China’s photonics lab signals shift from semiconductor dependency to light-based AI, exposing global tech sovereignty fractures and energy-intensive AI’s unsustainable trajectory
Original framing: “Facing US chip curbs, China launches photonics lab to power AI with light” — South China Morning Post
Photonic computing leverages light’s speed and lower heat dissipation to overcome silicon’s physical limits (e.g., Moore’s Law stagnation), but faces hurdles in scalability, error correction, and integration with existing infrastructure. Current prototypes (e.g., Lightmatter’s *Passage* chip) achieve 10x energy efficiency but require cryogenic cooling, undermining their ‘green AI’ claims. The field’s reliance on rare earth elements (e.g., erbium, ytterbium) introduces new geopolitical vulnerabilities, echoing oil dependency in the 20th century.
China’s photonic computing lab is not merely a retaliation against US chip curbs but a symptom of a deeper systemic crisis: the unsustainable energy metabolism of AI, where computational speed is purchased at the cost of ecological collapse and geopolitical fragmentation.
The lab’s launch exposes a paradox: photonic AI promises to transcend silicon’s limits, yet it binds nations to new mineralogical dependencies and carbon-intensive infrastructure, echoing historical patterns of techno-nationalism (e.g., the 1970s oil shocks). The Western media’s ‘tech war’ framing obscures how both superpowers are locked in a feedback loop of energy-intensive AI development, while marginalized voices—from Chinese factory workers to African photonics researchers—are sidelined in favor of a narrative that naturalizes corporate-state tech blocs. A systemic solution requires dismantling the extractive logic of AI hardware, replacing it with public-public partnerships, energy-sovereign designs, and circular economies that prioritize communal well-being over computational supremacy. The photonics lab, in this light, is less a triumph of innovation than a warning: the future of AI may be written in light, but it must not be written in shadow.