SemiAnalysis
semianalysis.skystack.xyz
SemiAnalysis
@semianalysis.skystack.xyz
Bridging the gap between the world's most important industry, semiconductors, and business.


This is an automated Substack Account of https://newsletter.semianalysis.com
Discover more/Create at: @skystack.xyz
AWS Trainium3 Deep Dive | A Potential Challenger Approaching • Step-Function Software & System Improvements, “Amazon Basics” GB200 NVL36x2, NL72x2/NL32x2 Scale Up Rack Architecture, Optimized Perf per TCO, Trainium4
AWS Trainium3 Deep Dive | A Potential Challenger Approaching
Step-Function Software & System Improvements, “Amazon Basics” GB200 NVL36x2, NL72x2/NL32x2 Scale Up Rack Architecture, Optimized Perf per TCO, Trainium4
newsletter.semianalysis.com
December 6, 2025 at 5:49 AM
TSMC Overseas Fabs – A Success? • Why Morris Chang Said U.S. Fabs Will Fail, Wafer Cost and Economics of Taiwan vs. U.S. Fabs, TSMC Supply Chain Details
TSMC Overseas Fabs – A Success?
Why Morris Chang Said U.S. Fabs Will Fail, Wafer Cost and Economics of Taiwan vs. U.S. Fabs, TSMC Supply Chain Details
newsletter.semianalysis.com
December 4, 2025 at 5:41 PM
TPUv7: Google Takes a Swing at the King • Potential End of the CUDA Moat?, Anthropic’s 1GW+ TPU Purchase, The more (TPU) Meta/SSI/xAI/OAI/Anthro buy the more (GPU capex) you save, Next Generation TPUv8AX and TPUv8X versus Vera Rubin
TPUv7: Google Takes a Swing at the King
Potential End of the CUDA Moat?, Anthropic’s 1GW+ TPU Purchase, The more (TPU) Meta/SSI/xAI/OAI/Anthro buy the more (GPU capex) you save, Next Generation TPUv8AX and TPUv8X versus Vera Rubin
newsletter.semianalysis.com
November 28, 2025 at 2:47 PM
Microsoft's AI Strategy Deconstructed - From Energy to Tokens • "The Big Pause", AI Tokens Factory Economics Stack, OpenAI, Neocloud Renting, GitHub Copilot, MAI and Maia
Microsoft's AI Strategy Deconstructed - From Energy to Tokens
"The Big Pause", AI Tokens Factory Economics Stack, OpenAI, Neocloud Renting, GitHub Copilot, MAI and Maia
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
ClusterMAX™ 2.0: The Industry Standard GPU Cloud Rating System • 95% Coverage By Volume, 84 Providers Rated, 209 Providers Tracked, 140+ Customers Surveyed, 46,000 Words For Your Enjoyment
ClusterMAX™ 2.0: The Industry Standard GPU Cloud Rating System
95% Coverage By Volume, 84 Providers Rated, 209 Providers Tracked, 140+ Customers Surveyed, 46,000 Words For Your Enjoyment
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
How to Kill 2 Monopolies with 1 Tool • Substrate X-Ray Lithography, a New American Foundry, $10k Logic Wafers
How to Kill 2 Monopolies with 1 Tool
Substrate X-Ray Lithography, a New American Foundry, $10k Logic Wafers
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Nanoimprint Lithography: Stop Saying It Will Replace EUV • NIL basics, why it won’t replace EUV, details of Canon’s tool, possible applications
Nanoimprint Lithography: Stop Saying It Will Replace EUV
NIL basics, why it won’t replace EUV, details of Canon’s tool, possible applications
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
InferenceMAX™: Open Source Inference Benchmarking • NVIDIA GB200 NVL72, AMD MI355X, Throughput Token per GPU, Latency Tok/s/user, Perf per Dollar, Cost per Million Tokens, Tokens per Provisioned Megawatt, DeepSeek R1 670B, GPTOSS 120B, Llama3 70B
InferenceMAX™: Open Source Inference Benchmarking
NVIDIA GB200 NVL72, AMD MI355X, Throughput Token per GPU, Latency Tok/s/user, Perf per Dollar, Cost per Million Tokens, Tokens per Provisioned Megawatt, DeepSeek R1 670B, GPTOSS 120B, Llama3 70B
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
xAI's Colossus 2 - First Gigawatt Datacenter In The World, Unique RL Methodology, Capital Raise • On Site Turbines, Mississippi Expansion, Solaris Energy, Can xAI afford it?, Middle East Funding, Tesla, Talent Exodus, API revenue, Consumer Growth, RL Environment
