👀In Case You Missed It
Amazon reshapes its AI command structure, appointing longtime AWS executive Peter DeSantis to run a new organisation spanning chips, infrastructure, frontier models, and quantum, signalling tighter vertical control of the AI stack.
Google flips Gemini 3 Flash to default across Gemini and AI Search, prioritising speed, multimodal throughput, and cost efficiency as competition shifts from raw capability to scale economics.
Google and Meta quietly align on “TorchTPU”, making Google’s TPUs more PyTorch-native in a direct attempt to weaken Nvidia’s CUDA-driven software lock-in.
US senators open a formal probe into hyperscalers’ impact on electricity prices, marking the first time AI data centres are treated as a consumer-cost and national energy issue rather than just tech infrastructure.
Read more here
🧠 Cloud & Infrastructure Watch
Google launched Gemini 3 Flash as a speed/efficiency-first frontier model and made it the default across Gemini (and AI Mode in Search), while Deep Research adds auto charts/diagrams/animations.
Read MoreOpenAI is opening ChatGPT’s in-app “App Store” to developer submissions, signalling a platform shift from chatbot → OS layer.
Read MoreAmazon is reorganising leadership under Peter DeSantis across AI, custom silicon and quantum, tightening execution across the stack.
Read MoreAmazon’s Alexa+ for Ring adds conversational AI that uses video descriptions to infer who’s at the door.
Read MoreLuma shipped a new Dream Machine capability to generate video from a start + end frame, pushing tighter control for creators.
Read MoreGoogle rolled out verification markers for AI-generated content inside the Gemini app, adding provenance signals at the model layer.
Read MoreOpenAI rolled back parts of ChatGPT’s model-router system after reliability complaints, prioritising deterministic outputs.
Read MoreLinkedIn accelerated profile verification tooling globally, with identity + AI trust becoming a platform feature.
Read MoreSalesforce moved to acquire Qualified, folding agentic AI deeper into sales and marketing workflows.
Read MoreMozilla confirmed AI features are coming to Firefox but will remain fully opt-in by defaulte.
Read More
💻 Hardware & Semiconductor Moves
Micron forecast stronger AI-driven memory demand and raised spending expectations, reinforcing HBM as a core AI bottleneck.
Read MoreHDD prices are spiking as AI infrastructure demand ripples into storage supply chains.
Read MoreA US foundry claims a “true 3D chip” combining logic + memory on one die, signalling a new path to AI efficiency.
Read MoreChina revealed a photonic AI chip prototype claimed to outperform Nvidia A100-class GPUs on speed per watt.
Read MoreNvidia confirmed it is reviewing expanded H200 production capacity after export approvals.
Read MoreNebius unveiled AI Cloud 3.1 built around Nvidia Blackwell Ultra compute.
Read More
