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...
For two months, a model called "Owl Alpha" sat near the top of OpenRouter's leaderboards. First place on Hermes Agent by call volume.
Google spent the first half of 2026 shopping its TPU chips to the companies that actually matter in AI compute — the Neoclouds. CoreWeave.
Every large language model you use right now — GPT-5.6, Claude, Gemini, Llama — generates text the same way: one token, then the next, then the next.
GDDR6 spot prices tripled since autumn 2025. From roughly 2.
Qualcomm's two-move play in the last three weeks tells you everything about where the AI hardware market is headed. First, the confirmed $3.
Three million dollars. That's the total exercise cost if both OpenAI and Meta decide to cash in every last warrant AMD handed them.
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.
NVIDIA's GR00T N1 foundation model for humanoid robots trained on 780,000 synthetic episodes spanning 6,500 hours of simulated manipulation.
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.
NVIDIA's Blackwell B200 debuted at 0.11 per million tokens on SemiAnalysis's InferenceMAX benchmarks.
NVIDIA quietly shipped one of the most interesting open model architectures I've seen this year, and most of the coverage buried the lede.
AMD just crossed a threshold that matters more than any spec sheet: one million tokens per second from a single cluster, verified by MLPerf.
Most open models fight for the same throne: highest MMLU score, best SWE-Bench pass rate, flashiest reasoning demo.
Every robotics lab has the same complaint: the model architecture works, the sim environment works, but there's never enough training data.
A chip vendor hiking prices 67% overnight would normally send customers scrambling.
Most open model launches follow a predictable script: bigger parameters, higher benchmark scores, a claim about beating GPT-something on MMLU.
Amazon will spend 200 billion on capital expenditure this year. Alphabet somewhere around 175 billion.