korrents

Dylan Patel

@dylan-patel · 50 positions · 0 changes of mind

Founder and chief executive of SemiAnalysis, the research firm that tracks the semiconductor and AI data-centre supply chain.

Dylan Patel did not write this page.

We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.

  1. Our reading

    An H100 is worth more today than it was when it was new, because the models it runs got better faster than its replacements arrived.

    Their wordsif improvement stopped you know here the value of an H100 is now predicated on the value that GPD 5.4 four can get out of it instead of the value that GP4 can get out of it and the margins and all that stuff that these labs are doing and they're in a competitive environment so their margins can't go to infinity. Um so you sort of have this like dynamic that is quite interesting in that an H100 is worth more today than it was 3 years ago.
    spoken · machine transcript hear it at 0:15:28 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 1st of 28 in this recording

    Interesting

  3. Our reading

    Committing to compute years in advance is the durable margin advantage in AI, because every increment of new capacity transacts at the new price.

    Their wordsA companies that have locked up, you know, and and don't have commitment issues, you know, have these 5-year contracts for compute, they've kind of locked in a humongous margin advantage because they've locked in compute for 5 years at a price of what it transacted at 5 years ago or three years ago or two years ago, whatever it is.
    spoken · machine transcript hear it at 0:21:32 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 2nd of 28 in this recording

    Interesting

  5. Our reading

    The model companies' margins will rise this year, because being capacity-constrained forces them to destroy their own demand.

    Their wordsUm I think at least this year we're going to see margins for the model vendors go up a lot, right? Because they're so capacity constrained, they have to demand destroy demand, right? there is there's no way they can continue anthropic can continue at the current pace without destroying demand.
    spoken · machine transcript hear it at 0:24:36 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 3rd of 28 in this recording

    Anthropic

    Interesting

  7. Our reading

    TSMC would rather give wafers to a slow, stable CPU business than to its fastest-growing AI customer.

    Their wordsTSMC is much more excited to give allocation to Graviton than they are to tranium because they view CPU business as more stable long-term growth right and as a company that is conservative and doesn't want to ride cycles of growth too hard you actually want to allocate to the uh the market that is more stable and lower growth rate first before you allocate all the incremental capacity to the fast growth rate market.
    spoken · machine transcript hear it at 0:27:20 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 4th of 28 in this recording

    semiconductors

    Interesting

  9. Our reading

    Anthropic saw the coming shortage of chips before Google did, and bought Google's own TPUs out from under it.

    Their wordsBecause Enthropic saw it before Google. And then Google had Nano Banano and Gemini 3 which caused their user metrics to skyrocket and leadership at Google was like oh and then they started making the statement of we have to double compute every is it 6 months or I don't remember the exact number that they said.
    spoken · machine transcript hear it at 0:32:25 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 5th of 28 in this recording

    AnthropicGoogle

    Interesting

  11. Our reading

    The binding constraint on AI is no longer power or data centres but the semiconductor supply chain, which has by far the longest lead times.

    Their wordsYeah, I think the biggest bottleneck is compute and for that the longest lead time supply chains are not power or data centers. They're actually the semiconductor supply chain themselves, right? It switches back from being power and data center uh as a major bottleneck to chips.
    spoken · machine transcript hear it at 0:34:51 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 6th of 28 in this recording

    data centerssemiconductors

    Interesting

  13. Our reading

    By the end of this decade the ceiling on AI compute is set by ASML, the lowest rung of the supply chain, and not by anything above it.

    Their wordsSo to scale compute further right there's some different bottlenecks this year next year uh but ultimately by 2829 the bottleneck falls to the lowest rung on the supply chain which is ASML right ASML makes the world's most complicated machine i.e. an EUV tool.
    spoken · machine transcript hear it at 0:37:03 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 7th of 28 in this recording

    semiconductors

    Interesting

  15. Our reading

    Hundreds of billions of dollars of AI value rest on about a billion dollars of lithography tooling that cannot be scaled quickly.

