{
  "source_file": "0930",
  "source_platform": "Substack",
  "author": "AI Made Simple\nDevansh",
  "title_listed": "How OpenAI Made Chip Design 3x Faster",
  "keywords": "芯片设计, AI加速, Jalapeño芯片, EDA, 硬件协同设计",
  "domain": "大模型算法",
  "content_type": "案例分析",
  "is_paid_flag": "免费",
  "url": "https://www.artificialintelligencemadesimple.com/p/how-openai-made-chip-design-3-faster",
  "publish_date": "2026",
  "corpus_stage": "预训练/硬件",
  "notes": "介绍OpenAI的Jalapeño定制芯片如何利用AI将芯片设计周期从24个月缩短至9个月",
  "id": "bb731dabc969",
  "collected_at": "2026-10-01T20:25:18",
  "title": "How OpenAI Made Chip Design 3× Faster [Guest] - by Devansh",
  "body_markdown": "It takes time to create work that’s clear, independent, and genuinely useful.\nIf you’ve found value in this newsletter, consider becoming a paid subscriber\n.\nIt helps me dive deeper into research, reach more people, stay free from ads/hidden agendas, and supports my crippling chocolate milk addiction.\nWe run on a “pay what you can” model\n—so if you believe in the mission, there’s likely a plan that fits (over here)\n.\nEvery subscription helps me stay independent, avoid clickbait, and focus on depth over noise, and I deeply appreciate everyone who chooses to support our cult.\nHelp me buy chocolate milk\nPS\n– Supporting this work doesn’t have to come out of your pocket. If you read this as part of your professional development, you can\nuse this email template\nto request reimbursement for your subscription.\nEvery month, the Chocolate Milk Cult reaches over a million Builders, Investors, Policy Makers, Leaders, and more.\nIf you’d like to meet other members of our community, please fill out this contact form here (\nI will never sell your data nor will I make intros w/o your explicit permission\n)\n-\nhttps://forms.gle/Pi1pGLuS1FmzXoLr6\nI recently came across this research on\nOpenAI’s new Jalapeno chip\nby\nAdvanced Notes\n. I appreciated the way it went beyond the (which are very good, as you will see in the table below), to instead provide a complete story on how AI may now be compressing the hardware-design cycle itself. Reading this article felt like a crash course in some of the important things to look out for when it comes to AI for chips.\nI’ve been thinking about how faster chip design may reduce the cost of being wrong about where AI is going. General-purpose GPUs dominate partly because nobody wants to lock a three-year silicon bet around a workload that may change in six months. If you can bring that cost down enough, we might see a world where GPUs discover the next important workload, after which you might design custom chips to solve that problem more efficiently (assuming it’s important enough). This would likely have some major implications for the market dynamics, software (I anticipate a rise in heterogenous computing), and even in AI/software capabilities(by enabling people to explore new kinds of computational workloads/paradigms that are a bit more awkward on current systems).For eg. take t\nhe “\nAn optical neural chip for implementing complex-valued neural network\n” which drops some serious heat in enabling a more bottom-up development of CVNNs. The authors “\nhighlight an optical neural chip (ONC) that implements truly complex-valued neural networks\n” w/ superior learning and efficiency.\n(Read a lot more about this topic here)\n.\nI could see this process enabling more complex valued chips, which can help us explore so many new directions that aren’t feasible today. As someone obsessed w/ researching the structure of intelligence and  computation, this implication causes my Hisoka to rise; hence this guest post. As you read it, here are some questions that might be worth playing with:\nIs nine months an outlier, or the beginning of a real compression in chip-development timelines?\nIf custom silicon gets much faster to build, how specialized do chips become?\nWho captures the value: hyperscalers, EDA incumbents, startups, or open source?\nHow powerful does the AI → better hardware → better AI feedback loop become?\nI don’t think we know yet. But this piece made me take those questions much more seriously. Hope you’ll enjoy it as much as I did.\n1. A Very Hot Jalapeño\nJalapeño, OpenAI’s first custom silicon, was one of the highlights of this week’s\nHot Chips conference at Stanford\n. Richard Ho, OpenAI’s head of hardware, walked the room through benchmark numbers that beat Nvidia’s Blackwell systems on throughput per watt and on latency, and most of the coverage since has understandably\nfocused on that comparison\n. But what drew my attention was the project timeline: nine months from initial RTL to tapeout.