China Released a Quantum Photonic AI Chip That is 108x Faster Than NVDIA! скачать в хорошем качестве

China Released a Quantum Photonic AI Chip That is 108x Faster Than NVDIA!

China just flipped the AI hardware script. A research team working under ChipX (Chip Hub for Integrated Photonics Explorer) with Turing Quantum unveiled an optical quantum–photonics chip built on a 6-inch thin-film lithium-niobate wafer—packing 1,000+ integrated optical components into a single, industrial-grade processor. They claim up to 1,000× speedups on select AI workloads, a two-week deployment cycle instead of six months, and an on-ramp to real data centers in aerospace, biomedicine, and finance. In short: photons instead of electrons, far less heat, massive parallelism, and acceleration that targets the exact bottlenecks choking today’s GPU farms. The pilot line is already online—about 12,000 wafers/year with ~350 chips per wafer—and the project just took the Leading Technology Award at the 2025 World Internet Conference. This isn’t a lab demo. It’s a production path. Why this matters: AI power draw is exploding, data-center cooling is maxed out, and even Nvidia knows optics are the future. China’s monolithic integration—waveguides, modulators, couplers, filters on one wafer—suggests a scalable quantum-classical hybrid built to cluster, stack, and route light for near-instant signal movement. If this architecture networks to “million-qubit” quantum-inspired grids as promised, the balance of compute could shift from GPU-bound to photon-first far sooner than expected. Then the week took a darker turn. Anthropic reported what it calls the first AI-orchestrated cyber-espionage operation, allegedly tied to Chinese state-linked actors, who stepped a model through small “benign” tasks that added up to a coordinated intrusion campaign across major organizations. Whether you buy every claim or not, one takeaway is clear: the next front in the AI race is hardware + autonomy + security. If optical accelerators enter racks worldwide, the first things to break may not be GPUs—it could be our power budgets and our threat models. In this video, we break down how the photonic chip actually works, why thin-film lithium-niobate is such a big deal, where it beats GPUs (and where it doesn’t), how fast this could scale in real data centers, and what the Anthropic story tells us about the risks riding alongside the hardware revolution. And if you want the real story behind the world’s fastest-moving AI breakthroughs, make sure to like and subscribe to Evolving AI for daily coverage.