[![SemiAnalysis](https://substackcdn.com/image/fetch/$s_!-cKQ!,w_40,h_40,c_fill,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc56db922-37c2-4d28-baac-2f4653e1ab00_512x512.png)](https://newsletter.semianalysis.com/)
# [![SemiAnalysis](../images/013.png)](https://newsletter.semianalysis.com/)
SubscribeSign in
[Home](https://newsletter.semianalysis.com/)
[Main Site](https://semianalysis.com)
[Models & Research](https://semianalysis.com/models-research/)
[Die Yield Calculator](https://semianalysis.com/die-yield-calculator/)
[Compliance](https://semianalysis.com/compliance-polices/)
[Archive](https://newsletter.semianalysis.com/archive)
[About](https://newsletter.semianalysis.com/about)
## Page not found
LatestTopDiscussions
[How GLM5.3 Sparse Attention Affects HBM Memory Usage](https://newsletter.semianalysis.com/p/sparse-savings-persistent-demand-inside-glm53)
[GLM-5.3, KV Cache Offloading, HiSparse, AgentX TileRT, InferenceX DeepSeek Sparse Attention, IndexShare, Single-rollout Asynchronous Optimization…](https://newsletter.semianalysis.com/p/sparse-savings-persistent-demand-inside-glm53)
Sep 28 • [Kimbo Chen](https://substack.com/@kimbobachen), [Alec Ibarra](https://substack.com/@adibarrasemi), [Wenyao Gao](https://substack.com/@wenyaogao), [Pratt Bhatt](https://substack.com/@pratt407051), [Bryan Shan](https://substack.com/@cheese01), and [Dylan Patel](https://substack.com/@semianalysis)
80
2
![](../images/014.jpeg)
[Intel Panther Lake Teardown](https://newsletter.semianalysis.com/p/intel-panther-lake-teardown)
[Taking a look inside Intel’s latest consumer chip and 18A process node](https://newsletter.semianalysis.com/p/intel-panther-lake-teardown)
Sep 26 • [Adith Shankar](https://substack.com/@semishankar), [Daniel Sanchez](https://substack.com/@danchezsemianalysis), [Allison Elliott](https://substack.com/@allisonelliott1), [Sarah Lawrence](https://substack.com/@sarahlaw77), [Afzal Ahmad](https://substack.com/@afzalahmad), [Andrew Wagner](https://substack.com/@semiwags), [STEEL Team](https://substack.com/@steelsemi), and [Dylan Patel](https://substack.com/@semianalysis)
103
8
![](../images/032.png)
[The Chinese AI Infrastructure Boom: Introducing the SemiAnalysis China Datacenter Model](https://newsletter.semianalysis.com/p/the-chinese-ai-infrastructure-boom)
[1,000+ facilities across 60+ operators mapped, built retail-first and flipped by AI, largest hyperscaler leases 1/5 national capacity, 100MW in 12…](https://newsletter.semianalysis.com/p/the-chinese-ai-infrastructure-boom)
Sep 25 • [Everlyn](https://substack.com/@everlyn832977), [Dylan Patel](https://substack.com/@semianalysis), and [Patrick Schaabi](https://substack.com/@patrickschaabi)
136
12
![](../images/body-3e6affd0a71ab2f9fd64.png)
[ClusterMAX 3.0: The Industry Standard GPU Cloud Rating System Returns](https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard)
[In gory detail: reliability, performance, support, pricing—and, of course, security—in our most thorough analysis of GPU cloud providers globally.](https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard)
Sep 23 • [Jordan Nanos](https://substack.com/@jnanos), [Sam Harshe](https://substack.com/@sjharshe), [Samuel Kruse](https://substack.com/@samdkruse), [Pratt Bhatt](https://substack.com/@pratt407051), [Billy Cao](https://substack.com/@aliencaocao), [Jack Carson](https://substack.com/@jdvcarson), and [Dylan Patel](https://substack.com/@semianalysis)
