Skip to content

Instantly share code, notes, and snippets.

@coolk8

coolk8/yt_transcript.html

Last active June 15, 2026 17:12
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save coolk8/162be70c69907d4ebe91bf35c0b7df52 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save coolk8/162be70c69907d4ebe91bf35c0b7df52 to your computer and use it in GitHub Desktop.
Download ZIP
Stanford MS&E435 — Economics of the AI Supercycle (Transcript)
Raw
yt_transcript.html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Stanford MS&E435 — Economics of the AI Supercycle</title>
<style>
:root {
--bg: #ffffff;
--card: #f6f8fa;
--border: #d0d7de;
--text: #1f2328;
--muted: #656d76;
--accent: #0969da;
--ts-bg: #eaeef2;
--hover: #f0f4f8;
--highlight: #ddf4ff;
}
* { box-sizing: border-box; margin: 0; padding: 0; }
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Georgia, serif;
background: var(--bg);
color: var(--text);
line-height: 1.9;
-webkit-font-smoothing: antialiased;
}
header {
position: sticky; top: 0; z-index: 100;
background: var(--bg);
border-bottom: 1px solid var(--border);
padding: 20px 0;
}
.header-inner {
max-width: 800px; margin: 0 auto; padding: 0 24px;
}
h1 {
font-size: 1.35rem;
font-weight: 650;
line-height: 1.4;
margin-bottom: 6px;
}
.meta {
color: var(--muted);
font-size: 0.85rem;
font-family: 'SF Mono', Consolas, monospace;
}
.meta a { color: var(--accent); text-decoration: none; }
.meta a:hover { text-decoration: underline; }
.actions {
margin-top: 14px;
display: flex; gap: 10px; flex-wrap: wrap;
}
.btn {
padding: 6px 14px;
border-radius: 20px;
background: var(--card);
border: 1px solid var(--border);
color: var(--text);
font-size: 0.8rem;
cursor: pointer;
transition: all 0.2s;
}
.btn:hover { border-color: var(--accent); color: var(--accent); }
main {
max-width: 800px;
margin: 0 auto;
padding: 32px 24px 120px;
}
.para {
display: flex;
gap: 16px;
padding: 10px 14px;
border-radius: 8px;
margin-bottom: 4px;
transition: background 0.15s;
cursor: default;
}
.para:hover { background: var(--hover); }
.para.highlight { background: var(--highlight); }
.ts {
flex-shrink: 0;
width: 52px;
font-family: 'SF Mono', Consolas, monospace;
font-size: 0.78rem;
color: var(--accent);
padding-top: 3px;
text-decoration: none;
opacity: 0.7;
}
.para:hover .ts { opacity: 1; }
.txt {
flex: 1;
font-size: 1.02rem;
}
.progress {
position: fixed; top: 0; left: 0; height: 3px;
background: var(--accent);
width: 0%;
z-index: 200;
transition: width 0.1s;
}
@media (max-width: 600px) {
.ts { width: 42px; font-size: 0.7rem; }
.txt { font-size: 0.95rem; }
h1 { font-size: 1.1rem; }
}
</style>
</head>
<body>
<div class="progress" id="progress"></div>
<header>
<div class="header-inner">
<h1>Stanford MS&amp;E435 — Economics of the AI Supercycle</h1>
<div class="meta">
Spring 2026 · Applications, Applied AI · 49 min · Guest: <strong style="color:var(--text)">Tuhin Srivastava</strong> (Founder &amp; CEO, Baseten)<br>
▶ <a href="https://www.youtube.com/watch?v=Qh7Oxvo5sJI" target="_blank">youtube.com/watch?v=Qh7Oxvo5sJI</a>
</div>
<div class="actions">
<button class="btn" onclick="window.print()">📄 Print / PDF</button>
<button class="btn" onclick="document.body.style.fontSize = parseInt(getComputedStyle(document.body).fontSize)+1+'px'">A+ Larger</button>
<button class="btn" onclick="document.body.style.fontSize = parseInt(getComputedStyle(document.body).fontSize)-1+'px'">A− Smaller</button>
</div>
</div>
</header>
<main>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">00:09</span>
<span class="txt">inference is about to go up a billionx. Not a thousandx, not a million x, a billion x. And the guy who&#x27;s making it go up a billionx is here with us today to thank you so much for joining us. Please join us. Thank you. Tuhin, founder and CEO of Bay 10. You&#x27;ve had a long long scenic journey to the to the day here today. Lots of windy turns. Tell us about it. You know, a lot of students in the class who want to be entrepreneurs. Um I think we&#x27;ll find inspiration in in in the journey.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">00:48</span>
<span class="txt">Awesome. Thank you for having me. It&#x27;s really nice to meet you all. Um yeah, my name is Tin. I&#x27;m the CEO, one of the co-founders of Ban. um based since about 7 years old. Um I&#x27;ve been working in technology since 2012. Um I&#x27;m originally from Sydney, Australia. I came here for uni. Um actually started my career in finance. So I um I moved home after uni um to work at a Australian investment bank called McQuary. um in my in my first day over there they said you want to go and move to New York and work on privatizing toll roads and airports. So I moved to New York and I and I worked in privatizing toll roads and and airports and um about 2 years in I was bored out of my mind and I decided that I should go um back to engineering which I&#x27;d studied in undergrad and I moved to Boston um to work on doing some research for machine learning. This is 2012 or 2013. No, 2011, 2012. Um, and what we were doing is we&#x27;re using traditional machine learning techniques to track the prognosis of neuromuscular disease.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">01:56</span>
<span class="txt">Um, I did that for about a year and a half.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">01:58</span>
<span class="txt">Um, put out some papers. Um, and really quickly realized that if I was going to do machine learning, even back then, that I probably should be in um, San Francisco. So, I moved to San Francisco and got involved in early stage technology as an engineer. Um, I think probably the more important piece is I kind of fell in love with early stage technology and small teams and and building products and</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">02:23</span>
<span class="txt">when no one really cares. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">02:24</span>
<span class="txt">Um did that for a few years, started a couple of companies starting 2015. They went nowhere, but you know, I had the bug. And so, um in 2019, me and my two co-founders, me and my two co co-founders, uh a guy called Phil and a guy called Amir, who I&#x27;d worked with, um for the better part of 10 to 15 years. Is that is that a me thing? Oh, they&#x27;re just straining up your water.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">02:54</span>
<span class="txt">Okay, cool. Um the</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">02:56</span>
<span class="txt">some more energy.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">02:57</span>
<span class="txt">Yeah. So they um you know I I know these guys for 10 or 15 years and we we we had this idea that machine learning was going to be pretty big and um let&#x27;s build a infrastructure business alongside it um to maybe capture the index of it. Obviously machine learning was a lot bigger than we expected and it came a lot faster and so what we&#x27;ve been doing for the last four years is production inference</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">03:18</span>
<span class="txt">um powering the fastest growing AI companies in the world.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">03:21</span>
<span class="txt">Yeah. you have a salivating list of customers. Um, actually you had a post I uh I copied your post here which is these are all the founders that uh based has a good opportunity to work with. You might know some of these faces. Some of them have been here in class. This is uh probably the best index of AI founders</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">03:44</span>
<span class="txt">who work with base 10</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">03:46</span>
<span class="txt">and so maybe maybe break it down for us. What pick an example or two of of of faces up here when they work with B 10? What does that mean that they work with B 10? What&#x27;s an example of their work with us and what does that enable for their business and their customers?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">04:02</span>
<span class="txt">Yeah, may maybe I&#x27;ll pick two fun ones. I&#x27;ll pick T</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">04:05</span>
<span class="txt">who&#x27;s on top. I&#x27;ll pick um T actually Stanford. You know, you know T very well. And I&#x27;ll pick Chev. Um, and so they they run very different businesses. So D runs a company called Whisper Flow. I don&#x27;t know how many people here know Whisper Flow.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">04:24</span>
<span class="txt">How many heard of Whisper Flow?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">04:26</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">04:26</span>
<span class="txt">Awesome. Awesome. Um, that&#x27; make him very happy. Um, the um, so Whisper Flow is a is basically a text keyboard uh, speech speech to keyboard uh, speechto text app um, for voice typing more or less. you know, they they run a lot of custom models that they&#x27;ve built in house or like modified in house in a very particular way um to make that experience possible. And so um what we do for them is we run all the optimizations, we run all the infrastructure so that you know that latency from the time that you talk to when text shows up um is as quick as possible. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">05:05</span>
<span class="txt">Um there&#x27;s language models, there&#x27;s audio models, there&#x27;s um there&#x27;s actually I think three or four language models in the middle there and two audio models to make that happen and all of them run on base 10.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">05:18</span>
<span class="txt">Um Chev who&#x27;s also a good friend of mine is the CEO of a company called a bridge. Um, a bridge is a healthcare um, it&#x27;s an ambient scribe um, used by used in almost every healthare system in the US that&#x27;s deeply integrated with EMRs. Um they run about 20 different models. Um everything from um obviously speech to text models to go from what is happening in the in the patient um in the operating room or the patient&#x27;s room to um everything that goes to turning that into a clinical note that is deeply integrated with EMR. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">05:58</span>
<span class="txt">Again, you know, over like dozens of models, um, lots of health, lots of requirements from a reliability perspective, from from a speed perspective. Um, and every almost actually I think every single one of those models runs on base 10. Um, and you know, I I think that&#x27;s kind of emblematic of kind of what we want to do. Our thesis is that um, AI is going to be absolutely massive. um inferences the cogs of AI value being delivered.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">06:30</span>
<span class="txt">Um and today about 90% or 95% of spend on inference is going to frontier models and about 5% is going to custom models and we believe that the way that all these folks who are building amazing applications</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">06:44</span>
