Oxide Is Crossing the Chasm  ## Metadata - Author: [[Changelog Master Feed]] - Full Title: Oxide Is Crossing the Chasm - Category: #podcasts - URL: https://share.snipd.com/episode/840e57ce-7401-4ceb-8914-2187212af68c ## Document Notes ## Highlights - **Capital As Strategic Insurance** - Oxide raised $100M intentionally to avoid running out of capital and to enable scaling operationally. - Bryan Cantrill frames VC funding as necessary fuel but warns about raising too much too early. Transcript: Bryan Cantrill And, you know, we are, you know, we obviously just raised 100 million bucks and we were very mindful about that and making sure that we, I mean, the flip side, at least for oxide, is that It is, we have always believed that if this company can cross the chasm, it is a generational company. It's a huge company waiting to be built. And the number one risk for the company actually is running out of money is actually capital. I mean, VCs hate to hear this, right? They're like, oh, that's your number one risk. It's like, yeah, that's why we're here. I mean, it's like, we've got money. That's why we need money. It's like, oh, all right. Well, is there any other way we can be helpful? It's like, no, you can be helpful by like, you know, by running a check. It's like, it's like, wow, God, okay. That's kind of boring. We'd love to introduce you to like, can we introduce you to potential customers? Like, no, that never works. No, you don't even, we're good to go. Yeah, we're good to go. We actually need, so that is, you know, I think that I would skirt the question by saying, you know, we're mindful of all, all seasons of Silicon Valley and all stages of a company that Represents. And how do we, how do we be the company we want to be without reliving some of the seasons of Silicon Valley, um, without instituting Piper pulse, um, without, you know, which I still think Is one of the, uh, and actually the system at Facebook was actually called pulse. I'd not realized this got some X metaphors. You're like, you realize that that is like actually real. And that's real. And also the chairs ad, do you know there are the tables that's based on a chair's ad tables, which is a spoof ad. And you're like, this is such good satire. And so biting. And it's like, Oh, that's because actually Facebook paved ([Time 0:05:16](https://share.snipd.com/snip/954ad920-de58-4cd6-b333-612b140f0390)) - **Courtship Beats Cold Pitches** - Meet VCs early and build courtship over time so they trust how you will use capital. - Seek partners who de-risk big rounds and align with your growth needs. Transcript: Bryan Cantrill And I, I think that, you know, when we, uh we were not necessarily raising. Right. So when we had raised our series A and, you know, when you are a venture back company, you always have to be willing to take the right meetings with venture capitalists because you always Need that next round. And especially a series B, I think, is the most difficult series to raise because you're kind of a teenager as a company. You are, you know, teenagers can be tough because they've got adult needs, but they don't have adult strengths. You know, they're not like, they're not actually like contributing yet. They're on the cusp of it. They're close, but you are, you know, I got two kids in college, so I'm definitely, you know, I'm feeling this right now. You're just like, okay, why don't I give you my bank account and you just give me whatever is left over? Let's do it that way. But, you know, you're kind of a teenager to series B and you need to have, so you're constantly like trying to find the right partner for that and the right. And we, you know, we had a bunch of firms that are just like not very good, you know, so you're kind of you're not taking every meeting by a long shot. But we this firm that really wanted to get to know us and, you know, the first meeting was fine. But then they came by for a second meeting and that second meeting was really good and they'd done a lot of homework and this is USIT. And, you know, the third meeting was real, you know, and you kind of have, it's courtship that happens over a long period of time. And, you know, they got to know us really well and they felt like, you know, when you're thinking about your next kind of financing, we think you guys need to go big. And we, and it was, it was bigger than we were like, wow, okay. But really the way they structured it was great from our perspective and to kind of de-risk some of the risks that we saw of taking that much capital. And it gave us, it was really important because again, like the way from it still to me, the, the, the greatest risk to oxide is that we just don't make it. We run out of capital before we were able to build, it's really hard to build a business. It's especially hard to build a business where it's like, we're selling big ([Time 0:07:56](https://share.snipd.com/snip/a6880991-e4d0-4f3b-bfd4-72fda39bc37e)) - **Funding Builds Customer Confidence** - Big financing signals credibility to customers and can unlock trust for multi-rack purchases. - Oxide uses capitalization to convert pilot interest into large, committed orders. Transcript: Bryan Cantrill And I, wow, that's, this is great. You know? And I, so I think that the, the, the capitalization is not just the capital that, that we need to go expand manufacturing and to go, uh, cause the other, I mean, the other really significant Aspect of this is that we, and we knew this was going to happen, but it's been fun to see it happen where a customer buys one rack or maybe two. And