📰 Semi Analysis

20GW sits inside a restricted local boundary, 1,525MW actually slips, 2.3GW nationwide including New York

Jensen Sandbagging Performance Again, 2x more Annual Profit Per GigaWatt, The More you Buy, The More you Earn, AgentX, InferenceX, Extreme Co-Design

Robot Models, Silicon Efficiency, Jetson Thor vs. B300 TCO, Deployments, The Network Wall

Same Bandwidth, Fewer Dies: How 4-hi HBM Cuts Inference Costs and Makes Scarce DRAM Go Further

The $11T AI Buildout, Nvidia’s Backstop Economics, and the Limits of Nvidia’s Balance Sheet

Dumb Science Experiments vs. Money Printing Machines

For most of its short history, AI lived behind a screen.

InferenceX, Up to 50% Better Performance per Dollar, Rapid Externalization of TPU stack, Growing Customer Base, Ironwood, TPUv8i, Reducing CUDA Moat

Korea hosts a Squid Games, National AI Tournament, the best non-Chinese open source model then gets eliminated, why Nvidia needs open source, implications Hynix

OpenAI vs HuggingFace, Container Escapes, Kernel Bypass, Network Policies, Security Keys, Multi-tenant Grafana, and a ClusterMAX 3.0 Preview

OpenAI’s self-designed ASIC compared with Rubin, Jalapeño’s TCO, throughput per MW, and spicy deets

$3 Million USD dataset open sourced, 1 Mil+ Context Length, Multiturn, Sub Agents 95%+ KVCache HitRate, GB300 NVL72, MI355, B200

Comparing open vs. closed models across the eras of frontier models

Double the Performance with Double the Power

How America's greatest grid wasted billions and is putting ratepayers at risk by using bad models, no not those models.

Can TileRT software on NVIDIA GPU compete with Cerebras, Groq LPU, SambaNova? Batch Size 1, Disaggregated engine, high throughput engine Prefill, high interacti

Inference at 100B/GW/year, SpaceX's stellar pace, Microsoft's 10GW 2026 Awakening, Azure Can Grow Tiple-Digits

or why DeepMind's long term failure is GCP's short term gain

Kimi K3’s architecture: compressed memory, attention across depth, latent expert routing, and the inference performance

The Labor Problem and Modularization to the Rescue

Agentic Kernel Generation, Improvement in Software Quality, Unstable Internal Development Clusters, Helios MI455X Production Ramp Hell, Up to 105% Discounts fro

3 bit Rubin LUT Based Tensor Core, SM140 Feynman, Rack Scale, Perf Per MegaWatt, Perf Per Dollar, Software Improvements, Public Rubin Software, PyTorch, vLLM, O

Meta Infrastructure has become bloated, with middle managers expending resources on over-engineered technology solutions that lose sight of broader organization

A top tier RL environment startup spawns out of thin air, the most aggressive compute ramp we've ever seen, 2000km+ scale-across, and some advice for Google Dee