xAI's Colossus 2 - First Gigawatt Datacenter In The World, Unique RL Methodology, Capital Raise
On Site Turbines, Mississippi Expansion, Solaris Energy, Can xAI afford it?, Middle East Funding, Tesla, Talent Exodus, API revenue, Consumer Growth, RL Environment
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Another Giant Leap: The Rubin CPX Specialized Accelerator & Rack • New Prefill Specialized GPU, Rack Architecture, BOM, Disaggregated PD, Higher Perf per TCO, Lower TCO, GDDR7 & HBM Market Trends
Another Giant Leap: The Rubin CPX Specialized Accelerator & Rack
New Prefill Specialized GPU, Rack Architecture, BOM, Disaggregated PD, Higher Perf per TCO, Lower TCO, GDDR7 & HBM Market Trends
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Huawei Ascend Production Ramp: Die Banks, TSMC Continued Production, HBM is The Bottleneck • H20 Shipments, Blackwell B30A, Bottlenecks to Chinese Chip Production, Export Controls, CXMT, SMIC, Cambricon
Huawei Ascend Production Ramp: Die Banks, TSMC Continued Production, HBM is The Bottleneck
H20 Shipments, Blackwell B30A, Bottlenecks to Chinese Chip Production, Export Controls, CXMT, SMIC, Cambricon
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Amazon’s AI Resurgence: AWS & Anthropic's Multi-Gigawatt Trainium Expansion • Anthropic multi-gigawatt clusters, Trainium ramp, best TCO per memory bandwidth, system-level roadmap, Bedrock and internal models
Amazon’s AI Resurgence: AWS & Anthropic's Multi-Gigawatt Trainium Expansion
Anthropic multi-gigawatt clusters, Trainium ramp, best TCO per memory bandwidth, system-level roadmap, Bedrock and internal models
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
H100 vs GB200 NVL72 Training Benchmarks - Power, TCO, and Reliability Analysis, Software Improvement Over Time • Joules per Token, TCO Per Million Tokens, MFU, Tokens Per US Annual Household Energy Usage, DeepSeek 670B, GB200 Unreliability, Backplane Downtime
H100 vs GB200 NVL72 Training Benchmarks - Power, TCO, and Reliability Analysis, Software Improvement Over Time
Joules per Token, TCO Per Million Tokens, MFU, Tokens Per US Annual Household Energy Usage, DeepSeek 670B, GB200 Unreliability, Backplane Downtime
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
GPT-5 Set the Stage for Ad Monetization and the SuperApp • How ChatGPT will monetize free users, Router is the Release, AIs will serve Ads, Google's moat eroded?, The shift of purchasing intent queries
GPT-5 Set the Stage for Ad Monetization and the SuperApp
How ChatGPT will monetize free users, Router is the Release, AIs will serve Ads, Google's moat eroded?, The shift of purchasing intent queries
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Scaling the Memory Wall: The Rise and Roadmap of HBM • HBM4, Custom Base Die, Shoreline Expansion, Process Flow, China Domestic Production, Samsung Qualification
Scaling the Memory Wall: The Rise and Roadmap of HBM
HBM4, Custom Base Die, Shoreline Expansion, Process Flow, China Domestic Production, Samsung Qualification
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Robotics Levels of Autonomy • Single-Purpose Robots Automating Hundreds Of Jobs, Pick And Place With Low Autonomy Is Expensive, General-Purpose Autonomy Navigating And Inspecting Large Sites, Targeting Low-Skill Labor In Early Pilots With Promise, Autonomy Capable Of Any Task In Resea
Robotics Levels of Autonomy
Single-Purpose Robots Automating Hundreds Of Jobs, Pick And Place With Low Autonomy Is Expensive, General-Purpose Autonomy Navigating And Inspecting Large Sites, Targeting Low-Skill Labor In Early Pilots With Promise, Autonomy Capable Of Any Task In Resea
