South Korea bets $880bn to win the AI era
THE SO WHAT
An $880 billion, decade-long commitment to chips, data centers, and robots is industrial policy at AI scale — and a clear signal that compute, not just models, is the battleground. If you rely on imported silicon or cloud, start mapping how national capex waves like Korea’s could reshape pricing, supply, and where it makes sense to build.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechRocket Lab plans to acquire Iridium for $8B and combine Rocket Lab's launch services with Iridium's satellite-based communications network to rival SpaceX
A launch provider buying a 66-satellite comms network is vertical integration in space—distribution plus service in one balance sheet. If your product depends on orbital bandwidth or timing, assume fewer, more integrated constellations and negotiate long-term capacity, not one-off rides.
Deep & Emerging TechRocket Lab to take on SpaceX’s Starlink with $8 billion acquisition
Calling Iridium a “shortcut” is code for buying time-to-market—Rocket Lab is trading $8B for an instant global comms footprint. Anyone planning to build on LEO connectivity should model a world with 3–4 vertically integrated networks, not a fragmented field of resellers.
Deep & Emerging TechSamsung doubles down on 1000-layer NAND for petabyte SSDs as it sets its sights on elusive 32TB M.2 solid state drives (you remember this right?)
Roadmaps to 1000-layer NAND and 32TB M.2 drives mean storage density is about to outpace many orgs’ data governance and backup practices. If you’re planning AI or analytics infra, assume local petabyte-scale SSDs are coming and design around bandwidth, failure domains, and data lifecycle now.
Deep & Emerging TechSandboxAQ Brings Large Quantitative Models to Google Cloud Marketplace
Physics-grounded Large Quantitative Models exposed via Google Cloud turn niche quant tooling into something any analyst with a chat interface can hit. If you’re in finance, materials, or pharma, this is a cue to inventory which hard simulation problems could move from specialist bottlenecks to self-serve workflows.