Naver AI Factory Push Meets AMD Inference Split

Two AI infrastructure signals point to sovereign cloud scale and split inference designs.

🗄️ Naver, Nvidia and Brookfield Scale Korea's AI Factory

Decoded: Nvidia's July 24 announcement said Naver, Nvidia and Brookfield plan to expand Korea's sovereign AI factory infrastructure, taking the initial Nvidia DSX deployment at Naver's GAK Sejong data center from 55 megawatts to 200 megawatts by 2028. The release said Naver intends to move toward 1 gigawatt of AI infrastructure, Nvidia plans to invest $1 billion in Naver, and Brookfield signed a nonbinding term sheet to fund up to $9 billion. Source: https://www.globenewswire.com/news-release/2026/07/25/3333160/0/en/NAVER-NVIDIA-and-Brookfield-to-Expand-Korea-s-National-AI-Factory-Infrastructure-Buildout.html

Why it matters: The AI buildout is becoming a capital-formation story as much as a chip story. Nvidia's upside increasingly depends on financing, power access, and sovereign-cloud customers that can turn GPU roadmaps into multi-year data center projects.

🤖 AMD and Cerebras Split the Inference Stack

Decoded: AMD said on July 23 that it is working with Cerebras on a disaggregated inference design pairing AMD Helios rackscale systems with Cerebras wafer-scale engines. AMD said Helios would handle prompt processing and large-context throughput, while Cerebras would accelerate token generation, with the joint system expected to deliver up to 5x higher tokens per second per watt and to appear first through Cerebras Cloud in the second half of 2026. Source: https://newsroom.amd.com/news/aai-2026-cerebras-inference

Why it matters: Inference is splitting into specialized stages instead of staying a generic GPU workload. If buyers optimize around latency, throughput and power per token, AMD's opportunity is not only replacing Nvidia boxes; it is becoming part of mixed compute fabrics tuned for specific AI economics.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team