<span class="txt">um with AI um are going to build profitable viable defensible companies is with custom models and hopefully they all run on base 10. 95% of their spend is going to frontier models.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">06:55</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">06:56</span>
<span class="txt">5% is going to open source models. This</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">06:59</span>
<span class="txt">or post-trained open source models. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:01</span>
<span class="txt">Got it. Actually, before we go there, I have a question for the two businesses that you highlighted. So,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:07</span>
<span class="txt">a bridge or whisper flow both fairly scaled businesses.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:12</span>
<span class="txt">Why would they come work with you? Why not? Typically, if I was a founder, I would go to my cloud provider. I go to AWS, GCP, Azure</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:21</span>
<span class="txt">or one of the NeoClouds that that are the AI clouds, Corviv, Nbius, others.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:27</span>
<span class="txt">Why Ban?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:28</span>
<span class="txt">Yeah, I I think a lot of it comes down to like the three or four core tenants of like the software we&#x27;re building. So</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:34</span>
<span class="txt">there&#x27;s obviously performance. You need these things to work as quickly as possible. [clears throat]</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:39</span>
<span class="txt">If you go to Core, weave, Nebus, AWS, GCP, um you&#x27;re kind of on your own to do those optimizations. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:46</span>
<span class="txt">Um the second piece is reliability and compute. Um you know they&#x27;re coming to us because they need um inference to run across clouds in a very fault tolerant resilient way.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:56</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">07:56</span>
<span class="txt">Um I think you know part of that is being multicloud and having access to compute from multiple sources.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:02</span>
<span class="txt">We kind of unlock multiloud for them. And the third one is like the development developer platform that we that we provide which gives them you know flexibility, security, observability and all those good things um to make that happen. And so</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:16</span>
<span class="txt">what we actually find is a lot of them do go</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:19</span>
<span class="txt">to those folks you talked about first</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:21</span>
<span class="txt">and then realize the pain of standing up that whole inference stack on top of compute</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:25</span>
<span class="txt">and realize they&#x27;ll just be better served coming to base 10.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:29</span>
<span class="txt">Fascinating. I imagine uh now going to the second thing you had said earlier 95% of them or or 95% of tokens are on Frontier 5% on post-trained or or open source of some kind.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">08:41</span>
<span class="txt">Yeah. If I was using an open source model or post- training an open source model, I am branching off the sort of the the the lineage of intelligence that might improve and I run the risk that the next GPT model or the next Opus model, let&#x27;s call it GPT 6 or or Opus 5, will be better than what I just postrained. Why go through the effort of branching out into this specific model uh uh when the</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:11</span>
<span class="txt">no work option</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:13</span>
<span class="txt">might just get there in in in a couple of weeks or a couple of couple of months.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:16</span>
<span class="txt">Yeah. Look, there&#x27;s a number of reasons I&#x27;d say. Um you know, I&#x27;ll give I&#x27;ll give you the the viable reason. I&#x27;ll give you the cynical reason. Um the the the viable reason the viability reason here is that you know you if you think about a company let&#x27;s take if you think about any truly um any company that&#x27;s doing things with these frontier models.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:38</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:38</span>
<span class="txt">Today open source models are about 90 days behind</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:42</span>
<span class="txt">closed source models.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:43</span>
<span class="txt">And the question that a lot</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:44</span>
<span class="txt">give a slide on this. Keep going.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:45</span>
<span class="txt">Yeah. Do I?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:47</span>
<span class="txt">Uh next one. Yeah. open source models about 90 90 days behind um Frontier models and you can run them about 70 to 70 to 90% cheaper.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">09:59</span>
<span class="txt">Wow.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:00</span>
<span class="txt">Than Frontier models.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:01</span>
<span class="txt">70 to 90%</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:02</span>
<span class="txt">cheaper. Yeah. And especially when you and when you add in a bit of you know I think you had Yash in here.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:07</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:07</span>
<span class="txt">Last week who probably told you the post- training narrative of specialized models and especially when you when you factor that in you can really do better, faster, cheaper.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:15</span>
<span class="txt">Mhm. um um for these models and and it&#x27;s really they come to us from like the viability reason when they really go from I have product market fit to I need to become a scale business and figure out how to you know get my gross margins</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:29</span>
<span class="txt">you know up from zero all the way to 40 50 60 70% hopefully um like you want to be</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:35</span>
<span class="txt">the cynical reason I&#x27;d say which is you know</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:38</span>
<span class="txt">um look what do you have that is defensible against a frontier model it&#x27;s probably some some workflow Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:45</span>
<span class="txt">some user signal um that only you have.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:48</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:49</span>
<span class="txt">Um if you keep working with the the Frontier Labs,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">10:54</span>
<span class="txt">to some extent, you know, you&#x27;re going to give them all the data and all that user signal and you know, they got they&#x27;re going to you know, the the who um I was talking to a prominent public office, I won&#x27;t say who who it was. Um he like he likened the Frontier Labs to the East India Company. you know, they they they they show off in India and like, you know, they&#x27;re they&#x27;re they&#x27;re making all these partnerships, but really what they&#x27;re doing is you&#x27;re giving them all the tricks on how</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">11:21</span>
<span class="txt">to rule that and before you know their post- training models kind of at um against those workflows that are sacred to you that only you know. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">11:30</span>
<span class="txt">So to be defensible against this and keep the thing that you know makes you special</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">11:35</span>
<span class="txt">you need int you need to own your intelligence</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">11:38</span>
<span class="txt">to some extent and that&#x27;s why you know they are running they are you know figuring out how they can use open source models and post train them um and really kind of build out that stack themselves to own their intelligence.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">11:50</span>
<span class="txt">Fascinating. Taking this further to it seems that if I had a larger volume of workloads, I am more incentivized to move to</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:03</span>
<span class="txt">open source or or post train model. Um you have a couple of customers that are very scaled customers a bridge, [clears throat] whisper flow, cursor</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:13</span>
<span class="txt">etc.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:14</span>
<span class="txt">They are obviously very large customers of the frontier models. Um is it also the case that the larger or the more um user base you have the more likely you are to adopt this or is it the other way is like the smaller uh guys are more likely to adopt this. I think I think the form I think the larger you are the more existential it becomes</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:34</span>
<span class="txt">right</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:34</span>
<span class="txt">to be able to to need to go towards this</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:37</span>
<span class="txt">because I think that&#x27;s when you like you know the bigger you get</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:40</span>
<span class="txt">the more you you know like I mean you&#x27;ve looked you maybe you can talk about this as well but you you know you look at these businesses</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:48</span>
<span class="txt">and as you get to massive scale</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">12:51</span>
<span class="txt">um if you&#x27;re just trading tokens um you know it starts becoming very very expensive and the more capital you need and I you need that path to profitability. Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:02</span>
<span class="txt">yeah,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:02</span>
<span class="txt">those post training becomes very important there.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:05</span>
<span class="txt">Yeah, I mean the leading coding companies that are not the frontier model companies themselves are, you know, still rumored to be negative gross margin and so I imagine is existential for them to to to be a viable business</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:21</span>
