28th, this campaign thrust American technology companies directly into active warfare – not as suppliers, but as integral participants, and now, deliberate targets.
Admiral Brad Cooper, CENTCOM Commander, revealed the stark reality: AI tools are dramatically accelerating the speed of military decision-making. Tasks that once consumed days are now completed in seconds, allowing forces to react faster than their adversaries. While humans retain final approval, the initial analysis, the crucial targeting, is now driven by machines.
At the core of this transformation lies Palantir’s Maven Smart System, powered by Anthropic’s Claude. Confirmed by NBC News, Palantir’s AI is actively identifying potential targets in ongoing strikes. Reports indicate AI enabled U.S. forces to engage over 1,000 targets within the first 24 hours of the operation – an unprecedented pace.
Retired Admiral Mark Montgomery highlighted the scale of this shift, stating the military now processes roughly a thousand targets daily, with a turnaround time of under four hours. This speed surpasses anything seen in previous campaigns. Even with a “supply chain risk” designation, Palantir CEO Alex Karp confirmed Claude remains operational within the targeting system.
This isn’t a temporary adaptation; it’s a fundamental change in military doctrine. The traditional Observe-Orient-Decide-Act loop has been streamlined to two steps: machines observe and decide, while commanders orient and act. Stripped of its complexity, the message is clear – the machine is now the primary thinker.
The seeds of Operation Epic Fury were sown in Ukraine. The integration of Palantir’s MetaConstellation – leveraging Starlink satellites, weather data, and even civilian smartphone uploads – dramatically accelerated Kyiv’s operational cycle, offsetting Russia’s numerical superiority. Ukraine effectively weaponized data.
Ukrainian forces retrained publicly available AI models using frontline combat data, boosting drone strike accuracy from a mere 10-20% to an astonishing 70-80%. AI-enhanced targeting, added to drone platforms for as little as $25, now accounts for 70-80% of battlefield casualties. This is the power of accessible, rapidly deployed AI.
Russia absorbed these lessons and shared them with Tehran. The Kremlin provided Iran with Shahed-136 drones, upgraded with AI navigation systems capable of defeating GPS jamming – technology initially tested in Ukraine. China, meanwhile, supplies approximately 80% of the critical technologies powering these Russian drones, a coalition largely overlooked in Washington.
Iran’s response was logical and chilling: if American military power relies on servers, then those servers must be destroyed. The IRGC-affiliated Tasnim News Agency published a target list naming Amazon, Microsoft, Google, Nvidia, IBM, Oracle, and Palantir – 29 locations across the UAE, Israel, Qatar, and Bahrain – labeling them “enemy’s technological infrastructure.”
This wasn’t mere rhetoric. Prior to the publication of the list, Iranian drone strikes had already damaged two AWS data centers in the UAE and a third in Bahrain. AWS confirmed structural damage, power disruptions, and fire suppression efforts. The financial markets reacted swiftly, with Nvidia’s stock plummeting nearly 9% in two days.
Wall Street finally recognized a truth strategists already understood: U.S. military power is as dependent on servers as it is on fighter jets. A less visible, yet potentially devastating, threat centers on helium – an essential, non-substitutable component in semiconductor manufacturing.
Qatar produces a third of the world’s helium. On March 2nd, Iranian drone strikes forced QatarEnergy to halt production, triggering a force majeure declaration two days later. Subsequent strikes caused “extensive” damage, cutting annual helium exports by 14%, with repairs expected to take years.
Spot helium prices have doubled. Industry experts predict a six-week resumption is “highly unlikely.” South Korea and Taiwan, sourcing 64.7% and 69% of their helium from Qatar and Gulf nations respectively, collectively control 36% of global chip production. No helium means no chips, and no chips mean no AI.
Anthropic, whose Claude model powers the Iranian targeting system, initially refused to allow its technology for autonomous lethal weapons or domestic surveillance. The Trump administration designated Anthropic a “supply chain risk” and attempted to revoke the contract, prompting a lawsuit. A Pentagon spokesperson stated bluntly that “America’s warfighters…will never be held hostage by unelected tech executives.”
Claude remains embedded, operating under a six-month phase-out agreement. OpenAI has offered classified network access, and Google has deployed AI agents for non-classified military applications. Crucially, no clear rules govern the permissible actions of these systems – not yet.
Senator Mark Warner, top Democrat on the Senate Intelligence Committee, acknowledged the unanswered questions surrounding the military’s use of AI in targeting. He is right to be concerned. Researchers warn of “automation bias” – the tendency for operators to defer to machine outputs, even when questionable.
The school strike in southern Iran, resulting in over 170 casualties, most of them children, serves as a tragic illustration of the potential consequences of accountability failures in AI targeting. Data centers are now legitimate wartime targets, and a gas commonly associated with party balloons has become a critical strategic chokepoint.
The world’s most powerful AI targeting system is actively engaged in combat while a legal battle over its control unfolds in Washington. Ukraine demonstrated AI’s transformative power on the battlefield. Iran is proving it can reshape everything else – supply chains, financial markets, and the very companies that built these tools.
Every American, reliant on cars, smartphones, pacemakers, or MRI machines, has a stake in the outcome. The battlefield has come to Silicon Valley, and Silicon Valley was profoundly unprepared.
