Timeline
Nvidia announced Cosmos 3 via X, a unified physical AI model using action-as-token.
RTX Pro 6000 Blackwell price raised to $13,250, a 55% increase
Nvidia qualified HBM4 for Vera Rubin from Samsung, SK Hynix, Micron
NVFP4 achieves 1.8x training speedup over FP8 on GPT-3 175B model with no accuracy loss claimed for models up to 70B.
NVIDIA announced NVFP4 4-bit precision format for Blackwell GPUs integrated into MaxText.
Qualified HBM4 from Samsung, SK Hynix, and Micron for Vera Rubin
Disclosed three mechanical innovations for wafer-scale chip cooling: vertical power delivery, flexible interposers, direct-impingement cooling
Announced CS4 wafer-scale chip staying on 5nm due to SRAM scaling flattening.
Cerebras published benchmark results for Kimi K2.6 on CS-3, claiming 981 tokens/sec and 6.7× speedup over GPU cloud.
Claims 10x training speed over Nvidia H100 for GPT-3-scale models using WSE-3
Ecosystem
Cerebras Systems
Nvidia
Evidence (11 articles)
China's Open-Source AI Surge: How Local Models Are Redefining Global Competition
Feb 12, 2026OpenAI Unleashes Real-Time Coding Revolution with GPT-5.3-Codex-Spark
Feb 12, 2026Cerebras Challenges Nvidia Inference Monopoly with Wafer-Scale Edge
May 20, 2026Cerebras Claims Performance Parity With Nvidia H100 on AI Training
Jun 13, 2026OpenAI's $100 Billion Horizon: How ChatGPT's Explosive Growth Is Reshaping the AI Industry
Feb 9, 2026Inference shift opens door for AI chip startups to challenge Nvidia
May 3, 2026Cerebras IPO Challenges GPU Scaling Orthodoxy
May 14, 2026Cerebras Shares Open at $385, 108% Above $185 IPO Price
May 14, 2026+ 3 more articles