    Their wordsoh 50 gigawatts of economic you know sort of capex in in the data center and what gets built on top of that in terms of tokens is even larger right it might be hundred billion dollars worth of AI value into the supply chain is held up by this $1.2 two billion dollars worth of tooling that simply just cannot expand its supply chain quickly.
    spoken · machine transcript hear it at 0:40:48 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 8th of 28 in this recording

    data centers

    Interesting

  17. Our reading

    The world can make roughly 200 gigawatts of AI chips a year by 2030, which makes a gigawatt a week a plausible quarter of the market rather than a fantasy.

    Their wordsUm, and then you stack on 70 this year, 80 next year, growing to 100 by 2030. You're at like 700 EV tools by the end of the decade. Um, 700 EV tools, three and a half tools per gigawatt. um assuming it's all allocated to AI which it's not but three and a half tools per gigawatt gets you to 200 gigawatts worth of AI chips for the data centers to deploy
    spoken · machine transcript hear it at 0:42:25 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 9th of 28 in this recording

    data centers

    Interesting

  19. Our reading

    ASML has never raised the price of a tool by more than it raised the tool's capability, which makes it the most generous monopoly in the industry.

    Their wordsYou can you can take the margin like Nvidia takes the margin. memory players are taking the margin, but ASML has never risen the price more than they've increased the capability of the tool. Um, and so in a sense, they've always provided net benefit to their customer.
    spoken · machine transcript hear it at 0:45:42 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 10th of 28 in this recording

    semiconductors

    Interesting

  21. Our reading

    The semiconductor supply chain will underbuild not because it cannot expand but because nobody in it believes the demand is real.

    Their wordsUm in general the semiconductor supply chain has not right it's lived through the booms and bust and uh we can talk a bit more about it but basically no one you know some players as of very recently have like woken up but in general no one really sees demand for 200 gawatts a year of AI chips or you know trillions of dollars of spend a year in the semiconductor supply chain.
    spoken · machine transcript hear it at 0:46:35 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 11th of 28 in this recording

    semiconductors

    Interesting

  23. Our reading

    Falling back to older 7-nanometre chips would not rescue AI compute, because the real gap between chip generations is twentyfold, not the threefold the flops suggest.

    Their wordsSo when you look at inference at let's say 100 tokens a second for deepseek and kimk 2.5 hopper versus blackwell the performance difference is on the order of 20x
    spoken · machine transcript hear it at 1:02:11 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 12th of 28 in this recording

    Interesting

  25. Our reading

    China will have a working EUV machine long before it can build them in quantity, because production hell takes years after the lab result.

    Their wordsI think they'll have working tools. I don't think that they'll be able to manufacture a bunch yet, right? You know, there's they're sort of having it work and then there's production hell, right?
    spoken · machine transcript hear it at 1:07:30 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 13th of 28 in this recording

    China

    Interesting

  27. Our reading

    As labs sell finished white-collar work rather than reasoning traces, Chinese models will lose the ability to distil their way to parity.

    Their wordsAs we move from, you know, hey, [clears throat] these companies are selling tokens where they provide the entire uh reasoning chain and all that to uh selling automated, you know, white collar work, right? Automated software engineer, send them the request, they give you the result back and there's a bunch of thinking on the back end that they don't show you. The ability to distill out of American models into Chinese models will be harder.
    spoken · machine transcript hear it at 1:11:44 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 14th of 28 in this recording

    AI and jobsAmerica

    Interesting

  29. Our reading

    You do not have to believe in AGI to believe in the timelines on which the United States wins the semiconductor race.

    Their wordsBut I don't know like I don't know what fast timelines means, right? Like I I like don't think you have to believe in AGI to have the timelines where the US wins.
    spoken · machine transcript hear it at 1:15:51 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 15th of 28 in this recording

    AGIAmericasemiconductors

    Interesting

  31. Our reading

    Nobody wants a slow model, which is why labs will not trade inference speed for cheaper memory even though they easily could.