<span class="txt">to shift that token uh uh the token volume towards open source.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:25</span>
<span class="txt">Yeah. Um now is there a performance gap that they notice as they move from let&#x27;s say for example cursor using enthropics models or GPD the latest openi models to moving to a post train model like they did recently there&#x27;s there&#x27;s a lot of hoopla on Twitter about their work with um composer um does the performance take a hit? Is there a way that the workload partitions that it does not or um how do the users experience the the the change in the underlying you know almost the engine of the Ferrari was was was swapped out.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">13:58</span>
<span class="txt">Yeah. Well, we don&#x27;t have any like you know any data like internally, but I think the the like you would hope that it gets better, right? You&#x27;d hope that you know as they post train models to use the signals that they have</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:11</span>
<span class="txt">the experience of the user gets better.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:13</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:13</span>
<span class="txt">And hopefully it also drives you know both performance from a latency perspective and reliability perspective</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:20</span>
<span class="txt">um higher because there&#x27;s a lot more control.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:23</span>
<span class="txt">Got it. Um makes tons of sense. And um what is the what is the business of basen? How do we monetize our users? You know, obviously there&#x27;s this there&#x27;s a whole slew of pricing options right now in AI. You&#x27;ve got outcome based pricing, token based pricing, GPU by the hour pricing and and others that I might not even know about.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:45</span>
<span class="txt">Where do we where do we fit in and and how does that work with your largest customers?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:50</span>
<span class="txt">Yeah, look, right now we mock up compute.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">14:53</span>
<span class="txt">Mhm. um we&#x27;re we&#x27;re re for for the most part. So what that means is that you know you come to use base 10 um you choose you bring your models you you load it up onto the base 10 inference stack</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:07</span>
<span class="txt">um and you basically you basically choose which compute you want to run it on</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:12</span>
<span class="txt">um and a B 10 H100 or B 10 B200 is more expensive than a raw B200. I think what&#x27;s interesting there to think about you know</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:21</span>
<span class="txt">um how much you like it is expensive way to get raw compute but ideally that&#x27;s that&#x27;s the value you&#x27;re getting the software stack</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:30</span>
<span class="txt">um others use token pricing</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:34</span>
<span class="txt">you know we we have some customers and you know we we we are pretty relatively unsophisticated</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:40</span>
<span class="txt">about this today um but you know we&#x27;re moving towards a world especially as we unlock more of those post- trainining workflows</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:46</span>
<span class="txt">within bas itself where</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:48</span>
<span class="txt">you know the narrative changes from how much how much are you paying for computing per token</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">15:54</span>
<span class="txt">and actually the app companies appreciate that a bit more because especially for those coming from using anthropic or open AI</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:01</span>
<span class="txt">um they have a like for like and how much how much cost savings there are</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:06</span>
<span class="txt">fascinating um I might follow up on something you said you said post training workflows</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:11</span>
<span class="txt">um could you um simplify that for us what does a post-training workflow look like</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:18</span>
<span class="txt">start to end. Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:20</span>
<span class="txt">and feel free to take an example if you can. Um, how does it work? How does it work with base 10 or how does it work with you know uh a large customer like cursor or bridge or whisper flow?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:30</span>
<span class="txt">Yeah, like you know we talk a bit about post training there um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:36</span>
<span class="txt">right here. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:38</span>
<span class="txt">So we have training um and so really what you&#x27;re doing as a customer there is you&#x27;re you&#x27;re designing what your utility function is.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:44</span>
<span class="txt">Mhm. and you&#x27;re saying, &quot;Hey, this is the thing that I want to optimize with this given model.&quot;</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:49</span>
<span class="txt">So, you you need to define that. Like, we can&#x27;t help you there. We don&#x27;t know your product. We don&#x27;t know your users. We don&#x27;t know what you&#x27;re optimizing for in your business.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:58</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">16:58</span>
<span class="txt">So, instead, you decide what your what the utility function is.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:03</span>
<span class="txt">Mhm. And then once you know that, start giving us a bunch of data</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:08</span>
<span class="txt">and choose an open source model and we kind of give you all the scaffolding to turn that into a a post-trained model.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:15</span>
<span class="txt">Um, and then once you have that, you know, how does that roll into the inference piece?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:20</span>
<span class="txt">Um, and so we kind of own that entire loop. So you bring your data, you bring your utility function.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:25</span>
<span class="txt">Um, you choose a base model. So you might say maybe just to make it very concrete you might say</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:32</span>
<span class="txt">um let&#x27;s say you&#x27;re building a speech to text model for for for a medical use case</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:38</span>
<span class="txt">it you choose you know the and maybe it is errors or transcription errors the thing you&#x27;re</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:45</span>
<span class="txt">you&#x27;re trying to minimize</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:47</span>
<span class="txt">so that&#x27;s your utility function you choose the model you want to use so let&#x27;s say you could use a Kimmy K25</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:52</span>
<span class="txt">and then you give us a data set</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">17:55</span>
<span class="txt">we have all that scaffolding set to turn that into a very very good post-trained Kimmy K25</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:01</span>
<span class="txt">made for that and then we obviously have all the um integration into inference</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:07</span>
<span class="txt">um predefined and so our customers who are coming to us with that and I think you&#x27;re going to see more and more of this</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:14</span>
<span class="txt">um really um they they come with data and what they know about their about their workflows and they leave with a post-trained um like specialized model running on base Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:28</span>
<span class="txt">They trust you with all that data, the keys to the kingdom.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:32</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:33</span>
<span class="txt">Is that a natural conversation for them to hand over the keys to the kingdom to base 10 or the East India Company?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:39</span>
<span class="txt">Um, you know,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:41</span>
<span class="txt">we&#x27;re the West India Company.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:41</span>
<span class="txt">We&#x27;re the way, you know, we&#x27;re good, you know, we we we are we are we&#x27;re the we&#x27;re the rebellion,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:47</span>
<span class="txt">right?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:48</span>
<span class="txt">We&#x27;re trying to arm the rebellion. Um, so</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">18:50</span>
<span class="txt">is that a natural conversation? I I think it&#x27;s fine. I think most people are okay with that. you know, we we have, you know, I think we have a track record of working with like amazing companies and the brand kind of does a lot of work there, but you know, we also have</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:02</span>
<span class="txt">like incredibly intense security posture internally,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:06</span>
<span class="txt">right?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:06</span>
<span class="txt">And, you know, like even look, we we work with</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:09</span>
<span class="txt">um multiple competitors and we have access to, you know, their models and their data and, you know, we have set up the boundaries internally</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:19</span>
<span class="txt">to make sure that there&#x27;s no leakage between that. Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:22</span>
<span class="txt">and so is it natural? I I I think it&#x27;s not not as bad as you&#x27;d think. You also got to remember that a lot of these companies are trying to work move very very very quickly.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:31</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:31</span>
<span class="txt">And I and I think they&#x27;re more they&#x27;re more interested on can you solve their user problem as opposed to we need to do everything from scratch.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:37</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:38</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:39</span>