they're like, Hey, um, what would 10 look like? What would 50 look like? What would a hundred look like? And, you know, for a long time, that was just kind of like, eh, just kind of pillow talk. You know, we were kind of dismissing it. You know, it's like, it's great. We obviously always took it seriously, but as time has gotten on, like those questions and conversations, but like, no, no, I'm like, I'm serious. Like I've got, like, I've got budget. I've actually used this thing. I know what this thing actually is and isn't. And like, I've got this new project coming up, but this would be perfect for it. And these are the kind of numbers I need. And you're like, those numbers get significant really quickly. And you do actually need a head of capital to be able to secure the financing that you need to be able. I mean, see, there is a flywheel that you need for which you need to like, this is not on the one hand, hardware is not as capital intensive as people think, because it's not like we're building Our own plant. Right. We're not benchmark electronics, our contract manufacturer. And like they like they've got their own ITAR facility. We're not building a manufacturing line. We're not. There's a lot of things we're not doing that people kind of associate with hardware. But on the other hand, it is also capital intensive because, yeah, you got to go procure a bunch of DRAM and a bunch of CPUs and a bunch of boards. And like you can't invoice someone on that until it's actually been like shipped. And so there is, especially as the numbers get large, the financing angle becomes really important. ([Time 0:13:32](https://share.snipd.com/snip/0bb4890c-4678-44dd-b274-179499c3e42f)) - **The $23 Close That Almost Stalled A Round** - During an earlier raise they were short by $23 because of share-rounding quirks and had to get another investor to buy an extra share. - Bryan recounts this to show how wire mechanics and timing get unexpectedly hairy in fundraising. Transcript: Bryan Cantrill Do they like it like it will like wires into your account like you raise around yep you raise around your account oh yeah oh well actually also i i will i i mean kind of a funny story in this Regard in our series a so uh not only does it, it wires into your account, but it's actually not yours until the round is closed, but it's like literally sitting in your account. Okay. It doesn't go to an escrow account. It's not like buying an escrow. You don't own it. But people will be like, it's in escrow, but like, you're like, it's okay. You keep saying it's an escrow, but it's like, it's kind like setting up the bank. Like escrow is a real thing that you can do. Escrows are a real thing you can do it, but like, no, no one's going to do that. So the, but there is, this is where, and things get super, like when you're raised, so you're raising a round and you're, you're raising a price round and you know, people are going to Invest at, and you know, people are investing conditional on other investors. So you need to wait until you kind of like hit the number that you've got on the paper before like the round is closed. Like you're not so you can be like waiting on a wire and you're not closed or you can. And, you know, one of the things that we did in our series A when we're extremely grateful to our investors for accommodating an early close. So we were going to raise, you know, we ultimately ended up raising what, $44 million in our series A and our investors did an early close at 30, which was great because we, this was 2022. It was really brutal to, I mean, the environment had turned very, very brutal. And, uh, we were, you know, so then, you know, like, okay, once I cross this, like this $30 million threshold, then like every, and not surprisingly, one of the larger investors, institutional Investor is like, look, we wire last. Like that's the way we operate. So we are going to operate, we are going to wire last. And I'm like, I'm kind of sympathetic to that. Like they want to make sure that everyone else has already wired before they wire. Well, they, they, they go to wire and you're like, Hey, you guys are $23 short. And we're like, Oh my God, we're $23 short. And like, how do you end up $23? And of course, everybody is like, Oh, I'll pony up $23. Like, are we short? Yeah. It's like a bunch of like grad students out for dinner or something. It's always comes up short. You're like, okay, who did there only like four of us here who didn't throw in and everyone kind of swears that they paid it up. And I'm like, of course, you know, it was, it speaks how it outside. Everyone's like, I'm in for $23. It's like, no, no, unfortunately it doesn't work that way. We can't just like put, you know, put a couple of, of, of bills on the table. The reason we're short is because of share math and shares get, shares will get rounded down and you end up and you have probably, I'm presumably you have many rounds that ended up being Short by a share or some very small number of multiples of shares. And most people don't care. But in this case, that investor is like, no, I'm not wiring until you have the $23. And we're like, okay. ([Time 0:20:24](https://share.snipd.com/snip/95de2e9f-d606-4d6a-9664-f54f4017cb27)) - **Safes Are Useful — Until They Stack** - SAFEs simplify fundraising logistics by deferring valuation and avoiding complex coordinated wires. - But raising too many SAFEs can dilute and complicate later priced rounds. Transcript: Bryan Cantrill What that says is like, look, let's not deal with all this. We have no idea how