\nThe Jalapeño ASIC and the host-plus-accelerator rack pairing, as shown at Hot Chips 2026. Source: OpenAI.\nNine months is a very aggressive schedule: Jalapeño is built on TSMC’s N3 node, about as leading edge as it gets, and a typical project would take at least 24 months, if not more. That means, the OpenAI team cut the design cycle by somewhere between half and two-thirds.\n2. Ho-Ho-Ho...\nIf you’ve been following Richard Ho for a while, this result shouldn’t feel like it came out of thin air. I first came across his name in 2018, when he co-presented a paper on open source\nRISC-V verification infrastructure\nat that year’s RISC-V Summit, and reached out to him afterward to ask more details. There was nothing about AI-accelerated design in that talk, it was standard verification tooling work. But two things about it turned out to matter later: it was open source, and it was about building shared infrastructure rather than a proprietary point solution. Both these habits will resurface later in this story. Looking back at his career since 2018, Google’s TPU program, the Argos video transcoding chip, a short stop at the photonics company Lightmatter, and now OpenAI, there is a fairly consistent thread running through it: someone who has spent close to a decade trying to make chip design move at something closer to software speed. (Coincidentally, my keynote at DVCon Europe back in 2020 was on “\nWhat can verification learn from the software world?\n”)\n3. The Prequel: Argos and the First Speed-Up\nThe first public marker of that thread that I can point to with a paper attached is Argos, a custom video transcoding chip Google built for YouTube. Ho was one of fifty-two co-authors (!) on the\n2021 paper describing it\n, and its most useful section for our purposes is 3.4, on the use of high-level synthesis. Instead of writing RTL by hand, Google’s team described the design in C++ using a commercial HLS tool, Mentor Graphics’ Catapult, paired with an in-house integration layer called Taffel, and reported results that read almost like a pitch for the whole approach: five to ten times less code to write and review, testing throughput multiplied by seven to eight orders of magnitude over conventional RTL simulation, and bugs that used to take weeks to isolate fixed and verified in under a day. Google gave a deeper look at the same project later that year at\nHot Chips 33\n.\nAlthough this was still not leveraging AI in the way we use today, that design flow applied software engineering discipline applied to hardware design. But it is the same underlying bet Ho would keep making for the next five years: that the bottleneck in chip design is process, not physics, and process can be improved and streamlined.\n4. I Have a Dream!\nThe next marker I’d point to is a keynote Ho gave at\nSynopsys’s SNUG conference in March 2025\n, titled “\nScaling Compute for the Age of Intelligence\n.” He opened, fittingly, with a bit of personal history: when he first arrived in the Bay Area, his plan was to study AI at Stanford, under Prof. Ed Feigenbaum, one of the founders of the expert-systems field.\nRichard Ho at SNUG 2025, recalling his start at Stanford under Prof. Ed Feigenbaum.\nSource: OpenAI/Synopsys.\nThe keynote spent most of its time on reasoning models and scaling laws before turning, late, to chip design specifically. Ho put up a diagram of the existing flow, architecture, RTL design, logic synthesis, physical design, verification, manufacturing, laid against a hardware development timeline that, by his own account, runs twelve to twenty-four months.\nThe existing flow, as Ho presented it at SNUG 2025. Source: Synopsys / OpenAI.\nThen he put up a second, much simpler diagram and called it his dream: an AI and a human designer in a loop of propose, measure, and optimize, running straight through to production.\n“ML for ML Hardware. My dream!” Source: Synopsys / OpenAI.\nHo was explicit that hardware multi-year timelines sit awkwardly against an ML research cycle that can shift in weeks, a gap borne out by\none industry tracker’s release-cadence data\n, which puts major frontier model releases at roughly four weeks apart by early 2026, down from about six months two years earlier. If model development moves in weeks and chip development moves in years, the chip becomes the constraint on how fast AI itself can improve.