<span class="txt">I think they get there&#x27;s a lot of trust they put in base 10, but I think hopefully that&#x27;s ear that&#x27;s earned and um from all the great customers we work with and their friends who are already using us for a lot of these things.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:49</span>
<span class="txt">Makes sense. Um the other big bet you you guys are making implicitly is that open-source will stay on the frontier for at least 3 to 6 months</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">19:59</span>
<span class="txt">behind the frontier. You know the economic model of front open open source frontier labs</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:06</span>
<span class="txt">is</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:08</span>
<span class="txt">TBD</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:08</span>
<span class="txt">TBD I was about to say TBD yeah</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:10</span>
<span class="txt">is not known and you know there&#x27;s a lot of um at least western open source labs like obviously Meta has decided to insource their their their their or not open source their latest models anymore.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:22</span>
<span class="txt">Talk talk about that for a second. What is the business model of open source? What is your best guess of how the world unfolds? uh because that is obviously one of the biggest underlying bets that you&#x27;ve made.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:33</span>
<span class="txt">Yeah. Um look, I I think the two big bets of base 10 are the existence of an application an independent application layer.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:41</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:41</span>
<span class="txt">And will open source be good enough that you can continue to post train most of like America so far has shown that</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:48</span>
<span class="txt">it it is not able to produce the best-in-class at least in the last two years</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:54</span>
<span class="txt">open source models like the best open source models today are coming</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:57</span>
<span class="txt">that&#x27;s wild</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:58</span>
<span class="txt">from China. I I think</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">20:59</span>
<span class="txt">why is that?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:00</span>
<span class="txt">Why can&#x27;t they produce them?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:01</span>
<span class="txt">Why can&#x27;t</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:03</span>
<span class="txt">America produce open source models that perform it?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:05</span>
<span class="txt">I mean, I think this, you know, there&#x27;s all the best research in the world right now working at two companies and, you know, I don&#x27;t think they&#x27;re despite their names maybe suggesting otherwise. Um, they&#x27;re not super motivated</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:17</span>
<span class="txt">um to produce that. Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:19</span>
<span class="txt">I I I think I think um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:22</span>
<span class="txt">and why have the non-American researchers decided to pursue the open source model? I mean I I I think it&#x27;s it&#x27;s to be relevant probably to start with, right? Like you know, we we we think like Moonshot and and Alibaba and um and uh Miniax, you know, there&#x27;s just a market opportunity</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:45</span>
<span class="txt">if all to take the counter position</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:48</span>
<span class="txt">of of of closed</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:49</span>
<span class="txt">of closed models.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:51</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:52</span>
<span class="txt">Um and Meta did that too, right, a few years ago. they they they famously have swung the other way.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:59</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">21:59</span>
<span class="txt">Um, look, we we open source models need to exist.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:03</span>
<span class="txt">Like I think it&#x27;s somewhat of a of a of a a matter of national security. If America does not have</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:11</span>
<span class="txt">good open source models or if the cost of intelligence is 70 or 90% cheaper</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:15</span>
<span class="txt">um in the east than it is the west like that, that that&#x27;s probably not a good outcome. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:19</span>
<span class="txt">Um I I do think you know um you know the the the best American companies are um are actually pretty well aligned around open source. You know Google produced Jamma</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:35</span>
<span class="txt">um Nvidia&#x27;s putting a lot of investment into the Neotron</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:38</span>
<span class="txt">that um family reflection AI</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:41</span>
<span class="txt">um you know hopefully will come out with a good open source model</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:45</span>
<span class="txt">in a few in a few years from now. I think it&#x27;s inevitability though. realistically like you know I if if we&#x27;re in a world in 10 years from now where there&#x27;s no good open source models</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:55</span>
<span class="txt">um you know that&#x27;s probably a very bad thing</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:58</span>
<span class="txt">right</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">22:59</span>
<span class="txt">um for the US but um I think I but I do believe there&#x27;s enough investment happening that it&#x27;s going to happen</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:06</span>
<span class="txt">did you uh I&#x27;m sure you did anthropic put out a big post this morning</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:10</span>
<span class="txt">I I saw that</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:10</span>
<span class="txt">the the the the future of the Chinese and American uh AI relationship of course very beneficial for Anthropic any any thoughts on that? Actually, maybe summarize the post for the class. So, how do you how do you</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:22</span>
<span class="txt">I I haven&#x27;t read I&#x27;ve been back to bed since the morning, but my my read on it was like there&#x27;s two scenarios that America that America&#x27;s at the frontier. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:30</span>
<span class="txt">And the other one is</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:31</span>
<span class="txt">that um China China um is is is neck and neck.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:37</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:38</span>
<span class="txt">More or less. And I think I think and again maybe you&#x27;ve read it. Um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:42</span>
<span class="txt">no, that&#x27;s that&#x27;s exactly it. They basically said, &quot;We either we lead it and you you shut it down or we&#x27;re neck and neck and that&#x27;s a war.&quot;</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:49</span>
<span class="txt">Yeah. And and and their recommendation was let&#x27;s shut it down. So shut down so there&#x27;s not a war.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:53</span>
<span class="txt">That&#x27;s right.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">23:55</span>
<span class="txt">Look, they&#x27;re awesome and like, you know, they do amazing work. I think it&#x27;s, you know, I I genuinely believe in life before you make any points, you should make your biases and your and your incentive functions very clear.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:07</span>
<span class="txt">Um I don&#x27;t think they acknowledge that.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:08</span>
<span class="txt">Yeah. Um, look, I I I I don&#x27;t know what to make of that. I I also just don&#x27;t I just think, you know, we put out a post yesterday about the world of many models. We think intelligence shouldn&#x27;t be owned by two people. I think in in in the history of time when too much power has been concentrated</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:24</span>
<span class="txt">in one or two parties, bad things</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:27</span>
<span class="txt">generally happen. And I also just don&#x27;t believe that thesis to some extent that,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:31</span>
<span class="txt">you know, these two companies with a massive profit motive should be the arbiters of morality for the rest of us. Um the [snorts] yeah [laughter] so so like you know when I think about like do do I think that it is it is concerning that all the best open source models come from China 100%. Like you know and they&#x27;re great you know we we we&#x27;ve spent time with all those teams and they&#x27;re amazing. It&#x27;s not so much that I don&#x27;t trust them. I just think that</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">24:59</span>
<span class="txt">you know America should have great open source as well.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:02</span>
<span class="txt">Yeah. Yeah. Yeah. It seems do do you think the velocity of uh American open source is increasing or decreasing over time in terms of talent and and resources capital GPUs?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:13</span>
<span class="txt">Yeah. Yeah. Oh, I think up until like a year ago was definitely decreasing.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:18</span>
<span class="txt">Okay.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:19</span>
<span class="txt">But I think it&#x27;s I think there&#x27;s uptick</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:21</span>
<span class="txt">now. I look I I think some very very important companies are behind this and I think we&#x27;ll see a lot this year</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:28</span>
<span class="txt">going the other way.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:29</span>
<span class="txt">Yeah. Yeah. Yeah. I think it&#x27;s necessary because otherwise we&#x27;re just going to have two companies left.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:34</span>