much you're worth. We got no idea how much you're worth. And let's also like not wait until everyone's ready to wire and this kind of like waiting for the, like, tell you what, we're, I'm just going to invest this amount of money. I'm going to invest 500 K. And then we, at some point you're going to have a price round, or you're going to have an event that happens in acquisition, what have you, we will price it then. And when we price it, because I'm doing this in advance, I get a discount on that, or I get, you know, there are different ways that people kind of structure those things to incentivize Them. But a safe is a, or there's a cap. I mean, that's another thing that people do. We'll say, all right, the, so you get this. So like, I'm going to invest, but I'm really worried that I'm going to invest 500K and then you're not going to raise for like years. And then you're going to raise at this multi-billion dollar valuation and i'm only going to own a little sliver of this so to prevent that from happening i'm going to put a cap in place So if you raise that above this valuation cap i get it as if it's at that valuation cap um and so the the and it's also a way for them to kind of be able to model it so but the safes are great for That because you can just like go right now. They can also be too great. They're kind of like candy. And entrepreneurs can raise too many safes. They don't understand how their safes are stacking, and they go to raise their price round. It's like, whoops, you got washed out because you didn't keep track of the safes. ([Time 0:26:32](https://share.snipd.com/snip/53fd1118-1015-4cdb-852f-69d1048132f2)) - **Being Unique Triggers VC Fear And FOMO** - Oxide expected VCs to be both attracted and terrified because the market was contrarian and unproven. - Being unique can create a fear/greed superposition among investors that slows decisions. Transcript: Bryan Cantrill Was so when we raised in 2019, my assumption was one of the myths. The first myth that we are going to dispel is that you cannot raise for this company. And I had many people tell me you cannot raise. You will not raise for this company. No venture firm will ever fund it. And it turns out it's more complicated than that because it's so contrarian that it's actually that FOMO, it does like kind of light VCs up because like, oh, this is like a big potential. And the problem we had actually was much more that VCs. Next bell y'all. Totally. That's right. Right. That they would be attracted to it, but then also terrified by it. And they would be kind of locked in that kind of, again, that kind of superposition. And we actually tried to actually, this is funny, Adam, you'll appreciate this. Like I did try and I can tell you that like, this is where, you know, season two, episode one definitely has some truth to it. You do actually want to disqualify from VCs to be like, no, this is like, you should not be doing this. Like, this is not something you should invest in. Like you're a bunch of software investors and they're like, Oh no, no, no. Now we want to invest much more. You're like, no, God damn it. Like I'm not actually trying to neg you. I'm not, I'm not trying to do what Eric did to you. I'm actually literally trying to talk. And actually there was a, there was a VC firm in particular that was so adamant. And I was like, I'm like, these guys are just like, not kind of technically lightweight. And they just did not appreciate the technical risk. And I'm like, you're never going to get over the line. And I'm like, you don't understand. We are going whaling. You should be fishing from the pier. Actually, you should be shopping in like the fish section and Safeway. Like that's, that is as close to it. And they're like, no, no, no, we want to go whaling. We want to go whaling. I'm like, no, God damn it. Like you don't actually, no, you don't actually want to go whaling. Like we are going like multi-year mission to far away ocean. So, but we assume like we're going to dispel this. We're going to show that you actually can raise. And there'll be a company that raises behind us. And so we were kind of geared for that from the beginning, that there'll be another company that raises behind us. And so we did one after we were completely done with our seed, like, eh, that was actually harder than I thought it was going to be. It's like, actually, eh, maybe. And in particular, one of the things we definitely did do with our series A is like, well, I'm our seed rather. I'm like, I'm definitely going to find the investors that want to invest in this because they believe in it outside of oxide. ([Time 0:40:36](https://share.snipd.com/snip/ac21baa9-42f0-45a6-a0c2-1cfdc19188d0)) - **Pay Cash First, Equity For Upside** - Give every employee equity and competitive cash so equity isn't treated as payroll substitution. - Pay living cash salaries first, then use equity to share future upside and retain talent. Transcript: Bryan Cantrill It, yeah. Uniform compensation. Uniform. Yeah. Because people are like, oh, you think everyone's equal. Like, I definitely don't think that. Adam Stacoviak Well, that's why I was trying to get at with just thinking about raising money and like, you know, people getting a slice of the pie. I'm sure there's equity involved in these people. It's not all cash compensation. And so how does that play out with that? Yeah. Bryan Cantrill You do get people who are like, oh, you're not talking about the total compensation because you're not talking about equity. Well, first of all, everyone at Oxide has got