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
Intel 18A Details & Cost, Future of DRAM 4F2 vs 3D, Backside Power Adoption (or Not), China's FlipFET, Digital Twins from Atoms to Fabs, and More • VLSI 2025 Roundup
Intel 18A Details & Cost, Future of DRAM 4F2 vs 3D, Backside Power Adoption (or Not), China's FlipFET, Digital Twins from Atoms to Fabs, and More
VLSI 2025 Roundup
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
DeepSeek Debrief: >128 Days Later • Traffic and User Zombification, GPU Rich Western Neoclouds, Token Economics (Tokenomics) Sets the Competitive Landscape
DeepSeek Debrief: >128 Days Later
Traffic and User Zombification, GPU Rich Western Neoclouds, Token Economics (Tokenomics) Sets the Competitive Landscape
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
How Oracle Is Winning the AI Compute Market • Stargate, OpenAI, ByteDance, Unique Datacenter Strategy, Investment Grade Neocloud, Revenue and EBIT Forecast
How Oracle Is Winning the AI Compute Market
Stargate, OpenAI, ByteDance, Unique Datacenter Strategy, Investment Grade Neocloud, Revenue and EBIT Forecast
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
AI Training Load Fluctuations at Gigawatt-scale - Risk of Power Grid Blackout? • 108GW Large Load Queue, Tesla Megapacks, Supercapacitors, Gigawatt-scale Batteries, PyTorch No Power Plant Blow Up
AI Training Load Fluctuations at Gigawatt-scale - Risk of Power Grid Blackout?
108GW Large Load Queue, Tesla Megapacks, Supercapacitors, Gigawatt-scale Batteries, PyTorch No Power Plant Blow Up
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
NVIDIA Tensor Core Evolution: From Volta To Blackwell • Amdahl’s Law, Strong Scaling, Asynchronous Execution, Blackwell, Hopper, Ampere, Turing, Volta, TMA
NVIDIA Tensor Core Evolution: From Volta To Blackwell
Amdahl’s Law, Strong Scaling, Asynchronous Execution, Blackwell, Hopper, Ampere, Turing, Volta, TMA
newsletter.semianalysis.com
November 20, 2025 at 3:27 PM
The New AI Networks | Ultra Ethernet UEC | UALink vs Broadcom Scale Up Ethernet SUE • LibFabric, Packet Spraying, Rail Optimized, Congestion Control, ECN, ACK, Flow Control, PFC, UEC Challenges
The New AI Networks | Ultra Ethernet UEC | UALink vs Broadcom Scale Up Ethernet SUE
LibFabric, Packet Spraying, Rail Optimized, Congestion Control, ECN, ACK, Flow Control, PFC, UEC Challenges
newsletter.semianalysis.com
November 20, 2025 at 4:27 PM
Scaling Reinforcement Learning: Environments, Reward Hacking, Agents, Scaling Data • Infrastructure Bottlenecks and Changes, Distillation, Data is a Moat, Recursive Self Improvement, o4 and o5 RL Training, China Accelerator Production
Scaling Reinforcement Learning: Environments, Reward Hacking, Agents, Scaling Data
Infrastructure Bottlenecks and Changes, Distillation, Data is a Moat, Recursive Self Improvement, o4 and o5 RL Training, China Accelerator Production
newsletter.semianalysis.com
November 20, 2025 at 4:27 PM
AMD vs NVIDIA Inference Benchmark: Who Wins? - Performance & Cost Per Million Tokens • MI300X, MI325X, H100, H200, B200, MI355X, VLLM, SGLang, TRT-LLM, ROCm CI Lack of Coverage, Inflated AMD Rental Prices
AMD vs NVIDIA Inference Benchmark: Who Wins? - Performance & Cost Per Million Tokens
MI300X, MI325X, H100, H200, B200, MI355X, VLLM, SGLang, TRT-LLM, ROCm CI Lack of Coverage, Inflated AMD Rental Prices
newsletter.semianalysis.com
November 20, 2025 at 4:27 PM
AI Arrives In The Middle East: US Strikes A Deal with UAE and KSA • 5 GW Datacenter, HUMAIN, G42, Diversion and Misuse Risks, Security Requirements, American AI Wins
AI Arrives In The Middle East: US Strikes A Deal with UAE and KSA
5 GW Datacenter, HUMAIN, G42, Diversion and Misuse Risks, Security Requirements, American AI Wins
newsletter.semianalysis.com
November 20, 2025 at 4:27 PM