The new wartime target isn't the autonomous weapon, it's the data center that trains it

By Aurora Israel · September 10, 2026.

Boaz Golany publishes in Aurora Israel an analysis with a simple, forceful thesis: the next major military conflict will not be decided by drones, robots or autonomous weapons alone, but by who controls the data centers, semiconductor plants, communications networks and industrial capacity that make that technology possible. According to the author, those assets are becoming first-order strategic targets, both for military resilience and for offensive planning.

The argument rests on a pattern Golany traces back to World War II: Germany and the Allies bombed factories, transport networks, oil facilities and power plants to undermine the enemy's war-making capacity. The author sees the same pattern in the war in Ukraine: Russia has struck Ukraine's power grid, defense industry and logistics on a sustained basis, while Ukraine has stepped up its attacks on Russian refineries, ammunition depots and defense-industrial facilities. His conclusion: military revolutions do not eliminate strategic infrastructure, they redefine it. If AI and autonomous systems are at the center of modern warfare, the industrial, technological and digital ecosystem that sustains them takes on equal strategic importance.

Golany describes a race already under way. China, he writes, makes no secret of its ambition to lead in military AI, autonomous systems and humanoid robotics, with sustained investment in unmanned systems across every operational domain, although the author acknowledges that the real scope of Beijing's military objectives is still unclear. The United States has accelerated its own investment, its allies are pushing similar initiatives and Russia is expanding its programs. The result, as he reads it, is an arms race that no longer revolves so much around nuclear arsenals or hypersonic missiles, but around AI, autonomous systems and the industrial base needed to sustain them.

The author himself introduces caveats worth keeping in view: AI software remains vulnerable to cyberattacks, electronic warfare, deception and the unpredictability of the real battlefield; governments are reluctant to delegate lethal decisions entirely to algorithms; and international law still has no settled rules governing fully autonomous weapons. Golany does not believe these limits will halt the trend, only slow its pace. His underlying question is no longer whether autonomous military systems will become widespread, but where each nation will look for the decisive advantage when they do.

Our take: the argument brings home, from the military angle, something we at Zendoric have been pointing out from the economic and regulatory angle —that the real contest over AI is fought in infrastructure and 'plumbing', not just in the smartest model— but here with a far more uncomfortable implication. If the data center, the semiconductor plant or the telecoms node become legitimate military targets, any dual-use infrastructure —civilian and military at once, which is almost all the AI infrastructure that exists today— becomes a potential target. That is exactly what is already happening with Ukraine's power grid or Russia's refineries: there is no need to imagine the future, it is already under way with classic energy and industrial infrastructure. Extending it to chips and data centers is, on paper, a small step.

It is worth applying, however, the same skepticism we usually bring to the US–China scoreboard: Beijing declaring an ambition to lead in humanoid robotics or autonomous systems is not the same as having the demonstrated capability to deliver it in real combat. Golany himself concedes as much in passing when he acknowledges that the scope of China's military objectives is 'still unclear'. The arms race he describes is, above all, a race of investment and strategic narrative; the outcome on the battlefield will keep depending on more prosaic variables —software reliability under jamming, the semiconductor supply chain, energy resilience— than on headlines about humanoid robots.

In the long run, the very dilemma the article poses illustrates almost to the letter our underlying thesis: the same compute infrastructure now being fought over as a military asset —chips, data centers, the energy to train them— is also the material basis of the abundance AI promises, from drug discovery to early diagnosis. The risk in this transition is not that such infrastructure exists, but that it is militarized before there is governance —targeting norms, rules on lethal autonomy, treaties— capable of keeping it from becoming the fuse of a great-power conflict. Golany gets something essential right: the question is no longer whether this is going to happen, but who governs that infrastructure and under what rules, and on that front politics is several steps behind investment.