equity. Very important. That's just like important principle for me that everybody has equity in the company. So that's just like bedrock principle. The equity also compensates for risk and risk in the company goes down over time. And there was way more risk in for employee number one than there is for employee number hundred. And it's like, those are not going to be. And, you know, it is kind of funny that people are like, oh, like, you know, everybody in the company should have if you say they should have equal compensation, which I'm like, again, Not the word we use. They should also have equal ownership. And you're like, how does that work? He's like, walk me through how that works. Like, I don't know, like, you know, you can do this. It's like you can have an employee owned cooperative. I think, you know, go down to your local organic grocery store, employee owned cooperative. And, and, you know, it's like the, I'm reminded of that old onion headline, a local app beat, local Applebee's hotbed of Machiavellian maneuvering. You know, I think that just like, there are a lot of challenges. It's like, we're sorry, we are not an employee, but that's not our objective. And I also think, by the way, this idea that equity is part of your compensation is misplaced. And this is a problem. And this is something that these large companies with product market fit that are public use the fact that they can kind of print these Confederate dollars in terms of their equity. And they use that as part of your compensation. Because if you view equity as part of your compensation, it means that you are selling your equity all the time. Like you're a net seller on your company. You know, you are, you're not buying hold on that. And I just don't think I, and I get where people kind of come from with it, but people don't understand. Startup equity is not liquid. Your Google equity and startup equity shouldn't use the same word because they're not the same thing at all. Like they are literally- You can sell one, you can't sell the other. You can sell one with infinite liquidity and the other one has got zero liquidity. I think people don't often understand what zero liquidity means. And, you know, one of the- It's not real. It's not real. Or- Well, it's real, but it's not real yet. Yeah, and actually, you know, in many cases, like from a liquidity perspective- It's not realized. The worst thing that happens is success, right? Because, you know, company kind of like stumbles and exits, and there are plenty of companies that exit for greater than their preference tax. They pay out to the common. You get your money back. And it was like, you know, you may be disappointed with that. It's not it's not the venture return that venture capital is running. But an angel investor be like, oh, this is great. Like, you know, I gave you 75K and I got back like, you know, 91K. Great. You know, and but like not if a company is successful, like I gave you 75 K, like where's my 75 K it's like, Oh, it's like, sorry, we're six years in, we're 10 years in, we're 12 years in. And you know, that, that liquidity event doesn't necessarily happen for a long time. So it's like, these are not in the same ballpark. And it's like for a startup, the, like your startup equity, it's very important that everyone has it. Very important. Everyone has a share in, in that kind of generational future. It's also important that like, that's not what actually pays for the kids' summer camp. What pays for the kids' summer camp is cash compensation. And so that cash compensation is very important to us. Adam Stacoviak So how do you do it then? Do you just offer opportunity for them to buy? Bryan Cantrill We make sure, I mean, everyone, when they join the company, gets equity as part of their package. And then, you know, we re-up people over time again to kind of, because it's important that employees have a share in that. ([Time 0:45:18](https://share.snipd.com/snip/e732af8e-f133-4a75-946c-8074bddb7ede)) - **Transparency Raises Hiring Standards** - Uniform pay transparency attracts higher standards and makes hires treat new colleagues as peers. - Bryan says this raises hiring bar and improves product and support quality. Transcript: Bryan Cantrill And one of our engineers is like, I think we should talk about the compensation. And we talked about it. And this is in March of 2021. And that is in many ways, the most important single thing we did in the history of the company, because it brought in so many new folks to the company. And not because they were drawn to the, the, it wasn't the fact that it was uniform per se. It was the fact that it represented a different way of conducting ourselves. It's what it represents that's important. And the reason that our engineer was advocating we talk about it is like when I talk about oxide values to friends and family, they think it's bulls**t until I talk about the comp. And when I talk about the comp, it's like a record scratch. People are like, what? You're serious? Oh, you're really're really serious well no one else does that does anybody else doing that no not to the best of my knowledge not at our time you even nope and we've been told over and over Again that like won't scale won't scale won't scale i'm like i don't know maybe i i'm whatever i don't know i've been told i was told that when we were 10 20 30 40 50 we're now at 70 we'll be And and or 80 on our way to 90. Um, we, and I would say that it is playing a more important role, not a less important role. Cause the thing it's been, it's had so many like