\n5. Solving Design Bottlenecks\nWhen describing the current flow, Ho said that Verification and Physical Design (two parallel branches in his diagram) are always racing each other, and whichever one takes longer ends up setting the pace for the whole project. So he backed up his proposal for a dream flow with two examples that were already real back then, at least internally at Google. The first was placement, the step where a chip’s largest physical blocks get positioned on the die, using the reinforcement-learning approach published in\nNature in 2021\nand later commercialized under the name AlphaChip.\nHo’s placement example: the 2021 Nature paper behind AlphaChip. Source: Synopsys / OpenAI.\nThe second example was fittingly around verification: Design2Vec, a Google Research verification tool that learns a representation of a chip’s RTL well enough to predict test coverage in seconds instead of running an overnight simulation.\nHo’s verification example: coverage prediction with Design2Vec. Source: Synopsys / OpenAI.\nHe also mentioned, almost in passing, an experiment closer to home: asking ChatGPT to design a small asynchronous FIFO for clock-domain crossings, a classic interview question from his own early career, and getting back working SystemVerilog along with a UVM testbench and documentation. Not full-chip design, he was careful to say, but a meaningful step.\nTwo established techniques, plus a glimpse of a third still in its infancy, all resting on the same idea: let a model absorb enough about the design to shortcut a long design cycle.\n6. Living the Dream...\nFast forward to August 2026, three OpenAI senior engineers - Richard Ho, Ravi Narayanaswami, and Chris Leary -\nstand on stage and introduce Jalapeño\n, a chip that, among its other attributes, was designed in nine months.\nHere is the timeline slide they shared in their presentation:\nOpenAI’s own project timeline for Jalapeño, presented at Hot Chips 2026. Source: OpenAI.\nHow did the team achieve this breakthrough design cycle?\nProbably Ho’s “dream flow” presented at SNUG was not a distant dream.\nFirstly, note the nine months from Initial RTL to Tapeout and the other silicon milestones in blue. Then note the green milestones of OpenAI Codex enabling the faster chip design work. So using Codex to speedup design & verification is a given, probably much beyond the small FIFO use case Ho shared at SNUG.\nThe second “trick” that Ho showed at SNUG was the AlphaChip placement Reinforcement Learning tool developed at Google.\nNow, even if Ho mentioned this AlphaChip work at SNUG while already at OpenAI, I don’t have confirmation that this tool was used in the Jalapeño project.\nWhat I can say: Anna Goldie and Azalia Mirhoseini, the two researchers who built the original AlphaChip placement work at Google, left in 2025 to found\nRicursive Intelligence\n, and are explicitly building a more ambitious, sellable version of that tool for chipmakers and hyperscalers generally, not just Google. Coincidentally, in a\nTIME100 AI profile published just this week\n, Anna put the company’s ambition plainly, wanting to take chip design from years down to “\nmonths, weeks, maybe even days eventually\n.” (BTW, we hosted Anna to speak at one of our\nsessions\nat the recent DAC)\nSeparately, and just as speculatively: It just happens that Broadcom, OpenAI’s silicon implementation partner on Jalapeño, is also\nGoogle’s TPU partner\n, a relationship that goes back to 2016 and is now in its seventh generation. MediaTek, which just joined this year as a design partner on\nGoogle’s TPU v8 program\n, has separately built its own version of the\nAlphaChip placement approach into its most advanced chips\n, extending the open-sourced technique rather than licensing it commercially. None of that proves anything about Jalapeño specifically, though it does mean the channels through which this kind of technique could move between companies clearly exist, whether or not they were used here.\n7. Accelerated HW Synthesis\nThe one piece of the Jalapeño flow OpenAI did put a name to is XLS (that stands for “Accelerated HW Synthesis”) - a hardware description language that let OpenAI’s own internal models search implementations against measured power, performance, and area, and iterate against a human-optimized baseline. OpenAI’s own slide claims a fifty-six percent improvement on a BF16 multiplier, smaller but still real gains on a couple of other blocks, and major changes landing right up until the day of RTL freeze, the kind of late-stage flexibility that a traditional flow, where physical layout gets locked in early, generally doesn’t allow.\n“AI helped us build Jalapeño faster and made it better.” Source: OpenAI, Hot Chips 2026.