The Key Takeaways From Micron's Better-Than-Expected Forecast
THE SO WHAT
Micron’s upside guide reinforces that AI demand is now a structural driver for memory, not a sidecar to CPUs and GPUs. For infra buyers, the message is simple: budget volatility is shifting to memory and storage—treat DRAM and HBM planning as a strategic lever, not a commodity line item.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechEuropean Space Agency signs up CGI and OVHcloud to build storage for the 500PB of Earth observation data it expects to have by 2035
ESA planning for 500 PB of EO data by 2035 is a reminder that geospatial and climate stacks are becoming petabyte-native. If your product touches EO or sensor data, assume the bottleneck moves from storage to indexing, retrieval, and model-ready preprocessing at this scale.
Deep & Emerging TechThis AI Can Estimate How Long You’ve Been Dead. How It Works Is the Weird Part
AI that predicts time of death from microbial “weather” on cadavers shows how far forensic and bio-surveillance models can reach into the physical world. Health, insurance, and public-sector operators should assume similar microbial and environmental signatures will be used for attribution and risk scoring, not just in labs.
Deep & Emerging TechThe investor behind Neko Health’s $700M bet on scanning your whole body
Preventative health is shifting from annual checkups to continuous, high-resolution body telemetry — $700M into Neko plus moves from Midjourney and Function Health is a bet that imaging and longitudinal data become the new primary care surface. If you build in health or wellness, assume your user shows up with a personal “digital twin” within a few years and design around interpretation, routing, and liability, not data capture.
Deep & Emerging TechMicron Forecast Tops Estimates After Demand Outstrips Supply
Micron’s upside on AI-fueled memory demand is a reminder that the bottleneck isn’t just GPUs — HBM and DRAM supply will dictate how quickly you can scale dense inference and training. If your roadmap assumes cheap, abundant memory, revisit cost models and timing for high-context or multi-modal workloads over the next 12–24 months.