    Their wordsThey could release claw slow mode and have an increase in tokens per dollar by a significant amount. Um they could probably like reduce the price of Opus 46 by you know 4x 5x and reduce the speed by another by maybe just like 2x like the curve on inference throughput versus speed is there already just on hm um and yet they don't um because no one actually wants to use a slow model
    spoken · machine transcript hear it at 1:17:38 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 16th of 28 in this recording

    Interesting

  33. Our reading

    What an accelerator needs from memory is bandwidth per unit of chip edge, and on that measure ordinary DRAM is an order of magnitude behind HBM.

    Their wordsand even if you take a generous interpretation of 128 * 8 gig transfers, you're at 128 gigabytes a second for the same shoreline versus 2 and a half terabytes a second. There's a there's an order of magnitude difference in bandwidth per edge area.
    spoken · machine transcript hear it at 1:22:09 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 17th of 28 in this recording

    Interesting

  35. Our reading

    The memory crunch will make ordinary phones and computers worse year on year, and that is why the public will come to hate AI more.

    Their wordsUm DRAM gets released goes to AI chips who are willing to do longer term contracts, willing to pay higher margins, etc., etc. because at the end of the day, the margin that they extract is much larger from the end user or whatever. Um, and so this this this probably leads to like people hating AI even more, right?
    spoken · machine transcript hear it at 1:26:55 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 18th of 28 in this recording

    Interesting

  37. Our reading

    The reason we cannot simply make more memory is that there is nowhere to put the tools, and a fab takes two years to build.

    Their wordsUh, Micron bought a fab from a company in Taiwan that makes lagging edge chips, right? Um, Heinix and Samsung are doing, you know, some pretty crazy things to try and expand capacity at their existing fabs, uh, that also have like very large knock-on effects in the economy. And so, hey, why can't we build more capacity is like there's nowhere to put the tools, right?
    spoken · machine transcript hear it at 1:30:38 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 19th of 28 in this recording

    Interesting

  39. Our reading

    Elon Musk can probably build the cleanroom for a million-wafer fab, but he cannot develop the process technology quickly, because that knowledge is accumulated rather than invented.

    Their wordsbut I think he can build the uh clean room. It'll take a year or two. Maybe initially it won't be super fast, but then over time you'll get faster and faster at it. But then the really complex part is actually developing the process technology and building wafers. And I don't think he can develop that uh quickly. I think that has a lot of built-up knowledge.
    spoken · machine transcript hear it at 1:35:12 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 20th of 28 in this recording

    Interesting

  41. Our reading

    About a fifth of the US grid could be freed for data centres just by covering the yearly peak with batteries or peaker plants.

    Their wordsThen all of a sudden, you've unlocked 20% of the US grid for data centers because most of the times that capacity is sitting idle and it's really only there for that peak, right? Which is a day or two, right?
    spoken · machine transcript hear it at 1:46:51 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 21st of 28 in this recording

    data centersbatteries

    Interesting

  43. Our reading

    Labour is a huge constraint on the AI build-out, and the way round it is to build data centres as modules in Asian factories rather than on site.

    Their wordsum humanoid robots maybe start to or robotics at least start to but the main factor is going to be for reducing the number of people is modularizing things and making them in factories in Asia
    spoken · machine transcript hear it at 1:52:02 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 22nd of 28 in this recording

    data centers

    Interesting

  45. Our reading

    Getting a data centre permitted in the empty middle of America is far easier than putting one in orbit.

    Their wordsAnd so I think people are figuring out how to build these things and permitting like I I just like ultimately like permitting and red tape in middle of nowhere Texas or middle of nowhere Wyoming or middle of nowhere like New Mexico is probably a hell of a lot easier than sending stuff into space
    spoken · machine transcript hear it at 1:57:13 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 23rd of 28 in this recording

    data centersAmericapermitting

    Interesting

  47. Our reading

    Space data centres solve the wrong problem, because the contended resource is chips rather than energy, and that will not change this decade.

    Their wordsSo, space data centers effectively are not li limited by, you know, hey, we have this energy advantage. It's actually just limited by the same contended resource. We can only make 200 gawatts of chips a year by the end of the decade.
    spoken · machine transcript hear it at 2:02:13 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 24th of 28 in this recording

    data centersenergy

    Interesting

  49. Our reading

    A lab should spend most of its compute on research rather than on building the next model, because research is where the tenfold yearly efficiency gains come from.

