Much of Dwarkesh's argument hinges on this statment which *was* accurate but will be increasingly inaccurate on a go forward basis imo:    “American labs port across accelerators constantly. Anthropic's models are run on GPUs, they're run on Trainium, they're run on TPUs. There are so many things you can do, from distilling to a model that's well fit for your chips.”   As system level architectures diverge (torus vs. switched scale-up topologies, memory hierarchies, networking primitives), true portability is eroding. The Mi300 and Mi325 had roughly the same scale-up domain size as Hopper while Blackwell’s scale-up domain is 9x larger than the Mi355 scale-up domain, etc. Many frontier models are now being explicitly co-designed for inference on specific hardware like GB300 racks. Codex on Cerebras is another example. Those models run less efficiently on other systems and the performance differentials will only widen. A model that runs well on Google’s torus topology will run less efficiently on Nvidia’s switched scale-up topology and vice versa - the data traffic is fundamentally different as a byproduct of the models being parallelized across the different topologies. Google’s internal teams - and increasingly the Anthropic teams as they become the most important customer of almost every cloud - have the luxury of operating across the stack (models, chips, networking) - but that is not the case for the rest of the market and other prospective users. Anthropic is the exception, not the rule. To wit, Anthropic and Google allegedly have a mutual understanding where Anthropic can hire the TPU engineers they need every year to ensure that they can continue to get the most out of the TPU. Given the overwhelming importance of cost per token to the economics of the labs, models will be run where they run best. Most extremely large MoE models will run best on GB300s given the importance of having a switched scale-up network like NVLink for MoE inference. When training was the dominant cost for labs and power was broadly available, labs were optimizing to minimize capex dollars. Model portability was a way to create leverage over suppliers. I think that drove a lot of the focus on portability. Today, inference costs as measured by tokens per watt per dollar are everything. Inference is way more important than training costs (inference is effectively now part of training via RL). Labs are therefore now optimizing for inference. This means increasing co-design and higher go-forward switching costs for individual models between systems. I do think this explains why Anthropic and Nvidia came together: Anthropic needed Blackwells and Rubins to inference at least *some* of their models economically. And Mythos might just end up being released coincident with the availability of Rubins for inference. TLDR: as labs shift their focus from training to inference, the costs of portability and the upside of co-design to maximize tokens per watt per dollar both rise. Portability is likely to begin decreasing as a result.   I think what I might have respectfully added to Jensen’s answer is that systems evolve under local selective pressures. The evolutionary pressure in America is a shortage of watts so it makes sense for Nvidia to optimize, as an American company, for power efficiency and tokens per watt and stay on copper as long as possible. China has a surfeit of watts. Chinese AI systems are already taking advantage of this with the Huawei Cloudmatrix 384 and Atlas SuperPoD having an optical scale-up domain that is much larger than anything offered by Nvidia today at the cost of *much* higher power consumption and much lower tokens per watt. The networking primitives for this Huawei system are very different than those for Nvidia’s systems and a model that runs well on Nvidia will not run well on that system and vice versa. This means that if a Chinese ecosystem gets momentum, Chinese models might stop running well on American hardware. And when Chinese models run best on American hardware, America is in a better position as this gives America a degree of leverage and control over Chinese AI that it risks losing to an all-Chinese alternative ecosystem.   This architectural fork makes porting and distillation less effective and strengthens the pro-American national security case for selling China deprecated GPUs imo. Also I will attest that I did not wake up a loser this morning.

Excited to announce the SemiAnalysis Teardown Engineering & Evaluation Lab (STEEL) We have been building a state-of-the-art teardown lab in Oregon with $10s of millions Capex being spent capable of analyzing the world’s most advanced and important chips and process technologies over the last year and half. We have already generated revenue on advanced datacenter chip teardowns. This is a bit of inconvenient timing for TechInsights as they are private equity owned and currently being sold while having enjoyed virtually no credible competition for decades. This has led to TechInsights underinvesting in CAPEX. SemiAnalysis exceeds TechInsights in revenue despite no venture or private equity ownership and being founded only 6 years ago. Because we have no external investors and are founder led, we move faster, build faster, and we can release client chip teardowns for free regularly, while focusing on datacenter for our major clients. STEEL has a state of the art fleet of tools. We are less than a year and half into the journey here, but our goal is to be number 1 just like we are in AI, Datacenter, and Semiconductor market data, consulting, and technically analysis. Our largest customers which, include all of the hyperscalers and world's largest semiconductor companies, are excited for us to shake up the market.

New Top 10 companies in the world. 8/10 design silicon chips. 1/10 manufactures silicon chips for the 8. 1/10 drills for oil.

everyone bashing Semianalysis, FundAI, etc. should instead look within. why are we trading everything on leverage, responding FAST before reading digesting and being quick to freak out? the issue is the borderline insane overreactions to these notes, not the notes themselves

The AI buildout has been actively rotating from building chip factories to filling data centers. We've been tracking the ocean containers landing at US ports for ~50 AI-supply companies. The last 12 months: (cooling & power) ↑ Foxconn 243% · Vertiv 82% (fab tools) ↓ Tokyo Electron −77% · Applied Materials −66% You can see the shift in the freight before it hits the revenue line.

KV Cache re-use is the most important thing for agentic rollouts. We've integrated Mooncake Store into prime-rl with vLLM, you can now use it as a drop-in replacement for native CPU/Disk offloading, giving you cross-node prefix cache reuse to make your agents go brrr🚀

Local inference not looking so crazy anymore now that companies are revolting over token spend

Westmag is building American robot actuators and drone motors at scale. In 2025, @westmagco raised $11M led by @a16z, with participation from @FoundersFund, @LuxCapital, NFDG, @MenloVentures, and other top investors. Since then, we’ve been building industrial capacity, crawling up supply chains, and securing high-volume customers. Now, we’re ramping production at our factory in South San Francisco to deliver against committed offtake orders from high-volume customers. Westmag is committed to scaling quickly in the US to deliver millions of drone motors and robot actuators to the surging domestic and global market. We’re building the great American motor and actuator company.

IMPORTANT: it is important to understand that the CoreWeave & Microsoft photos are still Engineering/Quality Samples, and there is still some time before the software stack bring-up finishes & first production tokens are generated. The VR200 & MI455 rack metric to watch out for is time to first at-scale production token TTF-(ASP)-T. You can clearly see in the CW rack photos that none of the scale-out 800G OSFP cages are even populated.

Pitch Mango Pineapple. Higher margins.

I think Anthropic likely trained the Mythos base model from roughly October to December using on the order of 6.7e26–1.0e27 flops Since then, the RL-to-base-model-training flop ratio is plausibly somewhere around 0.5 and 3, depending on how much of the expanded Trainium 2 fleet was actually allocated to Mythos RL. The reason this range is plausible is that public AWS/Anthropic statements imply Anthropic-accessible Trainium2 capacity grew from roughly 500k chips around Rainier’s launch to over 1M Trainium2 chips for Claude training and serving.

Today, @AyarLabs announced it has joined the @nvidia NVLink Fusion ecosystem, introducing co-packaged optics as a foundational building block for hyperscalers and system innovators deploying heterogeneous compute in NVIDIA AI factories. Press Release: bit.ly/4oa8epa

If a baker makes a cake with their own recipe using ingredients from a mutual friend, then gives it to you, would you pass it off as your own? Mediatek makes a CPU with their own recipe with arm ingredients and gives it to Nvidia. Broadcom makes a TPU with their own recipe but with Google ingredients and gives it back to Google.