Samsung signs a $200 billion chip deal with Broadcom covering memory, foundry and advanced packaging through 2030
THE SO WHAT
A $200B, five-year Samsung–Broadcom pact hardens AI chips as a long-cycle, vertically coordinated supply chain, not a spot market. If you're building on AI infra, assume fewer, larger upstream relationships and more pricing power concentrated in those networks.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechSamsung Inks $200 Billion Chip Supply with Broadcom
$200B of foundry plus networking silicon between Samsung and Broadcom is a long-duration bet that AI infra demand is not a bubble but an industrial cycle. If you’re planning large-scale AI workloads, treat chip and package availability as a geopolitical, multi-vendor procurement problem — not a simple cloud SKU choice.
Deep & Emerging Tech'The government has all but forgotten that most data centres require water to cool their systems': Industry trade board warns UK government over data center thirst for resources
Tripling UK data center capacity by 2030 without a water plan turns cooling into a permitting and political risk, not just an engineering one. If you’re siting AI infra in water-stressed regions, assume future constraints and price in alternative cooling, reclamation, or relocation now rather than after moratoria hit.
Deep & Emerging TechSources: CXMT expelled Huawei-linked SiCarrier staff from its R&D zone amid a pricing dispute, as Chinese memory makers flex their newfound clout to hike prices
Chinese memory makers using newfound pricing power — to the point of expelling Huawei-linked staff over disputes — signals that DRAM and NAND are no longer buyer’s markets in the AI boom. If your AI infra bill assumes flat or falling memory prices, revisit those models and consider multi-region, multi-vendor hedging now.
Deep & Emerging TechNvidia and SK Group unveil a $500B+ AI initiative that includes an SK Hynix partnership to secure next-gen memory supply for Nvidia and joint development of HBM
A $500B+ program tied to SK Hynix and HBM locks in memory as a strategic choke point alongside GPUs. If your roadmap assumes fungible HBM supply or stable pricing, revisit it — capacity is being pre‑allocated at nation‑state and mega‑cap scale, not left to the spot market.