SK Hynix and SanDisk reveal first specs for High Bandwidth Flash — but don't get too excited, you probably won't be able to buy it anytime soon
THE SO WHAT
512 GB stacks at 3 TB/s show where memory is headed—toward flash that behaves more like HBM for AI and high-throughput workloads. For now this is roadmap, not capacity you can book, but it should already inform how you think about future bottlenecks and data locality in your architectures.
READ THE SOURCE
MORE FROM THE WIRE
Here's how SpaceX imagines its massive Terafab chip factory will look
A planned $16.8B first phase for SpaceX’s Terafab chip plant underscores how AI and space are pulling semiconductors into mega-project territory usually reserved for energy and heavy industry. If you rely on advanced chips, start mapping exposure not just to vendors but to their upstream fab footprints — supply resilience is becoming a site-selection and permitting question, not just a pricing one.
Deep & Emerging TechFlock Highlighted Police Departments Using Its Tech. Now 4 Face Allegations of Misuse
When your flagship customer stories become case studies in misuse, the real product is your governance model. Any vendor selling surveillance or high-stakes analytics should assume marketing materials will be discovery fodder — build auditable controls, clear policies, and kill switches before scale.
Deep & Emerging TechICE’s DNA Collection Increases, SpaceX’s Rocket Crashes Into the Moon, and the AI Backlash Grows
Mass DNA collection of non-criminals — including children — going into an FBI database is a reminder that data infrastructure often outpaces consent and oversight. For any organization building identity, biometrics, or AI systems, assume the political and legal climate around data rights is about to get much less forgiving.
Deep & Emerging TechMicron’s stock claws back to buck the memory-chip selloff
Micron shrugging off a broader memory selloff suggests investors are starting to separate generic DRAM cycles from AI-tied demand. If you’re planning AI infra purchases, don’t expect memory pricing to behave like a classic downcycle — AI workloads are rewriting the floor.