When it comes to AI infrastructure, “The bubble is the correction, not the collapse,” said Khasm Labs CEO Jim Brisimitzis in a compelling RCR Wireless Reader Forum entry. With telcos taking the operational efficiency route on the way to value creation, he explained, “AI inferencing at the edge represents that inflection point — where telcos can move from connecting data to interpreting it, from being the transport layer to being the intelligence layer where proximity is the new power.” That proximity is the key to enabling telcos to become a “physical substrate of distributed AI,” with their edge presence combining with 5G’s latency and bandwidth capabilities to enhance their already geographically distributed compute fabric.

That distributed, intelligent fabric will act as a nervous system, carrying data between data centers and users. As that happens, network and data center operators are pivoting toward high-speed 800G and 1.6Tb/s optical networks. In these networks, fiber connections have to be pristine to avoid significant signal loss, data corruption, and network downtime. Some operators are experimenting with AI for defect detection, and that practice is showing potential for improving the detection accuracy of probe microscopes. Trained on very large datasets, AI models can inspect images pixel by pixel, also iteratively improving accuracy. When classical image analysis isn’t good enough, AI models self-examine textural information and identify hidden scratches, pits and contaminants. As a result, some operators are starting to layer AI capabilities on top of traditional probe microscopes’ core features.

When it comes to 800G and 1.6Tb/s optical networks, South Korea is becoming a leader, with network operators like KT and SK Telecom testing and adopting advanced optical networks as part of their aggressive 5G and data center rollouts. U.S.-based Nvidia has become central to some of the biggest buildouts, partnering with the South Korean government and major Korean industrial groups to deploy more than 260,000 GPU. Through the Ministry of Science and ICT (MSIT), Nvidia plans to install more than 50,000 next-generation GPUs across the National AI Computing Center and local cloud providers, including NHN Cloud, Kakao Corp., and Naver Cloud. Additionally, Samsung Electronics, Hundai Motor Group, and SK Group have each signed up for 50,000 GPUs from Nvidia. The deals start deep at the policy level, with Nvidia and the Korean government co-developing sovereign foundation models using Nvidia NeMo and Nemotron datasets.