US urged to consider military strikes to stop China achieving AGI first

South China Morning Post

Jacob Stokes, deputy director of the Indo-Pacific Security Program at the Centre for a New American Security, has urged the United States to prepare extreme measures preventing China from achieving artificial general intelligence first. Speaking at an online event on Thursday, the former White House official proposed state-backed espionage and military strikes against Chinese data centres.

This strategic framing elevates advanced computational infrastructure to the existential threat level historically reserved for nuclear weapons capability. Stokes recommended that the Department of Defense and the National Security Agency evaluate the specific intelligence baseline needed to justify such pre-emptive actions. Policy must work backward from scientific realities. Such rigorous analytical preparations aim to provide future US national security decision-makers with actionable intelligence thresholds for disrupting foreign technical breakthroughs well before deployment. Consequently, sovereign high-performance compute clusters are rapidly transitioning from commercial assets into primary kinetic targets within Great Power doctrine.

Comment

Treating computational clusters as counterforce targets redefines dual-use technology infrastructure as strategic counter-value nodes. The proposal to target data centres mirrors the logic applied during Operation Olympic Games, where the Stuxnet worm disrupted centrifuge operations at the Natanz enrichment facility to delay nuclear breakout. In the artificial intelligence domain, physical compute hardware replaces fissile material as the critical bottleneck governing technological supremacy. Consequently, the threshold for pre-emptive disruption shifts from weapons assembly to commercial semiconductor aggregation.

Targeting large-scale compute installations requires high-fidelity intelligence mapping of power grid feeds, liquid cooling distribution systems, and specialised high-bandwidth interconnects like Nvidia NVLink clusters. Disruption of these hyper-dense installations can be achieved via targeted cyber-attacks against industrial control systems or kinetic strikes on high-voltage electrical substations serving specific facility nodes. Disabling thermal management software at critical server nodes like the Wuzhen Supercomputing Centre would render training clusters inoperable without requiring total physical destruction.

Strategic Question for Discussion

Which presents a greater operational barrier to executing Stuxnet-style counter-compute operations — the challenge of identifying air-gapped training runs within commercial data centres, or the risk of immediate kinetic retaliation against sovereign infrastructure?

The primary operational hurdle lies in real-time target identification, as verifying whether a specific high-density cluster is executing dual-use frontier model training rather than benign commercial workloads presents severe intelligence attribution limits. While kinetic escalation risks can be managed through covert cyber vector deployment, the inability to reliably verify computational workloads within facilities like the Wuzhen Supercomputing Centre increases the probability of high-stakes intelligence failure. Consequently, counter-compute operations remain heavily constrained by target discrimination rather than deliverable effect.

Share your assessment in the comments below.

To read in details click here