    Their wordswell the model the compute efficiency gains you get from research are so large you actually want most of your compute to go to research not to development because you know all these researchers are generating new ideas trying them out testing them and continuing to march along this and push the prao optimal curve of scaling laws further and further and further
    spoken · machine transcript hear it at 2:10:54 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 25th of 28 in this recording

    scaling laws

    Interesting

  51. Our reading

    TSMC will never throw Apple off its leading node; Apple will simply stop mattering enough for TSMC to cater to it.

    Their wordsand so I don't think TSMC would kick out Apple. I think Apple will become a smaller and smaller and smaller percentage of TSMC's revenue and therefore be less relevant for TSMC to cater to their demands.
    spoken · machine transcript hear it at 2:19:46 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 26th of 28 in this recording

    semiconductorsApple

    Interesting

  53. Our reading

    Huawei, if it had access to TSMC, would arguably be a better chip company than Nvidia.

    Their wordsAnd Huawei has a bigger pool in China. It's very arguable that Huawei, if they had TSMC, would be better than Nvidia.
    spoken · machine transcript hear it at 2:23:30 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 27th of 28 in this recording

    Chinasemiconductors

    Interesting

  55. Our reading

    Evacuating TSMC's engineers and destroying its fabs would leave China with the strongest semiconductor supply chain in the world.

    Their wordsUm, just shipping out all the engineers and blowing up the fabs means China has a stronger semiconductor supply chain than the rest of the world, right?
    spoken · machine transcript hear it at 2:29:35 · all korrents from this recording

    ↗Dylan Patel — The single biggest bottleneck to scaling AI computeyoutube.com 28th of 28 in this recording

    Chinasemiconductors

    Interesting
  56. 13 months earlier

  57. Our reading

    OpenAI's 2022 bet on GPT-4 was a larger gamble than any training run since: all of its compute, for months, on an architecture nobody at Google believed in.

    Their wordsI think it's even more impressive what OpenAI did in 2022. At the time, no one believed in mixture of experts models at Google who had all the researchers. OpenAI had such little compute and they devoted all of their compute for many months, all of it, 100% for many months to GPT-4 with a brand-new architecture with no belief that, "Hey, let me spend a couple of hundred million dollars, which is all of the money I have on this model." That is truly YOLO.
    spoken · official transcript hear it at 0:50:25 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 4th of 44 in this recording

    OpenAIGoogle

    Interesting

  59. Our reading

    Training a frontier model achieves almost nothing by itself; all the economic and military power comes from deploying it, and deployment is what consumes the compute.

    Their wordsTo some extent, training a model does effectively nothing. They have a model. The thing that Dario is sort of speaking to is the implementation of that model, once trained to then create huge economic growth, huge increases in military capabilities, huge increases in productivity of people, betterment of lives.
    spoken · official transcript hear it at 1:03:30 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 7th of 44 in this recording

    Interesting

  61. Our reading

    If AI does not transform the economy within the next five to ten years, the export controls will have guaranteed that China wins the long run.

    Their wordsif you believe we're in this sort of stage of economic growth and change that we've been in for the last 20 years, the export controls are absolutely guaranteeing that China will win long-term.
    spoken · official transcript hear it at 1:18:56 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 10th of 44 in this recording

    China

    Interesting

  63. Our reading

    The only thing stopping China from building the largest data centre in the world is access to chips; the power and the industrial capacity are already there.

    Their wordsChina, if they wanted to build the largest data center in the world, if they had access to the chips, could. So it's just a question of when, not if.
    spoken · official transcript hear it at 1:20:02 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 11th of 44 in this recording

    data centersChina

    Interesting

  65. Our reading

    Building fabs in America does not reduce dependence on Taiwan: without Hsinchu, TSMC's Arizona plant would stop producing within a couple of years.

