OpenAI jalapeño: Better than Nvidia blackwell Dylan Patel · SemiAnalysis ·Aug 25, 2026 · 24 min read · AI & Tech
AgentX - InferenceXv3: Does cuda moat hold up in agentic inferencing? Dylan Patel · SemiAnalysis ·Aug 23, 2026 · 79 min read · AI & Tech
Cerebras's next generation CS-4: Fast just got faster Dylan Patel · SemiAnalysis ·Aug 18, 2026 · 11 min read · AI & Tech
$12B of US ratepayers' money wasted on a modeling mistake and pjm wants to do it again Dylan Patel · SemiAnalysis ·Aug 16, 2026 · 42 min read · Economics
Ultra-High interactivity on Nvidia GPUs? - TileRT InferenceX Dylan Patel · SemiAnalysis ·Aug 9, 2026 · 18 min read · AI & Tech
SpaceX 10GW in 2027 – why it’s real, will drive $300B arr for SpaceX, and why Microsoft will be the… Dylan Patel · SemiAnalysis ·Aug 7, 2026 · 12 min read · AI & Tech
Kimi k3, the manos, the mythos, the legendos Dylan Patel · SemiAnalysis ·Aug 3, 2026 · 25 min read · AI & Tech
The wild wild West of lego datacenters Dylan Patel · SemiAnalysis ·Jul 29, 2026 · 52 min read · AI & Tech
Can Amd break the cuda moat? Amd advancing AI 2026 Dylan Patel · SemiAnalysis ·Jul 24, 2026 · 58 min read · AI & Tech
Vera rubin NVL72 vs GB200 NVL72? Inference tco & architecture analysis Dylan Patel · SemiAnalysis ·Jul 22, 2026 · 20 min read · AI & Tech
Meta’s infrastructure team needs a culture reset Dylan Patel · SemiAnalysis ·Jul 21, 2026 · 15 min read · AI & Tech
The future of Meta superintelligence: A 1 year progress update Dylan Patel · SemiAnalysis ·Jul 9, 2026 · 18 min read · AI & Tech
Nvidia gpu debt backstop unleashes the AI project trinity: Capital, offtake and datacenters Dylan Patel · SemiAnalysis ·Jul 6, 2026 · 23 min read · AI & Tech
Meta compute: Everyone wants to be a cloud Dylan Patel · SemiAnalysis ·Jul 2, 2026 · 10 min read · AI & Tech
EMIB-T roadmap, custom hbm, HBM4 packaging challenges, microfluidic cooling, photonic… Dylan Patel · SemiAnalysis ·Jul 2, 2026 · 31 min read · AI & Tech
TokenBudgeting: Our conversations with enterprises on token spend Dylan Patel · SemiAnalysis ·Jun 30, 2026 · 10 min read · AI & Tech
US grid constraints: Towards 40GW+ of Behind-The-Meter datacenter by 2028? Dylan Patel · SemiAnalysis ·Jun 25, 2026 · 27 min read · AI & Tech
China’s cxmt is set to challenge dram incumbents Dylan Patel · SemiAnalysis ·Jun 23, 2026 · 28 min read · AI & Tech
Stop saying half of 2026 US datacenter capacity is canceled Dylan Patel · SemiAnalysis ·Jun 18, 2026 · 26 min read · AI & Tech
Rl systems mind the gap: Matching trainer and generator throughput Dylan Patel · SemiAnalysis ·Jun 16, 2026 · 25 min read · AI & Tech
Is smic n+3’s metal pitch smaller than Intel 18A’s? Dylan Patel · SemiAnalysis ·Jun 14, 2026 · 40 min read · AI & Tech
DeepSeekV4 1.6T day 0 to day 43 performance over time - Huawei, GB300 NVL72, MI355X, b200 Dylan Patel · SemiAnalysis ·Jun 9, 2026 · 31 min read · AI & Tech
China's unitree will dominate global robotics Dylan Patel · SemiAnalysis ·Jun 8, 2026 · 27 min read · AI & Tech
To boldly go: The case for space datacenters Dylan Patel · SemiAnalysis ·Jun 3, 2026 · 46 min read · AI & Tech
AI dark output: The visible cost of invisible output Dylan Patel · SemiAnalysis ·May 29, 2026 · 22 min read · AI & Tech
Anthropic growth and bedrock mix drive Aws margins higher while peers lag Dylan Patel · SemiAnalysis ·May 27, 2026 · 16 min read · AI & Tech
Inside the 800VDC revolution – part 1 Dylan Patel · SemiAnalysis ·May 26, 2026 · 42 min read · AI & Tech
The eda primer: From rtl to silicon Dylan Patel · SemiAnalysis ·May 11, 2026 · 50 min read · AI & Tech