91
9
![](../images/body-2ef8a370ed00c578df0e.png)
[Computation and Data Movement for Inference](https://newsletter.semianalysis.com/p/computation-and-data-movement-for)
[Mapping MoE models onto inference hardware: structure, flow, and efficient serving](https://newsletter.semianalysis.com/p/computation-and-data-movement-for)
Sep 21 • [Tanj Bennett](https://substack.com/@tanjbennett), [Bryan Shan](https://substack.com/@cheese01), and [Dylan Patel](https://substack.com/@semianalysis)
217
7
![](../images/body-05e1d617315041f4854a.png)
[Engrams Embedding Entendre: Codesign for Efficient DRAM/SSD Offloading](https://newsletter.semianalysis.com/p/engrams-embedding-entendre-codesign)
[New Model Architecture Implications for TAM of DRAM/NVMe, DeepSeek V4.1 Flash, AgentX, InferenceX, NVMe experiments](https://newsletter.semianalysis.com/p/engrams-embedding-entendre-codesign)
Sep 18 • [Bryan Shan](https://substack.com/@cheese01), [Cam Quilici](https://substack.com/@camjquilici), [Alec Ibarra](https://substack.com/@adibarrasemi), [Kimbo Chen](https://substack.com/@kimbobachen), [Myron Xie](https://substack.com/@myronxie), and [Dylan Patel](https://substack.com/@semianalysis)
94
5
![](../images/body-360bd2b8bceb50d04d3f.png)
[Everyone Says Datacenter Moratoriums Are Killing the US Buildout. We disagree ](https://newsletter.semianalysis.com/p/everyone-says-datacenter-moratoriums)
[300+ moratoriums mapped, 20GW sits inside a restricted local boundary, 1,525MW actually slips, 2.3GW nationwide including New York](https://newsletter.semianalysis.com/p/everyone-says-datacenter-moratoriums)
Sep 15 • [Maya Barkin](https://substack.com/@mayabkn), [Reyk Knuhtsen](https://substack.com/@reykknuhtsen), [Jeremie Eliahou Ontiveros](https://substack.com/@jeremieeliahouontiveros), and [Dylan Patel](https://substack.com/@semianalysis)
138
12
![](../images/body-870b84fe62c624c57c4c.png)
[Vera Rubin NVL72 Agentic Inference: 67x better Performance per Dollar](https://newsletter.semianalysis.com/p/vera-rubin-nvl72-agentic-inference)
[Jensen Sandbagging Performance Again, 2x more Annual Profit Per GigaWatt, The More you Buy, The More you Earn, AgentX, InferenceX, Extreme Co-Design](https://newsletter.semianalysis.com/p/vera-rubin-nvl72-agentic-inference)
Sep 14 • [Bryan Shan](https://substack.com/@cheese01), [Alec Ibarra](https://substack.com/@adibarrasemi), [Cam Quilici](https://substack.com/@camjquilici), [Wenyao Gao](https://substack.com/@wenyaogao), and [Dylan Patel](https://substack.com/@semianalysis)
100
6
9
![](../images/body-defcd7a9af191fd1ba9f.png)
[A Brain Too Big to Carry — On-Device vs Datacenter Inference](https://newsletter.semianalysis.com/p/a-brain-too-big-to-carry-on-device)
[Robot Models, Silicon & DRAM Efficiency, Jetson Thor vs. B300 TCO, Deployments, The Network Wall](https://newsletter.semianalysis.com/p/a-brain-too-big-to-carry-on-device)
Sep 14 • [Ivan Chiam](https://substack.com/@ivanchiam), [Gianluca Bencomo](https://substack.com/@bencomog), [Zane Fong](https://substack.com/@zanefong), [Bryan Shan](https://substack.com/@cheese01), [Dylan Patel](https://substack.com/@semianalysis), and [Reyk Knuhtsen](https://substack.com/@reykknuhtsen)
170
2
10
![](../images/body-eb253f7dfce5ae8c96d7.png)