\nUnder the hood, XLS is Google’s own open source hardware synthesis toolchain, published on\nGitHub\nunder an Apache license, and it is in a real sense Argos’s successor: Where Argos’s team wrote in C++ against Catapult, XLS designs are written in a language called DSLX and compiled down, through the toolchain’s own intermediate representation, to Verilog, with an LLVM-based just-in-time compiler standing in for the software-speed simulation that C++ and Catapult provided at Argos. One detail I find genuinely notable: Chris Leary, a listed contributor to the XLS project on GitHub, was one of three presenters, alongside Ho, at Jalapeño’s Hot Chips talk this week, a direct line from Google’s open source tooling into a leading-edge OpenAI and Broadcom tapeout.\nIt’s also, IMHO, a rather unique choice: Top-tier commercial chip design almost never runs on open source infrastructure, the industry defaults, nearly without exception, to Synopsys, Cadence and Siemens. XLS’s own GitHub page is candid that it remains experimental and is “not an officially supported Google product.” Whatever ran downstream of it for the actual physical implementation on TSMC’s N3 process was almost certainly commercial EDA, since\nXLS’s own open source path\n(Yosys and OpenROAD) targets open process kits like SKY130, not leading-edge commercial nodes. But the front end, the layer where OpenAI says its own models did the searching and optimizing, ran through code anyone can read on GitHub today.\nThe underlying message is that the key chip designers are telling the EDA industry, if you don’t move faster, we’ll find other solutions.\n8. Closing the Loop\nBut beyond the novelty of such an accelerated design cycle, the main insight that is worth sitting with is what it enables.\nIn the “Project timeline” slide OpenAI showed (see above), they added:\nHow to optimize compute for OpenAI’s workloads?\nOnce the importance of inference and agentic workloads came into focus, the need for a new design was clear.\nAs we already mentioned, Ho said the slow pace of designing chips compared to the fast update of models is a handicap on the silicon side.\nOpenAI’s first reasoning model, o1, shipped in September 2024, which is IMHO a plausible trigger for that realization. Reasoning models already require a slightly different type of hardware than “pure” inference, and agentic AI requires further optimization still. OpenAI’s own timeline states that the architecture concept was done in October 2024. By the time of that SNUG keynote, in March 2025, OpenAI was already six weeks into initial RTL work on Jalapeño. Ho, without ever saying so on stage, was already living out the dream flow he had just described to the room.\nTo fully close the loop, note the May’26 milestone says “Codex runs on Jalapeño.” OpenAI designed silicon that can run ChatGPT optimally, and Codex now runs optimally on this silicon.\nAlso note that this just in the same month of silicon back – a 1st silicon success!\n9. Is This the New Standard Pace?\nNone of this makes Jalapeño’s nine months a magic trick. It looks more like the payoff of a specific, decade-long institutional habit that moved from Google’s chip teams to OpenAI’s.\nI keep asking people, in different occasions, some version of the same question: how much will AI actually compress the chip design cycle – 10%? 50%? 90%?\nAt the GSA European Executive Forum earlier this summer in\nMunich\n, I polled the wide audience, and you can see their opinions in the slide below. Is 50% the new normal or is it just the beginning?\nLive poll results from my session at the GSA European Executive Forum, Munich, 2026.\nSource: Author’s own poll, Mentimeter.\nI also asked four VCs on my DAC panel (including one that invested in Ricursive), and you can read their answers\nhere\n.\nJalapeño’s timeline slide is, as far as I’ve seen, the first public answer that comes with a number large enough to take seriously.\nIt also leaves a question I didn’t set out to answer here: who actually owns this new layer of tooling (EDA? Startups? Open-source?), and what happens to the rest of the EDA industry if AI-driven design flows become standard practice rather than a Hot Chips keynote slide.\nThank you for being here, and I hope you have a wonderful day.\nDev <3\nI provide various consulting and advisory services. If you‘d like to explore how we can work together,\nreach out to me through any of my socials over here\nor reply to this email.