<span class="txt">Right.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:35</span>
<span class="txt">There&#x27;ll be no Apple player.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:36</span>
<span class="txt">Right.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:36</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:37</span>
<span class="txt">Right. Right. Right. Right. Um switching gears a little bit too. You were in a very um advantaged position where you see a lot of different hardware</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:45</span>
<span class="txt">that you run inference on. Right.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:47</span>
<span class="txt">Well, actually you run in two layers, a lot of different providers. Actually, I think you had this map um which we which we can talk about for a second,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">25:56</span>
<span class="txt">but I actually meant a layer underneath this, you know.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:00</span>
<span class="txt">Yeah,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:00</span>
<span class="txt">I assume a lot of the inference that you guys are delivering is on Nvidia hardware just because it&#x27;s the most</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:05</span>
<span class="txt">majority the majority. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:06</span>
<span class="txt">Prevalent,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:07</span>
<span class="txt">but then there&#x27;s a long tale, you know, obviously um Tranium announced in the last earnings call they had about $20 billion of revenue run rate. Um TPU, I&#x27;m sure, is is a big number. We don&#x27;t know what it is. They have they haven&#x27;t disclosed it at least. Then there&#x27;s Cerebras from this morning. Um then there&#x27;s everybody else edged, Maddox, posetron, others that we do not know about.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:26</span>
<span class="txt">Uh Dmatrix, Samanova, etc. Talk talk about the heterogenous compute ecosystem. How do you see the world unfolding on the on the hardware side?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:37</span>
<span class="txt">Yeah. Um, look, we we are like I I I think in general diversity like I I can&#x27;t I can&#x27;t sit here and say that I believe in a world of many models and then say, &quot;Oh, only one chip will exist.&quot; That that&#x27;d be hypocritical. Um, but that being said, um, [laughter] we we are we run the majority of our fleet on Nvidia chips. Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">26:59</span>
<span class="txt">like TPUs are very promising. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:01</span>
<span class="txt">Um, obviously all the new age the neo chips if you want to call them that</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:06</span>
<span class="txt">are very promising.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:08</span>
<span class="txt">Um, I don&#x27;t I don&#x27;t you know we haven&#x27;t seen anyone really use trrenium at scale. I think the advantage you have</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:15</span>
<span class="txt">with Nvidia is just obviously a fleshed out supply chain.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:18</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:18</span>
<span class="txt">A fleshed out supply chain.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:20</span>
<span class="txt">Um, you have and you know</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:23</span>
<span class="txt">with a very very low cost of capital and a very strong relationship with TSMC.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:27</span>
<span class="txt">Yeah. Um I I think that the the idea that anyone else can really compete with that today</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:33</span>
<span class="txt">is somewhat fanciful.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:34</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:35</span>
<span class="txt">Um you know to go one layer deeper like one thing we rely very very heavily on is CUDA</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:41</span>
<span class="txt">and the developer ecosystem.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:43</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:44</span>
<span class="txt">Um around CUDA</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">27:46</span>
<span class="txt">um there&#x27;s nothing like CUDA like CUDA is insane. Um I think you know these you know these new architectures of breaking out you know inference has like two core parts to it which is one is prefill one&#x27;s decode.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:01</span>
<span class="txt">Um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:02</span>
<span class="txt">you know this idea that all this will also just run on one chip.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:07</span>
<span class="txt">Mhm. which is historically what&#x27;s happened</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:09</span>
<span class="txt">probably isn&#x27;t isn&#x27;t isn&#x27;t</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:11</span>
<span class="txt">the the end state but you know Nvidia quite gro</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:15</span>
<span class="txt">um a lot of these um these neo these new chips you&#x27;re talking about is try to separate out</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:21</span>
<span class="txt">these concerns of</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:22</span>
<span class="txt">decode and like you know</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:24</span>
<span class="txt">where you do the memory bound stuff all on the GPUs and you do</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:28</span>
<span class="txt">the computer stuff on</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:30</span>
<span class="txt">um on um on a different chip so like the pre separating out so I think that is the world the world will go like heterogeneous architectures</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:40</span>
<span class="txt">for for these things. I think right now for us as a company we&#x27;re just trying to move as fast as possible and our customers are trying to move as fast as possible</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:48</span>
<span class="txt">and CUDA</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:49</span>
<span class="txt">availability of Nvidia GPUs um the ability to use stuff like TRTLM which is clearly just built for Nvidia chips.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:57</span>
<span class="txt">What is that?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">28:58</span>
<span class="txt">Um it&#x27;s a runtime it&#x27;s open source runtime that runs on top of Nvidia chips. It&#x27;s developed by Nvidia. Even Vlm and SG lang which are two other open source</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:06</span>
<span class="txt">um frameworks are you know native</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:10</span>
<span class="txt">to to Nvidia and so the ability to just move very very fast</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:16</span>
<span class="txt">um is is the thing we are optimizing for and that&#x27;s what Nvidia&#x27;s really good at today.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:21</span>
<span class="txt">Gotcha. Okay. So so a fairly</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:26</span>
<span class="txt">one-sided play there for now.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:28</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:28</span>
<span class="txt">With hopes for optimistic uh future. Um maybe a layer above that too and you know talk about this map a little bit. You&#x27;re working with so many different</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:37</span>
<span class="txt">clouds. Yeah. [snorts]</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:37</span>
<span class="txt">Clouds. Uh some of these have even announced an inference platform of their own.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:43</span>
<span class="txt">There&#x27;s no reason for others to not do so.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:45</span>
<span class="txt">Uh maybe talk about that like it seems to us through the class a lot of speakers have told us that compute is very scarce.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:51</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:52</span>
<span class="txt">Um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:53</span>
<span class="txt">probably true. How are you getting your compute and what&#x27;s the strategy going forward?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">29:57</span>
<span class="txt">Yeah. Look, we we work on top of 18 clouds. I think it&#x27;s 20 now actually.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:02</span>
<span class="txt">Um we and we have 87 different clusters where we&#x27;re we&#x27;re stitching compute together. This is the core um I&#x27;d say piece of technology that we have is that we have the ability to take a bunch of different GPUs from different places and stitch it together in one place</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:16</span>
<span class="txt">and make GPU fungeible</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:19</span>
<span class="txt">in that way. Um the reason we do this is twofold. one it&#x27;s it&#x27;s access</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:24</span>
<span class="txt">it gives you know is very very hard to find GPUs and</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:28</span>
<span class="txt">and if you add a constraint around which clouds they operate in that makes it even harder</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:33</span>
<span class="txt">and so what we do is that we take GPUs from anywhere we can stitch them together and abstract that away</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:39</span>
<span class="txt">from the customers</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:40</span>
<span class="txt">um as to the first thing and that&#x27;s a core part of the strategy and we&#x27;ll continue to do that we will rent and we will own um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:47</span>
<span class="txt">we will rent versus own</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:49</span>
<span class="txt">own yeah for the for the most part but we will also build our own ownership.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:53</span>
<span class="txt">Okay, we&#x27;ll come to this in a bit. Yeah,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">30:56</span>
