The Great AI Gamble

Project Syndicate | Harold James

US President Donald Trump and Chinese President Xi Jinping are escalating a high-stakes bilateral race to achieve artificial intelligence dominance. Both nations view advanced technology—termed "superintelligence" by Trump ahead of their high-level summit—as the decisive key to future economic, political, and military superiority. Strategic rivalry forces accelerated development.

This zero-sum logic parallels historical forced modernization programs, such as Stalinist industrialization, in which systemic risks were explicitly ignored to guarantee state survival. Washington and Beijing operate under rigid competitive imperatives that treat technological deceleration as equivalent to unilateral geopolitical surrender. Consequently, resource allocation and national defense planning in both superpowers prioritize rapid algorithmic scaling and hardware deployment over risk mitigation or safety controls. Unchecked technological competition heightens global strategic instability. The political focus on securing absolute technological supremacy locks both superpowers into a high-risk trajectory, ensuring that sovereign security imperatives systematically override international regulatory cooperation.

Comment

The race for artificial intelligence dominance between Washington and Beijing is fundamentally constrained by semiconductor manufacturing infrastructure rather than algorithmic design alone. US export controls enforced by the Bureau of Industry and Security aim to choke off advanced compute access, forcing Chinese foundries like Semiconductor Manufacturing International Corporation to adapt domestic fabrication techniques. Despite restrictions, state capital heavily subsidises yield inefficiencies to maintain hardware parity with Western competitors.

This industrial strategy relies on existing inventories of deep ultraviolet photolithography equipment supplied by ASML prior to regulatory restrictions. Chinese chip designers at Huawei leverage multi-patterning techniques on legacy seven-nanometre process nodes to bypass extreme ultraviolet equipment prohibitions. Consequently, compute scaling across Chinese data centres remains locked into higher energy consumption penalties per FLOP compared to Western facilities utilising Taiwan Semiconductor Manufacturing Company's three-nanometre lines.

Strategic Question for Discussion

Which factor will dictate the upper bound of military artificial intelligence scaling over the next decade—the hardware yields of Semiconductor Manufacturing International Corporation's multi-patterned nodes or the power grid infrastructure required to energise compute clusters?

The trajectory indicates that energy availability and power delivery infrastructure will displace lithography yield rates as the primary strategic bottleneck. While multi-patterning allows Semiconductor Manufacturing International Corporation to maintain limited hardware output, the resulting energy penalties place an unsustainable load on domestic grid allocation for megawatt-scale data centres. Consequently, severe power generation constraints will cap deployment velocity before silicon availability fully grounds operational scaling.

Share your assessment in the comments below.

To read in details click here