interesting side effects. One of the super interesting side effects is everybody is hiring a peer. And, uh, when you tell people that they're going to hire a subordinate, their standards for that person are going to be a lot lower. And they're maybe going to overlook a bunch of things that maybe they shouldn't overlook. You tell someone they're hiring a peer and they're like, OK, I want to make sure that this is like worthy of us. Right. And so it's like, yeah, it's it is not easy to get a job at Oxide. It is, we are very oversubscribed and we've got people that take hiring extremely seriously. ([Time 0:50:43](https://share.snipd.com/snip/8779b080-65db-4f14-8a73-6832f3553696)) - **Remove Pay As A Behavioral Lever** - Separate compensation from promotion incentives to avoid people gaming behavior for raises. - Use transparent, company-wide pay so employees focus on the right work, not personal gain. Transcript: Bryan Cantrill Is this something that I use my promo package or not? And great folks will not think about that too much. And they will do the right thing, even when the company is not rewarding the right thing. And we ask people, when are they most unhappy and why? And almost invariably, it's when they are doing the right thing and management is discouraging it. And to be in an environment where it's like, no, we're going to separate all that. And now allow people to do the right thing all of the time. So much more of the right thing just happens. Now, of course, you got to hire carefully. And that's what we do. But we we hire people that have that disposition. But man, you build a whole company that way. And it's extraordinary. ([Time 0:56:56](https://share.snipd.com/snip/b0bfc635-9ccd-4312-a275-53c4aa762ef0)) - **Investor Said Oxide Can't Scale To 600** - An investor told Bryan Oxide couldn't scale to 600 people, and Bryan replied he might never want that size and values disciplined hiring. - He emphasizes founder-led hiring to avoid overhiring and preserve culture. Transcript: Bryan Cantrill Honestly, it's a different question. No, it's a different question because they, you know, and I had, I did, I had an investor who turned to me at one point is like, you know, because they kind of forget that we've done this, Our investors, like they don't, I mean, it's not something they really think about. They know that our team is very good. They know we don't have attrition problems. They know. And also if they looked at one point early on, we're doing this one of our investors, like, you know, we did the math and like the, what you're paying people is like dead on the median, like Salary per employee that we have in our portfolio company. So it's like, you, you guys are not like off the mark one way or the other. So it's fine. Like, like doing it differently, like fine, knock yourselves out. So the VCs, I kind of don't care. But one of these investors turned to me is like, Oh, because we, in particular, one of our investors was like, was talking to Steve and it's like, you know, Steve, you've really done a Terrific job like you, you basically, you deserve a race. And Steve's like, that's, that's great news. That means we're going to give the whole company a race. And he's like, Oh, right. This, Oh, right. You people, you people in this again, you people. And, but we did, right. We did the give whole company a race and, uh, and it was great. But this investor turned to me, he's like, okay, you realize like you won't be able to do this at 600 people. I'm like, you know, I don't know that we'll ever be 600 people. I don't know that. I don't know that you, like, really? You can't build a billion-dollar company with lessons? I mean, like nobody has ever done even what we have done. Nobody has even been at this relatively small scale. So like, I'm just like, not, I, the right answer is, we don't know. And there are, will things change? Adam Stacoviak Maybe, I don't know. How many support people do you have like per rack or per customer or some sort of metric that you could take and then multiply to say like, well, at some point, because I looked it up, Dell, I'm fully admitting that they're probably overly bloated, et cetera. They got a hundred thousand people working there. So 600 seems small at scale. Maybe it's not, but it just like, that's not scaling up oxide very much from here. That's a 10 X basically. Totally. You got 70. Yeah. Bryan Cantrill Yeah, totally. And so I think that the, a fair question. I also think that like, how many support people would Dell have if they thoughtfully engineered their products, made sure that every, I mean, the number of people that Dell dedicates To telling customers that you're the only one problem. I've dealt with Dell support. Yeah, sure. Adam Stacoviak Yeah, I have too. Bryan Cantrill And it's like the you don't need very many people at a company to support a problem. It's the old Maytag, the Maytag repairman. Right. I'm among friends on this one. Right. Adam Stacoviak You are you are you with me on this? I know Maytag, but I'm not sure. Bryan Cantrill OK, so Maytag repairman was an ad from the eighties and the Maytag repairman had nothing to do. And he had nothing to do because Maytag Maytags work. This is when washing machines used to work. So, you know, those of you who have a modern appliance know that this is not true, but it used to be. They don't work anymore. They don't work anymore because it's a terrible economic model. I think that the, or that's what companies convince themselves of. We want to build something that works. And the way you build