\nI put a lot of work into writing this newsletter. To do so, I rely on you for support. If a few more people choose to become paid subscribers, the Chocolate Milk Cult can continue to provide high-quality and accessible education and opportunities to anyone who needs it. If you think this mission is worth contributing to, please consider a premium subscription. You can do so for less than the cost of a Netflix Subscription\n(pay what you want here)\n.\nSubscribe\nIf you liked this article and wish to share it, please refer to the following guidelines.\nShare\nThat is it for this piece. I appreciate your time. As always, if you’re interested in working with me or checking out my other work, my links will be at the end of this email/post. And if you found value in this write-up, I would appreciate you sharing it with more people.\nIt is word-of-mouth referrals like yours that help me grow.\nThe best way to share testimonials is to share articles and tag me in your post so I can see/share it.\nReach out to me\nUse the links below to check out my other content, learn more about tutoring, reach out to me about projects, or just to say hi.\nSmall Snippets about Tech, AI and Machine Learning over here\nAI Newsletter- https://artificialintelligencemadesimple.substack.com/\nMy grandma’s favorite Tech Newsletter- https://codinginterviewsmadesimple.substack.com/\nMy (imaginary) sister’s favorite MLOps Podcast-\nhttps://machine-learning-made-simple.medium.com/\nMy YouTube:\nhttps://www.youtube.com/@ChocolateMilkCultLeader/\nReach out to me on LinkedIn. Let’s connect:\nhttps://www.linkedin.com/in/devansh-devansh-516004168/\nMy Instagram:\nhttps://www.instagram.com/iseethings404/\nMy Twitter:\nhttps://twitter.com/Machine01776819",
  "metadata": {
    "title": "How OpenAI Made Chip Design 3× Faster [Guest] - by Devansh",
    "description": "What happens if hardware starts moving at AI speed?",
    "keywords": null,
    "author": "Devansh",
    "og:type": "article",
    "og:title": "How OpenAI Made Chip Design 3× Faster [Guest]",
    "og:description": "What happens if hardware starts moving at AI speed?",
    "og:image": "https://substackcdn.com/image/fetch/$s_!IZvP!,w_1200,h_675,c_fill,f_jpg,q_auto:good,fl_progressive:steep,g_auto/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe346a4e5-76cb-4fd9-be95-00321116b605_1846x510.png",
    "og:url": "https://www.artificialintelligencemadesimple.com/p/how-openai-made-chip-design-3-faster",
    "twitter:title": "How OpenAI Made Chip Design 3× Faster [Guest]",
    "twitter:description": "What happens if hardware starts moving at AI speed?",
    "twitter:image": "https://substackcdn.com/image/fetch/$s_!47g0!,f_auto,q_auto:best,fl_progressive:steep/https%3A%2F%2Fartificialintelligencemadesimple.substack.com%2Fapi%2Fv1%2Fpost_preview%2F213750828%2Ftwitter.jpg%3Fversion%3D4",
    "twitter:card": "summary_large_image",
    "article:modified_time": "2026-09-01T19:49:37.748Z"
  },
  "paywall": {
    "listed_paid": false,
    "detected": false,
    "marker": "",
    "public_only": false,
    "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙"
  },
  "crawl": {
    "success": true,
    "status_code": 200,
    "error": ""
  },
  "code_blocks": [],
  "images": [
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!IZvP!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe346a4e5-76cb-4fd9-be95-00321116b605_1846x510.png",
      "bytes": 222213,
      "reused": true,
      "local_path": "attachments/bb731dabc969/images/039.png",
      "src": "https://substackcdn.com/image/fetch/$s_!IZvP!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe346a4e5-76cb-4fd9-be95-00321116b605_1846x510.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!a22y!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2684a401-d982-40ce-b4cf-a2a3fa963078_800x164.png",
      "bytes": 27442,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-98bd4127300ae316aef1.png",
      "src": "https://substackcdn.com/image/fetch/$s_!a22y!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2684a401-d982-40ce-b4cf-a2a3fa963078_800x164.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!Gifo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff2099058-ba20-45b6-9939-02df7c0e4dfb_714x403.jpeg",
      "bytes": 45723,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-fe468e3df5077a00c94f.jpeg",
      "src": "https://substackcdn.com/image/fetch/$s_!Gifo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff2099058-ba20-45b6-9939-02df7c0e4dfb_714x403.jpeg",
      "alt": "Hot Chips 2026 OpenAI Jalapeno Slide 2 Jalapeño ASIC + System",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!bD5D!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F39bee364-9dd8-4ebd-8370-c5e9aba7a585_673x301.png",
