Why Microsoft's Maia 200 Is the Quiet Threat to Nvidia's Inference Lock-In
Maia 200 ships with 216GB HBM3e, native FP4, and a potential Anthropic deal that would crack Nvidia's grip on frontier-model inference.
Publication archive
Buying guides, explainers, news, tutorials, and opinions on technology and automotive topics
Maia 200 ships with 216GB HBM3e, native FP4, and a potential Anthropic deal that would crack Nvidia's grip on frontier-model inference.
AMD just became the first company to ramp an HPC processor on TSMC's 2nm node, and the 256-core Venice chip is aimed squarely at the AI training rack.

Google AI Mode now runs Gemini 3.5 Flash, hit one billion monthly users, and adds background agents that act like Google Alerts on steroids.
Sony's 66.8MP A7R VI lands at $4,499 with a stacked sensor and 30 fps burst. The upgrade case from an A7R V is narrower than the spec sheet suggests.
Record quarter, a 25× dividend hike, and an $80 billion buyback land alongside a China Data Center number that has collapsed to zero.

Gemini Spark runs on Google's servers, not your phone, and ships with a warning that it may share data or buy things without asking first.
A trojanized Nx Console extension sat live on the VS Code Marketplace for under 20 minutes - long enough to exfiltrate 3,800 GitHub-internal repositories.

Meta is recording 77,900 employees' keystrokes to train AI agents while cutting 8,000 jobs. The same workers are building their own replacements.
AMD's Gorgon Halo refresh pushes x86 unified memory to 192GB, with 160GB usable as VRAM and a $3,999 developer box landing first at Micro Center.
Google's new $100 AI Ultra plan halves what OpenAI and Anthropic charge at the top tier, but the fine print on usage limits changes the math.
SpaceX wants $75 billion at a $1.75 trillion valuation. The S-1 hides four numbers that matter more than the headline.
Tesla FSD won MotorTrend's January 2026 shootout, but it's the only Level 2 system NHTSA just benchmarked, and the only one without LiDAR redundancy.