something that works is when it doesn't work, you jump on it and you understand it in totality and you fix it. So the next customer doesn't have that problem. If every problem is only seen once because you've got such an ACEs support team and ACEs engineering team that they jump on it and they root cause every single thing from the very first Failure and they don't need to see a thousand failures 10 000 failures how many support people do you need a lot less than dell has so that's the that's the experiment we're doing right So how many do you have now four something like that. Four or five, something like that. Yeah. Dang. That is a small amount. Yeah. And it's, and you know, we will continue to hire, but like, we're going to continue to hire extraordinary folks, you know? Adam Stacoviak Sure. And that's the part that I say, how far does that go? You know, you, you have to be incredibly vigilant in your hiring, which right now you are. And a certain point, you might have the demand for hiring and you might actually be like, we compromised a few times and we're getting out to the mediocre people. You know? Nope. Not going to do it? Nope. Not going to do it. Not going to do it. Bryan Cantrill That is when, and this is where, and again, not to God, it's like, not to praise Paul Graham a second time, but this is where the founder. Oh, two times PR? Two times, I know. Founder mode. 100% founder mode. 100%. And this is where founder mode is actually extremely helpful. Because if you look at those companies, why does that essay exist? Because Brian Chesky woke up and suddenly there were a bunch of people at his company he didn't recognize and he didn't hire. And the failure mode is always the same. The failure mode is that they, they lose control of their hiring because they don't have good processes to begin with. The processes suck because they don't take it seriously. They, they know it's like, Oh, we want to have a great team, but no one actually understands what that hiring process looks like. They don't actually admittedly. I have, I've had some advantages of making some grievous mistakes in my past that forces up a very good hiring process. The, not an oxide, but the, they lose control of hiring. They hire, I mean, I don't want to say middle management is the root of all evil, but, you know, it's, you know, maybe a little too reductive. But the problem is that middle management grades their career on the number of reports that organization. It's like, if you hire software engineers based on the number of lines of code that they write, if that's how they value themselves, like you're going to get a lot of bad code and you're Going to get a lot of it. If you hire management that grades their career based on the number of reports that organization, guess who needs more recs? That manager you just hired. Really? We can't do it with fewer people? No, we can't. I need more people. Okay, you're trying to advance your own career. I get it. And this metastasizes and you wake up and you have too many people and they all do it. And so I turn to this investor. I'm like, you know what? I don't know if we're ever going to get 600 people. And you know what? We are also the only company in your portfolio that has not overhired. He's like, that is true. Actually. I'm like, you don't have to say actually. And you agree with me, by the way, you don't say actually, but the, um, and we've not overhired and we are not going to overhire. Yes. That's going to mean that it is painful and requires vigilance. ([Time 1:00:18](https://share.snipd.com/snip/7f8f4062-73d2-48cb-9e77-623a95977a0c)) - **AI Needs More Than GPUs** - AI workloads include significant CPU-backed general compute beyond GPUs for tasks like orchestration and code execution. - Oxide sees opportunity in serving inference and general-purpose compute needs, not just accelerator-heavy training rigs. Transcript: Bryan Cantrill Yeah. So, I mean, I think for sure. Something to know about that world is it's really easy to focus on GPUs. And they're important. There are a bunch of GPUs. We obviously don't, we don't have, we do not have an accelerator-based sled. We will at some point for sure. Something we've definitely got our eye on. When we do that, it almost certainly will not be focused on those training workloads. We focus on inference on, and then we can kind of talk about that. There's a lot to talk about in terms of inference. But I would say that one thing that people should be aware of is that there is a lot of general purpose compute involved in what people view as an AI workload. So I mean, I've been, I'm sure as you've been mesmerized by watching these things show their steps. And the number of times, like if you give chat GPT or Claude an image to do some, like, I want to recognize some things on this image, it will write a Python program. You know, I mean, you can watch it like write Python programs, run those Python programs, get the results of that. It will obviously search the web. It will do all sorts of things. Like, yeah, that didn't happen on GPUs. That's happening on CPUs. So just standard servers and APIs, et cetera, et cetera. Yeah. And also then like need to be containerized, right? Need to be like, especially if you're going to have something where like I'm running general purpose compute at behest of a user of this thing is the best the internet. And, you know, we had Simon Wilson on our, on, on the podcast, on our predictions podcast, which is great. And he was talking about Claude compute. The first thing he had Claude compute do