South Korean tech giants commit over $550B to ease ‘ RAMageddon’
THE SO WHAT
Samsung and SK Hynix committing $550B+ to memory fabs is a clear bet that AI-era scarcity will be in RAM and HBM, not just GPUs. If you’re planning large models or inference-heavy products, lock in long-term memory and storage assumptions now — pricing and availability are going to be strategic variables, not line items.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechDigital Realty says it plans to acquire a majority stake in three fully leased Northern Virginia data centers from Blackstone-managed funds in a $7.8B deal
Data centers are now treated like core, yield-bearing infrastructure—$7.8B for three fully leased Northern Virginia sites is a pricing signal on how the market values AI-era capacity. If your AI roadmap assumes “cheap colo later,” revisit that assumption and lock in power and space commitments earlier in planning cycles.
Cargo thieves have set their sights on data center supplies
When $1.3M of copper wire and equipment disappears from a truck yard, data centers stop being just a cyber target—your AI capacity depends on a physical supply chain that’s now attractive to organized theft. If you’re building or expanding facilities, treat cabling and critical components like high-value inventory and tighten logistics security and insurance assumptions accordingly.
Deep & Emerging TechForget Ryzen AI! AMD launches its SLOWEST processor in years, because 'not all customers can afford a new PC' — 2019 Zen+ CPU will at least be compatible with Windows 11
AMD shipping a 2019-era Zen+ CPU for budget Windows 11 laptops is a reminder that not every endpoint will be AI-accelerated anytime soon. If your product assumes on-device AI, design graceful degradation paths and consider cloud-assisted modes for the large base of low-end hardware that will persist for years.
Deep & Emerging TechSemiconductor stocks are surging while the ‘Magnificent Seven’ is struggling. This divergence of fortunes could be bad news for the market.
AI capex is now visibly accruing to the picks-and-shovels layer while the demand side that’s funding it trades down—your risk is being overexposed to buyers and underexposed to suppliers. If you’re building in the AI stack, assume your customers will be more valuation- and cash-constrained than your upstream infra vendors and price, contract, and runway accordingly.