      "bytes": 83262,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-cd4e33b07ed15b6d32f0.png",
      "src": "https://substackcdn.com/image/fetch/$s_!bD5D!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F39bee364-9dd8-4ebd-8370-c5e9aba7a585_673x301.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!IyPw!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F83357da2-ca1f-4cef-b6f9-a97615b6c71c_671x267.png",
      "bytes": 78051,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-0ca24b0e09131918ae0b.png",
      "src": "https://substackcdn.com/image/fetch/$s_!IyPw!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F83357da2-ca1f-4cef-b6f9-a97615b6c71c_671x267.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!nYdv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffe51a34f-da2b-428b-88eb-2e9d2007644e_643x256.png",
      "bytes": 71205,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-0753efa6adf2deac7ff1.png",
      "src": "https://substackcdn.com/image/fetch/$s_!nYdv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffe51a34f-da2b-428b-88eb-2e9d2007644e_643x256.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!P3tC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa670958d-ae72-4698-b671-4ed83ebc8ff4_677x269.png",
      "bytes": 92280,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-76a8abdf50f20e0ece8b.png",
      "src": "https://substackcdn.com/image/fetch/$s_!P3tC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa670958d-ae72-4698-b671-4ed83ebc8ff4_677x269.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!0HFd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F028ced73-bb71-47c3-85f2-4d900982d709_749x340.png",
      "bytes": 89959,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-4ff5dbc5e97ff1312069.png",
      "src": "https://substackcdn.com/image/fetch/$s_!0HFd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F028ced73-bb71-47c3-85f2-4d900982d709_749x340.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!spC8!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff7d314aa-318f-4e37-8bf2-2fbfa2c3978d_761x429.jpeg",
      "bytes": 30489,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-1064214f81585ae9d1cf.jpeg",
      "src": "https://substackcdn.com/image/fetch/$s_!spC8!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff7d314aa-318f-4e37-8bf2-2fbfa2c3978d_761x429.jpeg",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!e-_5!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6f5b855a-f714-49fa-8bb3-1e84c01bf51e_702x396.png",
      "bytes": 62090,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-315b0423df2a08df4b2d.png",
      "src": "https://substackcdn.com/image/fetch/$s_!e-_5!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6f5b855a-f714-49fa-8bb3-1e84c01bf51e_702x396.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!bvSD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3d9f7f6-8852-46cd-a3a5-cb51dbcfa761_779x441.png",
      "bytes": 13692,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-04845ca02eefed176cef.png",
      "src": "https://substackcdn.com/image/fetch/$s_!bvSD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3d9f7f6-8852-46cd-a3a5-cb51dbcfa761_779x441.png",
      "alt": "",
      "downloaded": true
    },
    {
      "status": "ok",
      "url": "https://substackcdn.com/image/fetch/$s_!rb4T!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53f2c14e-a83a-4d1b-a099-afc1d01a9f91_848x193.png",
      "bytes": 28344,
      "reused": false,
      "local_path": "attachments/bb731dabc969/images/body-cf1717b2687cabdf11fc.png",
      "src": "https://substackcdn.com/image/fetch/$s_!rb4T!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53f2c14e-a83a-4d1b-a099-afc1d01a9f91_848x193.png",
      "alt": "",
      "downloaded": true
    }
  ],
  "audio": [],
  "videos": [
    {
      "url": "https://www.youtube.com/watch?v=9y_4ucNO04U",
      "status": "ok",
      "local_path": "attachments/bb731dabc969/videos/9y_4ucNO04U.mp4",
      "bytes": 154369731,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "999.861000"
    },
    {
      "url": "https://www.youtube.com/@ChocolateMilkCultLeader/",
      "src": "https://www.youtube.com/@ChocolateMilkCultLeader/",
      "status": "reference_only",
      "kind": "channel_reference"
    }
  ],
  "links": [
    {
      "href": "https://artificialintelligencemadesimple.substack.com/subscribe",
      "label": "If you’ve found value in this newsletter, consider becoming a paid subscriber",
      "type": "a"
    },
    {