is download a UEFI vulnerability rootkit. And it's like, okay, I'll go execute this thing. And you're just like, you know, so the, that is, uh, there's a lot of general purpose compute out there. ([Time 1:11:47](https://share.snipd.com/snip/0ff9292a-2acb-4d35-9077-9cab207571c5)) - **A Supercomputer Customer Bought Non-GPU Racks** - Idaho National Labs pushed Oxide to take their call despite assumptions they only do non-accelerated compute. - The lab wanted Oxide to replace Dell/VMware for many non-supercomputing workloads, surprising the team. Transcript: Bryan Cantrill And, you know, we had our, our first customer, God bless them, Idaho national labs. Idaho national labs had to kick our door down because we're like, you're a supercomputing facility. We're not making accelerated compute. We're not taking your call And they're like, can we please explain to you why we're interested in oxide? And finally, Steve's like, I'm just going to take this call. These guys won't leave us alone. I'm like, it's a waste of time, but fine. And he gets up and they're like, okay, can we talk for a second? Because yes, we have a supercomputer. We also have a bunch of Dell VMware that does a bunch of stuff that we need. That's not the supercomputer. And can we, don't you guys replace the Dell and VMware? It seems like, I'm so sorry. I'm so sorry. We didn't take your call. Yes. That's what we do. I'm so sorry. I'm so sorry. And so there is that analog is, is where we are. ([Time 1:14:00](https://share.snipd.com/snip/de12e554-c6fa-4a02-b099-68823f55c423)) - **Treat Big Orders Like Operations Projects** - For large orders, map long-lead parts, partner with reliable manufacturers, and establish clear PO and supply-chain processes. - Be transparent with customers about delays and remediation steps to build trust. Transcript: Bryan Cantrill So actually this is where you need to go into the supply chain, right? You need to go find those log lean time parts. You need to go and we lean on, uh, or leverage partnerships that you've got. And when you talk about like big numbers like that, there's also a challenge of like the numbers get big, right? So now you're like, okay, well, we've got a lot of approvals that we need to go through internally. And there's a bunch of, you know, understandably, like you're going to actually like, this is a big number. And, you know, like, okay, well, when can we get it? And it's like, well, you know, it's like, we need to get the PO before we can, you know, how do we kind of kick stuff off without having a PO in hand? Because like you don't, and this is, you know, very mindful of, we're very mindful of history in everything we do, but the history of the Osborne computer company, right? Famously, Osborne went belly up when it actually had product market fit. Now there's a bunch, and there's a great book, great, maybe in air quotes, but a book called Hypergrowth out of print where Adam Osborne gave his version of events, but they lost it on Operations. It's kind of the TLDR is like, they had a product that people wanted and they died because they screwed up the operations of it. And they ended up with too much inventory of the wrong stuff at the wrong time. And they just went insolvent. And so like that can definitely happen. So when you have a customer is talking like, hey, I want 50 racks and I'm serious. I want 50 racks. And this is not just like pillow talk. This is like, what does this look like? Give me a quote. I need to go like, I'm going to go through my approval process. Then you need to be like, okay, we got to go light up bus parts, racks, rectifiers, compute sleds. We need to go figure out exactly what this thing is going to look like. We need to figure out where the long lead time parts are. We need to go on the inventory that's cheap. You just procure that. That's easy. And then you've got some decisions to make on, you know, people buying at that scale need some skew control. You know, maybe I want different TIMs or I want, you know, you've got to afford them that. That's a back and forth. So, you know, there's a whole galactic amount of detail. There's an entire galaxy of detail that is not the building of the thing, which already has enough detail, but is the actual the operations piece of this is extraordinarily complicated. It's why our operations team is terrific. Talk about being just as important as anyone else in the company. And boy, when you go do one of these exercises where it's serious and we're going, this is not a drill. And you really appreciate it. Like, wow, this is just extraordinary resourcefulness and agility and being able to, you know, probably unsurprisingly, we've got a couple of our operations folks have got operations Experience from the service. I got some, you know, ex-Air Force folks that, you know, are used to just like, how do we go deal with like very large scale operational issues and, and having to make up a bunch of stuff On the fly in terms of like, okay, we can't like your facility, your, you know, we, your facility in Toronto can't do this. But there's another facility that you've got in Malaysia. What does that look like? Oh, and then you, oh, by the way, then you have like the Joker in the white house spinning the tariff wheel. Can we, like, this is hard enough, by the way, without being like, there is like, okay, there's a tariff. There's not a tariff. No, no, no, no. Now we're going to have a tariff later. We're going to do this. It's like, okay, this is actually hard enough when everything stays still. But, and it's all challenging. And