AMP创始人：GPU利用率远低于预期，AI算力浪费严重

Last 4 days before regular tickets sell out at AI Engineer World’s Fair - this is the single biggest gathering of AI Engineers, Founders, Leaders, and Researchers in the world. Attendees get >$5000 worth of sponsor credits and talk tracks are looking FANTASTIC. Join us!The AI scaling debate always focuses on the question of “how do we get more GPUs?” but the better question may be: how do we make the most of ones we already have.The fact that a frontier lab like xAI could be running at sub-10% MFU (Model FLOPs Utilization) is just a hint at what the real problem may be.tweetFor context, older frontier-scale training runs were already much higher than 10%. GPT-3 was around 21% MFU. Gopher was around 32%. Megatron-Turing NLG was around 30%. PaLM reached around 46%. And our guest Anjney says best-in-class MFU today is closer to 60–70%.PaLM: Scaling Language Modeling with PathwaysIt’s not necessarily that xAI is uniquely incompetent (it’s clear they have talented folks) but rather the priorities may be flipped in the GPU arms race.While GPU access is a bottleneck, simply increasing CapEx won’t automatically translate to better models as frontier AI is increasingly a systems problem: scheduling, utilization, networking, kernels, frameworks, data pipelines, parallelism, cluster reliability, and the thousand small decisions that determine whether your theoretical FLOPs become real training progress.From building Discord’s developer platform and backing frontier AI companies like Anthropic, Mistral, Black Forest Labs, and Periodic Labs to now building AMP’s independent compute grid, Anjney Midha has spent years close to the real bottlenecks of AI scaling. In this episode, Anjney joins swyx at Periodic Labs to unpack why the AI race is not just about buying more GPUs, why 95% utilization would have been considered an outage at Google, and why the next era of AI infrastructure has to be more aligned, more efficient, and more responsible.We go deep on AMP’s vision for a compute grid that makes FLOPs flow like megawatts, the difference between full-stack AI labs and horizontal pooling, why AI data centers need community buy-in, and how compute markets could evolve into something closer to an independent system operator. Anjney also explains why DeepMind’s unpublished research points to a market failure, why end-of-life prediction remains one of the most important AI applications he has thought about for fourteen years, and why “output maxing” may become a new discipline for frontier systems.We also discuss Anthropic’s culture, why “luck favors the prepared mind” in coding models, how Claude cracked coding, why too much capital too early can make AI labs fragile, what Periodic Labs is trying to do with science and superconductors, why great researchers can become great CEOs, and why Silicon Valley is both deeply missionary and deeply mercenary.We discuss:Why 95% utilization was considered an outage at GoogleWhy AI infrastructure waste compounds at frontier-lab scaleWhy “move fast and break things” does not work for AI data centersHow data center backlash, power grids, and community incentives shape AI scalingAMP’s vision for making FLOPs flow like megawattsWhy compute needs an independent system operatorHow interruptible demand and dynamic prioritization worked inside GoogleWhy DeepMind research hoarding creates negative externalitiesAMP’s 1.2GW base-load ambition and the need for 6GW of spike capacityWhy end-of-life prediction could become one of AI’s most important healthcare applicationsFrontier Systems, output maxing, and full-stack alignmentWhy APIs and abstraction layers become lossy as organizations scaleSuperconductors, standards, and the dream of lossless systemsSF Compute, open protocols, and the future of compute marketplacesWhy non-NVIDIA chips can still benefit from NVIDIA’s reference architectureTrust boundaries and why chip startups need visibility into future model architecturesWhy VCs often underestimate researchers as CEOsScientists as star athletes of the mindWhy great CEOs need to be confrontational up and down the stackWhy leading the frontier matters more than “winning”How Anthropic cracked codingWhy culture is fragile, not a permanent moatWhy hardship was a feature, not a bug, for AnthropicWhy Anthropic’s P0 was coding from day onePeriodic Labs, physics as the constraint, and technical realitySilicon Valley mercenaries, missionary teams, and what happens after a breakthroughAnjney MidhaLinkedIn: https://www.linkedin.com/in/anjneyX: https://x.com/AnjneyMidhaAMP PBCWebsite: https://amppublic.com/X: https://x.com/amppublicTimestamps00:00:00 Introduction00:00:09 Why AI Compute Is Being Wasted00:03:17 Responsible Infrastructure and Data Center Backlash00:06:07 AMP Grid: Making FLOPs Flow Like Megawatts00:12:41 Foundry, Frontier Labs, and Research Hoarding00:14:42 Gigawatt-Scale Compute and End-of-Life Prediction00:24:08 Frontier Systems, Output Maxing, and Alignment00:27:38 Compute Markets, SF Compute, and Non-NVIDIA Chips00:32:57 Trust Boundaries, Co-Design, and Researcher CEOs00:38:17 AI Coachella and First-Principles Thinking00:42:43 Leading vs Winning in Frontier AI00:45:54 How Anthropic Cracked Coding00:48:25 Culture, Hardship, and Anthropic’s P000:54:03 Periodic Labs, Physics, and Silicon Valley Mercenaries00:56:26 Rishi Valley, Singapore, and Money as a Measure00:58:47 Closing ThoughtsTranscriptIntroduction: Anjney Midha, AMP, and Compute WasteSwyx [00:00:00]: We’re in Periodic Labs with Anjney Midha, CEO, founder of AMP. Welcome.Compute Utilization: Node Allocation, MFU, and AlignmentAnjney [00:00:09]: Thanks for having me. At Google, there are two types of utilization usually, right? That you’re measuring in these clusters. One is node allocation, and then the other’s MFU. Node utilization is usually like what percentage of cards in the data center are just, used, and that, if it’s not at, 95%-Swyx [00:00:29]: There is no excuseAnjney [00:00:29]: There’s no excuse, right? I think 95% at Google, which is where my co-founder, Seb, came from, he built the Borg, PBorg/GQM scheduler at Google, and there I think 95% was considered an outage, so 96% node utilization is, should be standard. And most single-tenant clusters are not running at that. So that’s one. And then MFU should be, I would say the best in class today is somewhere between 60 and 70%. I think this is a leadership question, right? Fundamentally it’s an alignment question, which is are the people who are funding the cluster and then deploying the cluster actually aligned? And sometimes theoretically they are, but in practice the number of people in the chain, the supply chain between, the capital and all the way to whoever’s managing the cluster and then whoever’s measuring what the output is, are just so many, degrees of separation away that, the, The Have you ever heard the radian metaphor, which is at the beginning of an arc, if you have two arcs that are two lines that are just off by a few degrees, that-Swyx [00:01:33]: It spreads outAnjney [00:01:34]: It spreads out, right? Or at scale. And I think what’s happening is a lot of cluster implementations and infrastructure, a lot of frontier labs and other teams, that’s what’s happening, is they’re, they initialize the plan, which is kind of like North Star with a team that wants to do good, but then they’re, required to scale so fast instead of iteratively that the wastage just compounds really fast at scale. And so I think we know the answer, which is just do iterative bring ups. If you spend time with people who’ve been in the semiconductor industry or the DSN industry for a long time, this is not new, and I don’t think AI should be an excuse. Sure. Something What is new? Okay. We have a lot of new capabilities, but that doesn’t mean just abandon common sense. Common sense should always be in fashion. ? AI scaling doesn’t change the in fact, if anything, AI scaling should be putting a premium on the value of common sense

原文超出正文长度上限，此处截断——上游还有内容，完整版见上方「原文 ↗」。