      "href": "https://artificialintelligencemadesimple.substack.com/p/help-me-take-ai-made-simple-to-the",
      "label": "We run on a “pay what you can” model",
      "type": "a"
    },
    {
      "href": "https://docs.google.com/document/d/1xy6CNE8S7ZIM1LPKc5qdjwLJcqj6lwxzv3HFz3gEU14/edit?usp=sharing",
      "label": "use this email template",
      "type": "a"
    },
    {
      "href": "https://docs.google.com/forms/d/e/1FAIpQLScCSWYlzouT8pzhfl0A2xdA0BxAPYg75h9F-WNkN8XuowpstA/viewform?usp=dialog",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://forms.gle/Pi1pGLuS1FmzXoLr6",
      "label": "https://forms.gle/Pi1pGLuS1FmzXoLr6",
      "type": "a"
    },
    {
      "href": "https://moshezalcberg.substack.com/p/hot-chips-and-spiced-up-jalapeno",
      "label": "OpenAI’s new Jalapeno chip",
      "type": "a"
    },
    {
      "href": "https://open.substack.com/users/23257572-advanced-notes?utm_source=mentions",
      "label": "Advanced Notes",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!IZvP!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe346a4e5-76cb-4fd9-be95-00321116b605_1846x510.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.nature.com/articles/s41467-020-20719-7",
      "label": "An optical neural chip for implementing complex-valued neural network",
      "type": "a"
    },
    {
      "href": "https://www.artificialintelligencemadesimple.com/p/your-data-is-not-real-a-case-for?utm_source=publication-search",
      "label": "(Read a lot more about this topic here)",
      "type": "a"
    },
    {
      "href": "https://www.servethehome.com/openai-jalapeno-asic-at-hot-chips-2026/",
      "label": "Hot Chips conference at Stanford",
      "type": "a"
    },
    {
      "href": "https://www.tomshardware.com/tech-industry/artificial-intelligence/hot-chips-2026-openais-jalapeno-ai-asic-unpacked-accelerator-developed-using-ai-achieves-efficiency-and-throughput-gains-against-power-hungry-blackwell",
      "label": "focused on that comparison",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!Gifo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff2099058-ba20-45b6-9939-02df7c0e4dfb_714x403.jpeg",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/watch?v=9y_4ucNO04U",
      "label": "RISC-V verification infrastructure",
      "type": "a"
    },
    {
      "href": "https://www.veriests.com/news-and-press/veriest-at-dvcon-2020/",
      "label": "What can verification learn from the software world?",
      "type": "a"
    },
    {
      "href": "https://dl.acm.org/doi/10.1145/3445814.3446723",
      "label": "2021 paper describing it",
      "type": "a"
    },
    {
      "href": "https://www.servethehome.com/google-vcu-video-coding-unit-at-hot-chips-33/",
      "label": "Hot Chips 33",
      "type": "a"
    },
    {
      "href": "https://www.synopsys.com/community/snug/snug-news/2025-keynote/thank-you.html",
      "label": "Synopsys’s SNUG conference in March 2025",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!bD5D!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F39bee364-9dd8-4ebd-8370-c5e9aba7a585_673x301.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!IyPw!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F83357da2-ca1f-4cef-b6f9-a97615b6c71c_671x267.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!nYdv!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffe51a34f-da2b-428b-88eb-2e9d2007644e_643x256.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.digitalapplied.com/blog/frontier-model-release-velocity-index-q2-2026",
      "label": "one industry tracker’s release-cadence data",
      "type": "a"
    },
    {
      "href": "https://www.nature.com/articles/s41586-021-03544-w",
      "label": "Nature in 2021",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!P3tC!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa670958d-ae72-4698-b671-4ed83ebc8ff4_677x269.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!0HFd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F028ced73-bb71-47c3-85f2-4d900982d709_749x340.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.servethehome.com/hot-chips-2026-openai-jalapeno-slide-03/",
      "label": "stand on stage and introduce Jalapeño",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!spC8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff7d314aa-318f-4e37-8bf2-2fbfa2c3978d_761x429.jpeg",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://www.ricursive.com/",
      "label": "Ricursive Intelligence",
      "type": "a"
    },
    {
      "href": "https://time.com/collection/time100-ai/2026/azalia-mirhoseini-anna-goldie/",