<span class="txt">but in terms of why you know them going after and a lot of those platforms having their own inference</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:03</span>
<span class="txt">things,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:04</span>
<span class="txt">it makes sense. You know, there&#x27;s a lot of value in that stack</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:08</span>
<span class="txt">in that inference stack that sits on top of GPUs.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:11</span>
<span class="txt">Um,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:12</span>
<span class="txt">you know, we we we welcome it. You know, we will partner with our competitors. We&#x27;re fine with that. We we we stand behind the value that we we are we are providing.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:21</span>
<span class="txt">Mhm. But it makes sense. You know, inference is very very sticky.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:25</span>
<span class="txt">Got it.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:25</span>
<span class="txt">I think truly like a lot of people think that inference is a commodity.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:28</span>
<span class="txt">I mean, you know, you you&#x27;ve obviously studied these markets.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:31</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:31</span>
<span class="txt">Um it it it just resembles the way that people used to buy databases.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:36</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:36</span>
<span class="txt">To some extent. You you choose once and you kind of just grow there.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:39</span>
<span class="txt">Yeah. Well, it&#x27;s also, you know, just was was with one of the uh great founders you had listed on here this morning. They said that inference is so sticky for us because ultimately this is the product that we deliver to our customers. Yeah,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">31:51</span>
<span class="txt">we don&#x27;t want any disruption in [clears throat] the core service that our customers are consuming. A disruption in that is actually you know sometimes even the biggest cost line item even more than their cloud spend.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:02</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:03</span>
<span class="txt">Is is the inference the intelligence.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:04</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:05</span>
<span class="txt">So I imagine it&#x27;s a price as large will be fiercely competed for</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:09</span>
<span class="txt">100%. And and I think you it is once you&#x27;re in token path like if your inference is down your product&#x27;s down.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:15</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:16</span>
<span class="txt">You don&#x27;t want to mess with that.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:17</span>
<span class="txt">Yeah. Um I put a bookmark on renting versus owning.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:22</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:22</span>
<span class="txt">Can you talk about that? You know, we&#x27;ve had um Chase from Cruso here who who who talked us through the whole economics of building a</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:29</span>
<span class="txt">data center of owning the whole thing.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:31</span>
<span class="txt">You have taken the other side of it which is to rent the</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:35</span>
<span class="txt">till now. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:36</span>
<span class="txt">For now. Okay. For now.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:37</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:38</span>
<span class="txt">Talk more about that. Uh it seems like an important decision uh and very different shape of the business.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:43</span>
<span class="txt">Yeah. Well, you know, for our our thesis was to some extent what what is the sticky part of the stack, right?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:48</span>
<span class="txt">And so our, you know, we would say that software is the sticky part of the stack here.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:53</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:53</span>
<span class="txt">Um I think what what you know, a lot of these clouds who are competing us with us would say is that access to GPUs</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">32:59</span>
<span class="txt">is is the sticky part of the stack.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:01</span>
<span class="txt">Mh.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:02</span>
<span class="txt">And you know, once you have that, the software part is easy. I think it&#x27;s probably arrogant both ways.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:06</span>
<span class="txt">You know, I think they&#x27;re slightly arrogant to think the software piece they could do they could do it very easily. I think we are 100% arrogant to be like oh we can do that too.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:13</span>
<span class="txt">Um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:14</span>
<span class="txt">to some extent like we we did that for speed and we did that because we had no we had no business building data centers.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:22</span>
<span class="txt">Um for us going forward though is becoming clear that you know what I what I would say is that no matter how much people tell you that there&#x27;s a supply problem</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:36</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:37</span>
<span class="txt">it is 10 times worse. [laughter] Uh, you know, I I think you saw in the GC G in the Google earnings, they had the GCP. Did you see this chart?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:45</span>
<span class="txt">I think they&#x27;ve got like a 10 10x backlog or something.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:48</span>
<span class="txt">Um, if you go out right now saying you want a,000 GPUs, truly you probably saying people are talking about Q2 next year.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:55</span>
<span class="txt">Q2 next year. So 12 months out, maybe 15 months out.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">33:58</span>
<span class="txt">15 months out. Um, we we have a um we have a cluster, a small cluster relative now. At the time it was big. um in one of in one of these clouds um that of B200&#x27;s Blackwell chips B200&#x27;s like great chip um it&#x27;s a bit old but still it&#x27;s amazing chip um our our unit price right now is 263 an hour</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:22</span>
<span class="txt">Mhm. for that and like it doesn&#x27;t matter what that is because I&#x27;ll just the relative piece of it matters more. Um that&#x27;s up for renewal in October</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:32</span>
<span class="txt">and they came to us already in May and said 510 is a new price.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:37</span>
<span class="txt">Wow.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:38</span>
<span class="txt">For next year</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:39</span>
<span class="txt">double</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:39</span>
<span class="txt">double. Um this is fine.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:42</span>
<span class="txt">Are you going to take it?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">34:43</span>
<span class="txt">No, absolutely not. The um the um that&#x27;s that&#x27;s egregious. Um the um and we have a bunch of other clouds that we can work with here. Um I think though like what is becoming very important is that um there&#x27;s there&#x27;s there was a slide that was doing the rounds in Twitter. There was a slide that was doing around Twitter of um of Nolan Brown</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:03</span>
<span class="txt">from OpenAI</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:04</span>
<span class="txt">who I think is you know close to the altimate folks as well. Um he came up with the 03 model and and the side basically said that in access to compute is the strategic advantage for inference. that&#x27;s becoming clear for us is that look we we we run in in you know right now I think the base 10 um inference service across all our customers is bigger than the openi API is bigger than Gemini</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:27</span>
<span class="txt">wow</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:28</span>
<span class="txt">in terms of how many tokens it does so we do around 30 trillion tokens a day</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:32</span>
<span class="txt">the base infrance service is bigger than openi&#x27;s API service</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:36</span>
<span class="txt">so not not chat GPT the API product</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:39</span>
<span class="txt">their API product okay</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:40</span>
<span class="txt">yeah at least in last report and it&#x27;s definitely bigger than the Gemini the Gemini thing and So if you project that out, we&#x27;ll need around 150,000 B200 equivalents.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:50</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:50</span>
<span class="txt">In 2 years from now.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:51</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:52</span>
<span class="txt">That&#x27;s just insane amount of compute.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:54</span>
<span class="txt">Could you could you translate that in dollars for the class?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">35:57</span>
<span class="txt">Yeah. Yeah. So um it&#x27;s about $7 billion of compute spend.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:02</span>
<span class="txt">Easy.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:03</span>
<span class="txt">$7 billion of comput spend. Um too big. It&#x27;s it&#x27;s it&#x27;s a scary amount.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:07</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:07</span>
<span class="txt">Um the the the idea that we&#x27;re going to be able to get access to that by renting</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:11</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:12</span>