[Long Live the Short King: Why 4-hi HBM Wins](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi)
[Same Bandwidth, Fewer Dies: How 4-hi HBM Cuts Inference Costs and Makes Scarce DRAM Go Further](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi)
Sep 13 • [Myron Xie](https://substack.com/@myronxie), [Bryan Shan](https://substack.com/@cheese01), [Harrison Barclay](https://substack.com/@harrison905588), [Minjae Kang](https://substack.com/@minjae00), and [Dylan Patel](https://substack.com/@semianalysis)
143
4
11
![](../images/body-b0fb8376e5080b2b9af3.png)
[Nvidia’s Backstop Universe – Heads I Win, Tails Who Loses? ](https://newsletter.semianalysis.com/p/nvidias-backstop-universe-heads-i)
[The $11T AI Buildout, Nvidia’s Backstop Economics, and the Limits of Nvidia’s Balance Sheet](https://newsletter.semianalysis.com/p/nvidias-backstop-universe-heads-i)
Sep 11 • [Daniel Nishball](https://substack.com/@danielnishball730869), [Oliver Kennon](https://substack.com/@oliverkennon1), and [Terence Ong](https://substack.com/@tarzaniskingg)
132
2
7
![](../images/body-1105ff4fec34b3ce0b87.png)
[What is So Hard About Behind-The-Meter Power For Datacenters? Part 1 ](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the)
[Dumb Science Experiments vs. Money Printing Machines](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the)
Sep 10 • [Ellie Holbrook](https://substack.com/@ellieholbrook), [Robert Boswall](https://substack.com/@robertboswallsa), [Jeremie Eliahou Ontiveros](https://substack.com/@jeremieeliahouontiveros), [Nicolas Bontigui](https://substack.com/@nicolasbontigui), and [Dylan Patel](https://substack.com/@semianalysis)
145
2
13
![](../images/body-d96d03a6c7e3667a3b42.png)
[TPU Inference Externalization Full Steam Ahead - InferenceX](https://newsletter.semianalysis.com/p/tpu-inferencex-full-steam)
[InferenceX, Up to 50% Better Performance per Dollar, Rapid Externalization of TPU stack, Growing Customer Base, Ironwood, TPUv8i, Reducing CUDA Moat](https://newsletter.semianalysis.com/p/tpu-inferencex-full-steam)
Sep 7 • [Alec Ibarra](https://substack.com/@adibarrasemi), [Cam Quilici](https://substack.com/@camjquilici), [Bryan Shan](https://substack.com/@cheese01), [Wenyao Gao](https://substack.com/@wenyaogao), [Daniel Nishball](https://substack.com/@danielnishball730869), [Zane Fong](https://substack.com/@zanefong), and [Dylan Patel](https://substack.com/@semianalysis)
128
1
10
![](../images/body-7ee0970ce024a49d2df9.png)
[Korea’s Trillion-Dollar Sovereign AI Investment: Nvidia Wins, Hynix Loses](https://newsletter.semianalysis.com/p/koreas-trillion-dollar-sovereign)
[Korea hosts a Squid Games, National AI Tournament, the best non-Chinese open source model gets eliminated, why Nvidia needs open source, implications…](https://newsletter.semianalysis.com/p/koreas-trillion-dollar-sovereign)
Sep 1 • [Max Kan](https://substack.com/@mlkan), [Ray Wang](https://substack.com/@rwang07semis), [Myron Xie](https://substack.com/@myronxie), and [Dylan Patel](https://substack.com/@semianalysis)
187
2
14
![](../images/body-77d1f7f4d0330730cfac.png)
### August 2026
[Most Neoclouds Suck At Security](https://newsletter.semianalysis.com/p/most-neoclouds-suck-at-security)
[OpenAI vs HuggingFace, Container Escapes, Kernel Bypass, Network Policies, Security Keys, Multi-tenant Grafana, and a ClusterMAX 3.0 Preview](https://newsletter.semianalysis.com/p/most-neoclouds-suck-at-security)