    Their wordsArizona is a paperweight. If Hsinchu disappeared off the face of the planet, within a year, couple years, Arizona would stop producing too.
    spoken · official transcript hear it at 1:43:24 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 12th of 44 in this recording

    Americasemiconductors

    Interesting

  67. Our reading

    Export controls slowed China at the leading edge and accelerated it everywhere else, because being locked out taught China to in-source the trailing edge.

    Their wordsSo there is an angle of, the US' actions, from the angle of the expert controls, have been so inflammatory at slowing down China's progress on the leading edge that they've turned around and have accelerated their progress elsewhere because they know that this is so important.
    spoken · official transcript hear it at 1:46:59 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 13th of 44 in this recording

    China

    Interesting

  69. Our reading

    Completely in-sourcing semiconductor manufacturing to the United States is possible, and it would take a decade and a trillion dollars.

    Their wordsAnd so going back, can the US build it here? Yes, but it's going to take a ton of money. I truly think to revolutionize and completely in-source semiconductors would take a decade and a trillion dollars.
    spoken · official transcript hear it at 1:47:20 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 14th of 44 in this recording

    Americasemiconductors

    Interesting

  71. Our reading

    The world has been at its most peaceful precisely when a single power was hegemonic, which is why the arrival of a second AI superpower is dangerous.

    Their wordsIt's an objective fact that the world has been the most peaceful it's ever been when there are global hegemons, or regional hegemons in historical context. The Mediterranean was the most peaceful ever when the Romans were there.
    spoken · official transcript hear it at 1:52:35 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 15th of 44 in this recording

    Interesting

  73. Our reading

    American export controls police floating-point operations while memory and interconnect matter just as much, and reasoning models make that the wrong axis to control.

    Their wordsFLOP is the vector that the government has cared about historically, but the other two vectors are arguably just as important. And especially when we come to this new paradigm, which the world is only just learning about over the last six months: reasoning.
    spoken · official transcript hear it at 1:58:50 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 16th of 44 in this recording

    governmentAmerica

    Interesting

  75. Our reading

    Most of the price gap between o1 and R1 is margin rather than efficiency: OpenAI's inference gross margins are north of 75% because nobody else could serve the capability.

    Their wordsOpenAI has a fantastic margin. When they're doing inference, their gross margins are north of 75%. So that's a four to five X factor right there of the cost difference, is that OpenAI is just making crazy amounts of money because they're the only one with the capability.
    spoken · official transcript hear it at 2:13:23 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 17th of 44 in this recording

    OpenAI

    Interesting

  77. Our reading

    Superhuman persuasion will arrive before superhuman intelligence, which means whoever builds a model can embed their ideals in it long before anyone reaches AGI.

    Their wordsThere's this very good quote from Sam Altman who... He can be a hyperbeast sometimes, but one of the things he said, and I think I agree, is that superhuman persuasion will happen before superhuman intelligence, right? And if that's the case, then these things before we get this AGI ASI stuff, we can embed superhuman persuasion towards our ideal or whatever the ideal of the model maker is, right?
    spoken · official transcript hear it at 2:27:58 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 20th of 44 in this recording

    AGIOpenAI

    Interesting

  79. Our reading

    The aha moment for reasoning models will come from computer use and robotics rather than from scientific discovery, because those are the infinitely verifiable playgrounds.

    Their wordsI think it's actually probably simpler than that. It's probably something related to computer use or robotics rather than science discovery.
    spoken · official transcript hear it at 2:48:43 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 24th of 44 in this recording

    Interesting

  81. Our reading

    We will have AGI-grade intelligence well before it reaches the economy, because what limits deployment is the cost of running it rather than the capability itself.

    Their wordsThe important thing about, hey, is cost a limiting factor here? My view is that we'll have really awesome intelligence, like AGI, before we have it permeate throughout the economy.
    spoken · official transcript hear it at 3:11:27 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 25th of 44 in this recording

    AGI

    Interesting

  83. Our reading

    Cheaper models raise the demand for compute rather than lowering it: H100 rental prices went up in the weeks after DeepSeek launched.