<span class="txt">is probably, you know, it&#x27;s probably not going to happen. And the only way we can guarantee that is is knowing that we can we have a strong relationship with the the chip providers and we can go buy them and put it up ourselves. And that that&#x27;s the reason we&#x27;ll do it is access to be able to f fulfill the demand that we have for inference.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:28</span>
<span class="txt">And there&#x27;s also an economic advantage to do it. It&#x27;s about 30% cheaper. Like I think if you think about a scaled out cloud like Oracle,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:35</span>
<span class="txt">it&#x27;s like what 30% gross margins.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:37</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:37</span>
<span class="txt">Um so it&#x27;s about 30% cheaper to do that.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:40</span>
<span class="txt">Yeah. Working integration. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:41</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:42</span>
<span class="txt">Fascinating. So the shape of the business is going to change quite a bit over time.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:45</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:46</span>
<span class="txt">But you want to own your destiny.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:48</span>
<span class="txt">There&#x27;s no other way.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:49</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">36:49</span>
<span class="txt">Like you know the if if you think about like what is what is the what the now I&#x27;ll give you the third risk but I&#x27;ll tell them the three risks right which is like open source the app layer and don&#x27;t have access to enough compute and uh and that&#x27;s the core risk of the business and like if we don&#x27;t take care of that we&#x27;ll be in a lot of trouble.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:09</span>
<span class="txt">Yeah makes sense. Um that makes sense. You know the um actually follow up to the question you said this 263 per hour renewing at 5</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:22</span>
<span class="txt">10 seems egregious.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:24</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:25</span>
<span class="txt">Where did you settle at?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:26</span>
<span class="txt">Oh yeah. We we haven&#x27;t responded yet. It&#x27;s in September. You know we don&#x27;t need to that.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:31</span>
<span class="txt">Yeah. Actually the the the question behind the question is when should we expect the compute scarcity to to get back to normal? Is this like a 12 15 month thing? Is it a multi-year thing? Is it a multi-deade thing? Like what&#x27;s your best I&#x27;m sure you guys have done some Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:48</span>
<span class="txt">thinking or economic modeling around this?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:50</span>
<span class="txt">I I I I I&#x27;m curious what you guys think. You guys I&#x27;m curious what your public&#x27;s people say about this. I don&#x27;t think it&#x27;s ever going to normalize.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:58</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">37:58</span>
<span class="txt">You know, like I I think it&#x27;s what you said earlier. It&#x27;s like if inference demand is a billionx what it is today,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:05</span>
<span class="txt">um like two two different things are happening, right? which is model like the applications are getting more agentic</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:12</span>
<span class="txt">and models are getting bigger</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:14</span>
<span class="txt">which both of those say things say there&#x27;s going to be a lot more inference which just means there&#x27;s going to be a ton more compute.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:20</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:21</span>
<span class="txt">Um you know like I think the analogy really is is like when you show up at JFK at 5:00 a.m. every day there&#x27;s no line. You go straight through.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:31</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:31</span>
<span class="txt">But you know by 8 a.m. there&#x27;s a line out the door.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:34</span>
<span class="txt">Yeah. Um, and we were out the door right now, but unlike JFK, which closes down from like 11:00 p.m. to 4:00 a.m., we don&#x27;t have any reset period.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:44</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:44</span>
<span class="txt">And so, it just keeps compounding.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:46</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:47</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:48</span>
<span class="txt">Actually, that&#x27;s an interesting analogy. Does the inference demand die down between midnight and 4:00 a.m. And are you able to reuse that somehow?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:54</span>
<span class="txt">Well, but there, you know, it&#x27;s it&#x27;s, you know, it&#x27;s, you know, what is it like when it</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:58</span>
<span class="txt">it&#x27;s afternoon somewhere else?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">38:59</span>
<span class="txt">I mean, there&#x27;s like the thing, right? It&#x27;s five, it&#x27;s 5:00, it&#x27;s beer o&#x27;clock somewhere. It&#x27;s like, you know, you can you can justify drinking a beer in any time because it&#x27;s 5 PM somewhere.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:07</span>
<span class="txt">Got it.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:07</span>
<span class="txt">I think that&#x27;s the same thing with inference where it&#x27;s like even if demand here is down in Europe it goes up, in China it goes up.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:13</span>
<span class="txt">Yeah. Yeah. Yeah. Makes sense. Um I&#x27;m I&#x27;m going to shift gears a little bit. Um you know, one of the biggest things we&#x27;ve been discussing in class, a lot of the students here will go on to build build businesses, start businesses, work at businesses. Um, if you were not building base stand right now, what&#x27;s your next best idea? What would you go start? What would you go build?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:36</span>
<span class="txt">Um, that&#x27;s a good question. Like, like I would be I&#x27;d be going more the Cruso route to be honest. I&#x27;d be I&#x27;d be going and investing in energy and power.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:46</span>
<span class="txt">Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:46</span>
<span class="txt">Um, and like you know, I think the buildout is like we&#x27;re just going to need so much space to put this compute. Mhm.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">39:54</span>
<span class="txt">Um what I I think like one of the one of the ideas I&#x27;m most excited about um which I hope someone does, maybe someone here will do is um you know I was I was driving over from Oakland</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:06</span>
<span class="txt">to San Francisco the other day and you always see the ports there and you see the containers there you like containers are one of the biggest economic drivers</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:14</span>
<span class="txt">in history</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:15</span>
<span class="txt">because they normalize they normalize the unit of trade. the um and so you know my idea like one of the things I&#x27;m really excited about is modular data centers</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:25</span>
<span class="txt">modular data center</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:26</span>
<span class="txt">data centers which is like take like and standardize the unit of compute</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:30</span>
<span class="txt">because whoever standardized units of compute now now we can just like really industrialize</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:37</span>
<span class="txt">like if you go and talk to all the folks who are putting up data centers right now</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:40</span>
<span class="txt">and the people who own this pallet space</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:42</span>
<span class="txt">everything is different</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:44</span>
<span class="txt">um so to the extent that you can fascinating</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:46</span>
<span class="txt">you you can modulize is the the</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:49</span>
<span class="txt">and have a shared consistent format.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:51</span>
<span class="txt">You&#x27;re creating an API for compute and then you can then like a whole industry will start around that.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:55</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">40:55</span>
<span class="txt">Cuz then you can have you know the same people servicing these things all over and you know you create and you know that that&#x27;s probably what I&#x27;d be working on.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:02</span>
<span class="txt">Fascinating. Fascinating. Well um I&#x27;m sure folks have taken that taken that note down. Um relatedly um outside of that is there a you know we play this long short game with every speaker. Is there is there a business is there a startup? Is there a founder that you&#x27;re particularly excited about right now or and the other way you think is more more hyped than than they deserve?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:22</span>
<span class="txt">I&#x27;d rather not play that one, but you know, for the for for the former look, I I I think you know, all the things I said, right, the um the [laughter] um the but yeah, no, like I I think you know, the whole any any part of the buildout.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:36</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:36</span>
<span class="txt">Um and then, you know, in terms of short, look, I I I think I&#x27;d be looking at companies that just aren&#x27;t innovating. Actually, I actually think the enterprises are fine.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:46</span>
<span class="txt">Enterprises are fine.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:47</span>
<span class="txt">Fine. Like, but I think they need to figure out how to take the thing that is unique to them.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:52</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">41:52</span>