Aug 30 • [Jordan Nanos](https://substack.com/@jnanos), [Sam Harshe](https://substack.com/@sjharshe), [Pratt Bhatt](https://substack.com/@pratt407051), [Billy Cao](https://substack.com/@aliencaocao), [Jack Carson](https://substack.com/@jdvcarson), and [Dylan Patel](https://substack.com/@semianalysis)
147
5
9
![](../images/body-26eeeec963cdaf15fcaf.png)
[OpenAI Jalapeño: Better Than Nvidia Blackwell](https://newsletter.semianalysis.com/p/openai-jalapeno-better-than-nvidia)
[OpenAI’s self-designed ASIC compared with Rubin, Jalapeño’s TCO, throughput per MW, and spicy deets](https://newsletter.semianalysis.com/p/openai-jalapeno-better-than-nvidia)
Aug 25 • [Bryan Shan](https://substack.com/@cheese01), [Myron Xie](https://substack.com/@myronxie), [Jordan Nanos](https://substack.com/@jnanos), [Wega Chu](https://substack.com/@wegachu), [Clara Ee](https://substack.com/@clarasemianalysis), and [Dylan Patel](https://substack.com/@semianalysis)
275
2
30
![](../images/body-7b9dbc38ea140d07d9d7.png)
[AgentX - InferenceXv3: Does CUDA Moat Hold up in Agentic Inferencing?](https://newsletter.semianalysis.com/p/agentx-inferencexv3-does-cuda-moat)
[$3 Million USD dataset open sourced, 1 Mil+ Context Length, Multiturn, Sub Agents 95%+ KVCache HitRate, GB300 NVL72, MI355, B200](https://newsletter.semianalysis.com/p/agentx-inferencexv3-does-cuda-moat)
Aug 24 • [Cam Quilici](https://substack.com/@camjquilici), [Bryan Shan](https://substack.com/@cheese01), [Alec Ibarra](https://substack.com/@adibarrasemi), [Daniel Nishball](https://substack.com/@danielnishball730869), [Zane Fong](https://substack.com/@zanefong), [Kimbo Chen](https://substack.com/@kimbobachen), and [Dylan Patel](https://substack.com/@semianalysis)
89
1
8
![](../images/body-4da25accd72e3cedf538.png)
[Are Open Models Catching Up?](https://newsletter.semianalysis.com/p/are-open-models-catching-up)
[Comparing open vs. closed models across the eras of frontier models, Is the gap narrowing?](https://newsletter.semianalysis.com/p/are-open-models-catching-up)
Aug 21 • [Evan Cloutier](https://substack.com/@evancloutier), [Max Kan](https://substack.com/@mlkan), [Jordan Nanos](https://substack.com/@jnanos), and [Dylan Patel](https://substack.com/@semianalysis)
255
4
29
![](../images/body-2b92b71cb10eb648fae9.png)
[Cerebras's Next Generation CS-4: Fast Just Got Faster](https://newsletter.semianalysis.com/p/cerebrass-next-generation-cs-4-fast)
[Double the Performance, Double the Power, Double the Fun](https://newsletter.semianalysis.com/p/cerebrass-next-generation-cs-4-fast)
Aug 19 • [Myron Xie](https://substack.com/@myronxie), [Bryan Shan](https://substack.com/@cheese01), [Wega Chu](https://substack.com/@wegachu), [Jordan Nanos](https://substack.com/@jnanos), and [Dylan Patel](https://substack.com/@semianalysis)
142
1
6
![](../images/body-2c9fbcde9685ad4d5641.png)
[$12B of US ratepayers' money wasted on a modeling mistake and PJM wants to do it again](https://newsletter.semianalysis.com/p/12b-of-us-ratepayers-money-wasted)
[American Grid design needs an overhaul, Why it is good to be full of cold air.](https://newsletter.semianalysis.com/p/12b-of-us-ratepayers-money-wasted)
