SemiAnalysis dropped a client note on August 16 that read like a plot twist in a supplier drama: Google is working with AMD on a variant of its 10th-generation...
Gartner published a prediction on Sunday that deserves more attention than it's getting: inference costs per agentic workflow will increase more than...
The number that matters in AMD's deal with OpenAI isn't six gigawatts. It's 160 million shares.
On August 1, DeepSeek's API processed 8 trillion tokens in a single day — 5 trillion from free-tier users alone.
Intel walked onto the Computex stage in June with a pitch that made a certain kind of sense: skip the HBM shortage entirely.
AMD handed OpenAI warrants for 160 million shares at a penny each. Nine months later, it wrote Anthropic a check for $5 billion.
Jay Parikh doesn't do subtle.
Spot-market H100 instances dropped below $1.20 an hour in June.
Midjourney slashed its inference bill from 2.1 million a month to 700,000 by moving from NVIDIA A100/H100 clusters to Google TPU v6e pods.
Moonshot AI's Kimi K3 weights went live at midnight UTC — 594 gigabytes of MXFP4 safetensors, the largest open-weight release in AI history.
On July 1, Meta announced Meta Compute — a cloud business unit that will sell surplus AI training and inference capacity, plus hosted access to its Llama and...
On July 1, AWS raised its EC2 Capacity Block prices for GPU instances by 20%. This was the second hike in 2026 — a 15% bump landed earlier in the year.
Three million dollars. That's the total exercise cost if both OpenAI and Meta decide to cash in every last warrant AMD handed them.
The math on GPU ownership used to be simple.
On Monday morning, three chipmakers woke up to a problem they didn't have on Friday. Intel dropped 6%.
Google just did something no major chip designer has tried at this scale: they designed two separate processors for a job everyone else handles with one.
OpenAI spent roughly thirty months building the most impressive video generation demo the industry had ever seen, then watched it bleed $15 million per day in...
Between December 2025 and March 2026, NVIDIA wrote two checks that contradict everything the company built over the last decade.
Pick up an NVIDIA B200 and trace where the roughly 6,400 manufacturing cost actually goes.