    Their wordsBut the funniest thing I think that comes out of this is Jevons paradox is true. AWS pricing for H100s has gone up over the last couple of weeks, since a little bit after Christmas, since V3 was launched, AWS H100 pricing has gone up.
    spoken · official transcript hear it at 3:17:11 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 26th of 44 in this recording

    Interesting

  85. Our reading

    The largest hole in America's GPU export controls was never physical smuggling but renting: ByteDance runs over 500,000 GPUs it does not own.

    Their wordsOne is ByteDance, arguably is the largest smuggler of GPUs for China. China's not supposed to have GPUs. ByteDance has over 500,000 GPUs. Why? Because they're all rented from companies around the world.
    spoken · official transcript hear it at 3:19:47 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 28th of 44 in this recording

    ChinaAmericasemiconductors

    Interesting

  87. Our reading

    The copyright problem in AI training already has a legal solution nobody is taking: train the models in Japan, where copyright does not apply to training data.

    Their wordsSo, Japan has a law which you're allowed to train on any training data and copyrights don't apply if you want to train a model, A. B, Japan has 9 gigawatts of curtailed nuclear power. C, Japan is allowed under the AI diffusion rule to import as many GPUs as they'd like.
    spoken · official transcript hear it at 3:32:10 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 31st of 44 in this recording

    energysemiconductorsnuclear power

    Interesting

  89. Our reading

    The bottleneck on America's AI buildout is transmitting power, not generating it: in parts of the US, moving electricity costs more than making it.

    Their wordsInteresting thing is certain regions of the US transmitting power cost more than actually generating it because the grid is so slow to build. And the demand for power, and the ability to build power, and re-ramping on a natural gas plant or even a coal plant is easy enough to do, but transmitting the power's really hard.
    spoken · official transcript hear it at 3:45:47 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 33rd of 44 in this recording

    energyAmerica

    Interesting

  91. Our reading

    Google will never become NVIDIA's competitor, because it has never had the DNA to sell a product and its TPU, cloud, DeepMind and search teams serve different customers.

    Their wordsBut Google has never had that DNA of like, "This is a product we should sell." The Google Cloud, which is a separate organization from the TPU team, which is a separate organization from the DeepMind team, which is a separate organization from the Search team. There's a lot of bureaucracy here.
    spoken · official transcript hear it at 4:04:27 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 34th of 44 in this recording

    Googlesemiconductors

    Interesting

  93. Our reading

    AMD's hardware is in many ways better than NVIDIA's and it does not matter, because the software gulf is too large and AMD has never resourced closing it.

    Their wordsAnd they're decent, their hardware is better in many ways than in NVIDIA's. The problem is their software is really bad and I think they're getting better, right? They're getting better, faster, but the gulf is so large and they don't spend enough resources on it or haven't historically, right?
    spoken · official transcript hear it at 4:08:39 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 35th of 44 in this recording

    semiconductors

    Interesting

  95. Our reading

    Software engineering agents will work before any other kind of agent, because code is the one domain where the machine can check its own answer.

    Their wordsBut really the software engineering agents I think can be done faster sooner than any other agent because it is a verifiable domain. You can always unit test or compile, and there's many different regions of it can inspect the whole code base at once, which no engineer really can.
    spoken · official transcript hear it at 4:31:34 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 38th of 44 in this recording

    Interesting

  97. Our reading

    When writing software becomes cheap, platform SaaS dies: companies stop buying Salesforce and start building their own business logic, as China already does.

    Their wordsBut what happens when every company can just invent their own business logic really cheaply and quickly? You stop using platform SaaS, you start building custom tailored solutions, you change them really quickly.
    spoken · official transcript hear it at 4:32:38 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 39th of 44 in this recording

    SaaSChina

    Interesting

  99. Our reading

    The danger from powerful AI is not that it kills everyone but techno-fascism: a few thousand people ruling whoever is left and the economy around them.

    Their wordsit won't be one person rule them all, but it will be, the thing I worry about is it'll be few people, hundreds, thousands, tens of thousands, maybe millions of people rule whoever's left and the economy around it.
    spoken · official transcript hear it at 5:03:38 · all korrents from this recording

    ↗DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459youtube.com 44th of 44 in this recording

    Interesting