Aug 16 • [Robert Boswall](https://substack.com/@robertboswallsa), [Reyk Knuhtsen](https://substack.com/@reykknuhtsen), [Jeremie Eliahou Ontiveros](https://substack.com/@jeremieeliahouontiveros), [Oliver Kennon](https://substack.com/@oliverkennon1), [Dylan Patel](https://substack.com/@semianalysis), and [Dashboard American](https://substack.com/@dashboardamerican)
119
4
9
![](../images/body-5cd472cb9f59cc7c48bf.png)
[Ultra-High Interactivity on NVIDIA GPUs? - TileRT InferenceX ](https://newsletter.semianalysis.com/p/ultra-high-interactivity-on-nvidia)
[Can TileRT software on NVIDIA GPU compete with Cerebras, Groq LPU, SambaNova? Batch Size 1, Disaggregated engine, high throughput prefill engine, high…](https://newsletter.semianalysis.com/p/ultra-high-interactivity-on-nvidia)
Aug 10 • [Bryan Shan](https://substack.com/@cheese01), [Daniel Nishball](https://substack.com/@danielnishball730869), [Cam Quilici](https://substack.com/@camjquilici), [Kimbo Chen](https://substack.com/@kimbobachen), [Alec Ibarra](https://substack.com/@adibarrasemi), and [Dylan Patel](https://substack.com/@semianalysis)
148
1
6
![](../images/body-26dd4da5559a11e48984.png)
[SpaceX 10GW in 2027 – Why It’s Real, Will Drive $300B ARR for SpaceX, and Why Microsoft Will Be the Largest Offtaker](https://newsletter.semianalysis.com/p/spacex-10gw-in-2027-why-its-real)
[Inference at 100B/GW/year, SpaceX's stellar pace, Microsoft's 10GW 2026 Awakening, Azure Can Grow Triple-Digits](https://newsletter.semianalysis.com/p/spacex-10gw-in-2027-why-its-real)
Aug 7 • [Jeremie Eliahou Ontiveros](https://substack.com/@jeremieeliahouontiveros), [Reyk Knuhtsen](https://substack.com/@reykknuhtsen), [Jordan Nanos](https://substack.com/@jnanos), [Max Kan](https://substack.com/@mlkan), [Dylan Patel](https://substack.com/@semianalysis), and [Muhammad Zuhair](https://substack.com/@mzuhair)
302
11
20
![](../images/body-21babee2207c8b91af12.png)
[Gemini is Cooked but GCP is Cooking](https://newsletter.semianalysis.com/p/gemini-is-cooked-but-gcp-is-cooking)
[GCP YoY rev growth >100%, DeepMind's long term failure is Google Cloud's short term gain](https://newsletter.semianalysis.com/p/gemini-is-cooked-but-gcp-is-cooking)
Aug 7 • [Max Kan](https://substack.com/@mlkan), [Joey Brookhart](https://substack.com/@joeysemianalysis), [Doug O'Laughlin](https://substack.com/@mule), and [Dylan Patel](https://substack.com/@semianalysis)
324
9
30
![](../images/body-6d1e9ffa056c4c4ed265.png)
[Kimi K3, The Manos, The Mythos, The Legendos](https://newsletter.semianalysis.com/p/kimi-k3-the-manos-the-mythos-the)
[Kimi K3’s architecture: compressed memory, attention across depth, latent expert routing, and serving performance](https://newsletter.semianalysis.com/p/kimi-k3-the-manos-the-mythos-the)
Aug 3 • [Kimbo Chen](https://substack.com/@kimbobachen), [Shubham Choudhari](https://substack.com/@shubham007272), [Bryan Shan](https://substack.com/@cheese01), and [Dylan Patel](https://substack.com/@semianalysis)
176
12
![](../images/body-8f7c7f667af3963accf0.png)
© 2026 Dylan Patel · [Privacy](https://substack.com/privacy) ∙ [Terms](https://substack.com/tos) ∙ [Collection notice](https://substack.com/ccpa#personal-data-collected)
[ Start your Substack](https://substack.com/signup?utm_source=substack&utm_medium=web&utm_content=footer)[Get the app](https://substack.com/app/app-store-redirect?utm_campaign=app-marketing&utm_content=web-footer-button)
[Substack](https://substack.com) is the home for great culture