Chinese AI developers published a safety result alongside only 9 of the 857 models they released between 2021 and September 15, 2026, according to a SemiAnalysis report titled "Beijing Will Not Pace t…
CoreWeave Inc. announced CoreWeave Forge, a development platform that connects training, inference and evaluation so production learnings feed back into models and agents, according to chief marketing…
OpenAI's Codex team made Auto-review, the second agent that approves Codex's risky actions, free for ChatGPT sign-ins and no longer drawing usage from a user's plan, as day two of a 28-day pledge post…
Reflection announced Beam, a 501B-parameter mixture-of-experts model with 23B active parameters trained from scratch on 23.8T tokens, with Apache 2.0 weights and a technical report due this month. Sep…
NVIDIA says its AI factories are engineered to maximize return on investment across three dimensions — earning capacity, useful life and demand — citing SemiAnalysis AgentX data that NVIDIA Vera Rubin…
Inferact, a startup whose team includes vLLM maintainers, released open-source "megakernels" for Google's TPU v7 (Ironwood) that delivered 709 tokens per second for a single user on Moonshot AI's Kimi…
SemiAnalysis's STEEL teardown lab found that Intel's Panther Lake, the first commercial chip to use backside power delivery (PowerVia) and Intel's first gate-all-around RibbonFET transistors on the 18…
TileRT powered GLM-5.3 to 469 tok/s of single-user generation throughput on 8× AMD Instinct MI355X GPUs, taking the top spot on the InferenceX AgentX leaderboard and beating the second-place NVIDIA GB…
CoreWeave earned SemiAnalysis' Platinum ClusterMAX rating for the third consecutive evaluation cycle in the ClusterMAX 3.0 report released September 23, 2026, making it the only provider to hold Plati…
Two independent investigations found that ZCode, a GLM-based coding agent, has been quietly uploading users' git history and full workspace snapshots to remote servers. The findings prompted a warning…
SemiAnalysis, the semiconductor and AI infrastructure newsletter founded by Dylan Patel in 2020, is projected to surpass $100 million in revenue this year, with the vast majority of the world's larges…
SemiAnalysis reported that DeepSeek's Engram architecture, which extends standard token embeddings with learned multi-token lookups, can lower required HBM capacity for models at the same quality by o…
AI Infra Summit in Santa Clara drew 8,000 attendees on day one, up from 3,500 last year, with organizers framing the event around the message that "the age of inference is here, and it's time to conne…
SemiAnalysis estimates that only about 2.3 GW of planned US datacenter capacity is genuinely delayed by local moratoriums and New York's executive order, despite four states acting in under two months…
SemiAnalysis reported that Nvidia's Vera Rubin NVL72 platform delivered up to 7x better token throughput per megawatt than Blackwell on its AgentX agentic inference benchmark, even on early pre-releas…
Nvidia's next-generation Rubin Ultra accelerator will ship with 192GB of HBM per GPU, down from 288GB in standard Rubin and B300, according to SemiAnalysis, which first reported the change. SemiAnalys…
Google Cloud CEO Thomas Kurian said at Goldman Sachs' Communacopia + Tech Conference that the company is signing up new customers twice as fast year-over-year, with deals over $100 million also doubli…
SemiAnalysis benchmarks show Google's TPUv7 Ironwood delivers up to 50% better performance per dollar than NVIDIA's Blackwell Ultra in inference tests, marking one of the first independent verificatio…
By the end of 2028, two frontier AI labs are on track to control most of the world's usable compute, according to Dylan Patel, founder of SemiAnalysis, who calculated that their compute grows 3x annua…
SemiAnalysis published the first third-party inference results for Google's TPUv7 Ironwood, showing up to 50% better performance per dollar than NVIDIA's B200/B300 in apples-to-apples comparisons. The…