<span class="txt">And post- train models and go do inference on them. Um the but but I think the ones that aren&#x27;t doing that, I&#x27;d be very very</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:00</span>
<span class="txt">scared about.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:00</span>
<span class="txt">Yeah. Yeah. A lot of uh students right now thinking about what to study, particularly those who are starting their educational journey. Uh what advice would you have for folks? It feels like uh quicksand right now. Yeah, I I you know I don&#x27;t know. I actually think you should just study whatever that&#x27;s fine. like you know they they&#x27;re like you know oh man such a such a I think you know</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:24</span>
<span class="txt">yeah you change careers</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:25</span>
<span class="txt">change careers you know and like you know the you could become an expert in anything in six months so just do the things that&#x27;s fun you know like I I don&#x27;t but but I think in terms of like</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:34</span>
<span class="txt">you know if you if you are you know like actually financing models of of you know I spending a lot of time thinking about you know how to finance data centers right now</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:45</span>
<span class="txt">okay</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:45</span>
<span class="txt">just think you know something I&#x27;m going deep on and you know I think you the project financing stuff is actually pretty nifty there and so study. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">42:54</span>
<span class="txt">Yeah. Fascinating. Well, we&#x27;ll open it up for questions. We&#x27;ve got five more minutes. Uh go ahead.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">43:01</span>
<span class="txt">What&#x27;s your take on like the futures?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">43:04</span>
<span class="txt">Oh, I I don&#x27;t know. Like it&#x27;s obviously interesting. Um I I I think the market is way too um what&#x27;s like the it&#x27;s way it&#x27;s if you if you went like saw how compute deals are done it it&#x27;s basically a drug it&#x27;s a drug market like tr truly it&#x27;s like you have a guy I and um</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">43:32</span>
<span class="txt">who&#x27;s your guy we have so many guys you know the and Like there&#x27;s a guy at our office who kind of just sits in the corner and he just call he just calls people all day asking for comput it&#x27;s dead. Yeah. And it&#x27;s and it&#x27;s absolutely terrifying. So like I&#x27;m very bull like there&#x27;s clearly going to be markets here but I don&#x27;t think the market has the you know it&#x27;s it&#x27;s like um it looks more akin to you know a is it&#x27;s not a mature market. Um but I I think a good way to think about like electricity markets are</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:06</span>
<span class="txt">it&#x27;s not as efficient. slippage. Yeah, a little slippage. Yeah,</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:11</span>
<span class="txt">go ahead.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:13</span>
<span class="txt">So, bases is that 10,000 eggs. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:18</span>
<span class="txt">What&#x27;s the benefit? What&#x27;s the</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:22</span>
<span class="txt">Yeah, I I said that, right? The antithesis is that there&#x27;s two there&#x27;s only two people who can get computer in the world and they own everything. That&#x27;s um you know, that&#x27;s one. The second one is that you know we we are dependent on open source models getting very good and you know there&#x27;s not enough open source models. That&#x27;s number two. Um and three is that you know um if you believe if you truly believe in like the the like the AGI thesis you know and you just play it out and you keep pushing and pushing and pushing like we all have nothing to do.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:56</span>
<span class="txt">But then I would argue maybe inference is the only market left.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">44:59</span>
<span class="txt">That&#x27;s right. is the only mocking left. And</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">45:01</span>
<span class="txt">that&#x27;s right. Yeah, that&#x27;s right. Go ahead.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">45:07</span>
<span class="txt">Everybody talks about how important open source is, but we&#x27;re not really funding it. Well, every except um we&#x27;re not really funding it in the United States. Do you think given the national security?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">45:27</span>
<span class="txt">Yeah. Yeah, definitely. I I think like the US government it pro I think it already is getting involved and thinking about these things. Um I would argue that actually like you know when you go look at Nvidia when you go look at Microsoft and Google they&#x27;re actually putting a lot of work behind open source. Um I just don&#x27;t think we&#x27;ve seen the the the like has it hasn&#x27;t borne fruits just yet and it will come. So I I think like actually there is quite a lot of investment um but we&#x27;re just not seeing it and I think the government is getting involved but yeah definitely I think you know you have to create artificial incentives and you know there will have to be a alliance and um um for all these things to be incentivized top down. Yeah</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:11</span>
<span class="txt">you saw the other side happen two days ago. The Chinese government invested in deepseek.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:16</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:16</span>
<span class="txt">Yeah. Anyway, go ahead.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:20</span>
<span class="txt">Sorry. What&#x27;s to stop and open AI from open sourcing one of their lagging but still quite good models and then providing a workforce to both enterprises. Are they still focused on research? Why can&#x27;t they do that?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:36</span>
<span class="txt">Open already does</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">46:37</span>
<span class="txt">they they do that to some extent but in a lot of ways like it&#x27;s for them to invest in post training to some extent it&#x27;s to almost give up on their their what they think the thesis is right. It&#x27;s like, you know, one of the most fascinating things about an open anthropic is, you know, if the thesis is AGI is everything, the way they do every anything at all, they should every every dollar should go to pre-training.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">47:03</span>
<span class="txt">Yeah. Um and so but you know they they already do that to some extent but their incentive to push dollars towards that is never going to be the case because they can they the reason why you know like all these models exist so they can just make the most money on the most expensive model like they they want they want to push the capability gap the capability argument as much as possible. All right, go ahead. One final question</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">47:28</span>
<span class="txt">about the idea of the model modularized data center. I&#x27;m just wondering what would be the difference between like virtual machine.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">47:36</span>
<span class="txt">Question is what&#x27;s the difference between a modular data center and a virtual machine?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">47:40</span>
<span class="txt">Yeah, that that&#x27;s a good question. Um, well, I I think you know the virtual machine sits a couple layers higher up in the stack. I I&#x27;m just trying to figure out like how do you how do you create make it cheaper to put up more compute and so that you know like that that&#x27;s a in a lot of ways you&#x27;re kind of asking like modular containers you&#x27;re kind of talking about containers which is once the compute exists you need that and it kind of is that just two two layers deeper</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:08</span>
<span class="txt">yeah it&#x27;s at a different layer of abstraction</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:11</span>
<span class="txt">I&#x27;ll take one more was there one here go ahead</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:15</span>
<span class="txt">thanks for coming I&#x27;m just curious what you think about where you more like</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:24</span>
<span class="txt">it&#x27;s every it&#x27;s everywhere like disag you&#x27;re really separating our concerns the kernel stuff is continuing like a big like the amount of people in Nvidia and every you know even Stanford I think who work on stuff like thunder kittens and um things like that there&#x27;s a lot of opportunity there um but the entire stack is changing like that&#x27;s that&#x27;s like the most fascinating thing about this market is that Everything is like quicksand like you you throw everything away, you know, like Ilia might come out with a new with a new architecture for a model and everything it&#x27;s up everything is up for grabs again.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:56</span>
<span class="txt">Ilia</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:57</span>
<span class="txt">um Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">48:58</span>
<span class="txt">Yeah. Do you know what he&#x27;s working on?</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">49:00</span>
<span class="txt">I I I mean I have sense. Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">49:02</span>
<span class="txt">Okay. That was all about [laughter]</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">49:04</span>
<span class="txt">All right. All right.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">49:06</span>
<span class="txt">Yeah.</span>
</div>
<div class="para" onclick="this.classList.toggle('highlight')">
<span class="ts">49:07</span>
<span class="txt">Perfect. Well, thank you so much. This was fun. Yeah. Appreciate you making</span>
</div>
</main>
<script>
window.addEventListener('scroll', () => {
const h = document.documentElement;
const pct = (h.scrollTop / (h.scrollHeight - h.clientHeight)) * 100;
document.getElementById('progress').style.width = pct + '%';
});
document.querySelectorAll('.ts').forEach(el => {
el.style.cursor = 'pointer';
el.title = 'Click to copy';
el.addEventListener('click', e => {
e.stopPropagation();
navigator.clipboard.writeText(el.textContent);
const old = el.textContent;
el.textContent = '✓';
setTimeout(() => el.textContent = old, 800);
});
});
</script>
</body>
</html>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment