Semiconductor
Custom HBM Moves the Memory Bottleneck to the Base Die
Custom HBM won't end AI memory scarcity - SK hynix's HBM4E roadmap shifts the real bottleneck to base-die co-design, qualification, and yield-limited variants.
AI
Meta Iris Chip: Not the Last Hyperscaler, the Loudest of Five
Meta's Iris AI chip enters production in September 2026, part of a 7-to-14GW compute plan. Why it targets inference, not Nvidia's training crown - and what it changes.
Semiconductor
Micron $250B US Bet: The $500M Wafer Deal That Matters More
Micron's $250 billion US investment grabbed headlines, but a $500M stake in America's only advanced 300mm raw-wafer plant is the real supply-chain move. Here's why.
Semiconductor
Nvidia H200 China Sales: A Gate Opens, With Conditions
Beijing may let Alibaba, ByteDance, and DeepSeek buy Nvidia H200 chips - under 200,000 units, training only. What the two-government gate means.
Semiconductor
A Second Chip Means a Second Codebase
You design with the tools you are allowed to own. What that leaves out is who pays to make them usable. DeepSeek's reported inference chip has no named manufacturer and no prototype, and the durable cost it implies is not…
AI
The Cheapest Token Is Not the Cheapest Task
Two frontier models shipped inside 48 hours in July, and the scoreboard is the least interesting part of it. xAI answered the pricing question with one aggressively cheap sheet. OpenAI answered the same question with a menu of three. Put…