      "label": "TIME100 AI profile published just this week",
      "type": "a"
    },
    {
      "href": "https://63dac.conference-program.com/session/?sess=sess298",
      "label": "sessions",
      "type": "a"
    },
    {
      "href": "https://www.cnbc.com/2025/11/24/broadcom-is-joining-alphabet-in-the-ai-rally-why-investors-are-jumping-in.html",
      "label": "Google’s TPU partner",
      "type": "a"
    },
    {
      "href": "https://www.tomshardware.com/tech-industry/semiconductors/google-splits-its-tpu-into-two-chips-for-the-first-time-with-training-and-inference-variants",
      "label": "Google’s TPU v8 program",
      "type": "a"
    },
    {
      "href": "https://deepmind.google/blog/how-alphachip-transformed-computer-chip-design/",
      "label": "AlphaChip placement approach into its most advanced chips",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!e-_5!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6f5b855a-f714-49fa-8bb3-1e84c01bf51e_702x396.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://github.com/google/xls",
      "label": "GitHub",
      "type": "a"
    },
    {
      "href": "https://antmicro.com/blog/2023/09/accelerating-digital-block-design-with-googles-xls",
      "label": "XLS’s own open source path",
      "type": "a"
    },
    {
      "href": "https://www.gsaglobal.org/forums/taking-the-pulse-in-munich/",
      "label": "Munich",
      "type": "a"
    },
    {
      "href": "https://substackcdn.com/image/fetch/$s_!bvSD!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3d9f7f6-8852-46cd-a3a5-cb51dbcfa761_779x441.png",
      "label": "",
      "type": "a"
    },
    {
      "href": "https://moshezalcberg.substack.com/p/4-vcs-back-to-dac",
      "label": "here",
      "type": "a"
    },
    {
      "href": "https://linktr.ee/iseethings404",
      "label": "reach out to me through any of my socials over here",
      "type": "a"
    },
    {
      "href": "https://artificialintelligencemadesimple.substack.com/p/read-this-if-you-want-to-share-ai",
      "label": "If you liked this article and wish to share it, please refer to the following guidelines.",
      "type": "a"
    },
    {
      "href": "https://www.artificialintelligencemadesimple.com/p/how-openai-made-chip-design-3-faster?utm_source=substack&utm_medium=email&utm_content=share&action=share",
      "label": "Share",
      "type": "a"
    },
    {
      "href": "https://www.instagram.com/yourgodandsavior/",
      "label": "Small Snippets about Tech, AI and Machine Learning over here",
      "type": "a"
    },
    {
      "href": "https://artificialintelligencemadesimple.substack.com/",
      "label": "AI Newsletter- https://artificialintelligencemadesimple.substack.com/",
      "type": "a"
    },
    {
      "href": "https://codinginterviewsmadesimple.substack.com/",
      "label": "My grandma’s favorite Tech Newsletter- https://codinginterviewsmadesimple.substack.com/",
      "type": "a"
    },
    {
      "href": "https://open.spotify.com/show/7wZygk3mUUqBaRbBGB1lgh?si=b93afa69de994c88&nd=1&dlsi=ac0f8d9ac35642d5",
      "label": "My (imaginary) sister’s favorite MLOps Podcast-",
      "type": "a"
    },
    {
      "href": "https://www.youtube.com/@ChocolateMilkCultLeader/",
      "label": "https://www.youtube.com/@ChocolateMilkCultLeader/",
      "type": "a"
    },
    {
      "href": "https://www.linkedin.com/in/devansh-devansh-516004168/",
      "label": "https://www.linkedin.com/in/devansh-devansh-516004168/",
      "type": "a"
    },
    {
      "href": "https://www.instagram.com/iseethings404/",
      "label": "https://www.instagram.com/iseethings404/",
      "type": "a"
    },
    {
      "href": "https://twitter.com/Machine01776819",
      "label": "https://twitter.com/Machine01776819",
      "type": "a"
    }
  ],
  "repair_version": 2,
  "repaired_at": "2026-10-14T10:47:16Z",
  "source_url": "https://www.artificialintelligencemadesimple.com/p/how-openai-made-chip-design-3-faster",
  "papers": [],
  "actual_video_files": [
    {
      "local_path": "attachments/bb731dabc969/videos/9y_4ucNO04U.mp4",
      "bytes": 154369731,
      "valid": true,
      "height": 1080,
      "width": 1920,
      "duration": "999.861000",
      "status": "ok"
    }
  ],
  "body_chars": 19210,
  "expected_body_images": 12,
  "discovery_status": "ok",
  "delivery_status": "ok",
  "missing_reason": "",
  "raw_html": "attachments/bb731dabc969/pages/source.html",
  "local_html": "attachments/bb731dabc969/pages/article.html",
  "body_base_path": "attachments/bb731dabc969/pages/",
  "body_markdown_path": "attachments/bb731dabc969/pages/article.md"
}