it's a lot of detail. You know, you just have to get and you've got to embrace that detail and you've got to go headlong into it. And then I think the other thing that's been important for us is then be transparent with folks about where we are and where we're not. Like if we, you know, if and when we hit a snag, we're going to be transparent about what it is. And we're not going to present this to you as like, hey, this is a problem. We don't know what to do. It's like, here's the problem. You know, we've had, we've had an issue here. This is what we're doing to resolve it. Just we're letting you know. And you know, that is the way we've operated as a company with that deal of transparency. But if you are a Dell HP Supermicro customer, you get these same delays, but it's just like, you don't have any transparency about what's going on. It's like, where's this thing I paid for? It's like, Oh, you know, I don't know. No one will tell me. It's like, no, no. We want to be sure that we are giving people that kind of transparency. ([Time 1:19:37](https://share.snipd.com/snip/d352ae08-9a7f-4b72-b867-5ff30ccc3ad9)) - **Target 8–12 Week Lead Times** - Oxide aims for an eight-to-twelve-week delivery lead time and prioritizes reducing lead times through capacity expansion. - Bryan notes they've generally moved faster than customers but growth tightens that margin. Transcript: Bryan Cantrill If you were to buy those right now, I mean, our goal is to be at like eight to 12 weekly times. So we would be, and in general, we have been, yeah, exactly. In general, we have been able to move faster than our customers. Mostly that is no longer becoming the case. As the numbers get bigger and the urgency gets higher, that is not going to be the case. But it has been the case that like, no, like we're ready to ship it. We need an address from you. They're like, oh, we still don't know which DC it's going to and, you know, or what have you. So it's like, we've been able to block on, but yeah, we want to be at that kind of eight to 12 week kind of lead time. ([Time 1:27:10](https://share.snipd.com/snip/8400d258-049c-4e5c-ad88-0c3e2ea57ba2)) - **Several Deep Technical Breakthroughs** - Key technical breakthroughs: booting x86 without BIOS, a programmable switch integrated at rack scale, dynamic distributed updates, and a custom service processor/root-of-trust. - Bryan says these under-the-hood advances enabled Oxide's product and are rare in the industry. Transcript: Bryan Cantrill I mean, I, I, what we've been able to do, it is just sad, a sad state of the industry that the bias, a, a construct that dates back to when I was a child and I'm, you know, 51 years old, uh, being Able to boot an X86 without a bias is like a man on the moon. And that was extraordinarily difficult. And that was one of those things like you don't pull it off, you've got nothing. You don't have a product at all. Being able to, it's the approach we took to holistic boot is something that the hyperscalers haven't even done. The approach that we took to the switch, the programmable switch, we, and being able to integrate that in with the rack is extraordinary. And I can't believe we got it to work and then Intel killed the part, right? Intel, we were using Intel Tofino. Intel killed the part. And so we have had to go to, we had to go find another part. And we were really excited about the part we found at the Excite Labs called the X2, which we're really excited to go build on. And then the stuff we've been able to build on that X2 is very exciting. That's not what we're shipping today. It's really extraordinary. The things we've been able to do with the control plane, with this distributed system that was, you know, we had done several control planes in the past and really wanted to improve on Them. And actually seeing that come to fruition, being able to do an update and being able to dynamically update this thing, because we take a distributed system that we update across an air Gap. The fact that we are, and when we shipped our first product, it was important that we could update this thing at all. We had something that we called the update, the minimum update. That minimum update required the rack to be parked. You would stop VMs, update the software, and start the VMs again, which is great that you could update it at all. Obviously, that's not the cloud experience. Um and the the ability to build a true distributed system to allow that thing to be updated truly dynamically and to watch that unfold over i mean that that is a that was a long long long Software project and required uh just extraordinary work from a bunch of people and and and And so watching that come to fruition is amazing. What we've been able to go do on the service processor and getting rid of the traditional BMC, we've been able to go do with the root of trust, having a true root of trust, being able to attest All the way up the stack, having the root of trust truly attest the SP, having the SP attest what runs on the host CPU, pulling that together into a trust quorum where you have, I mean, what We're pulling off is really extraordinary where you get this group of sleds together and they themselves decide there are enough of us to share a secret. So in other words, if you just walk off with a sled, you don't have anything. The ability to do this with distributed storage, the ability to stitch all this together and then get the power observability that we had always... There's so many things that I had always wanted and I didn't understand why I didn't have them. And I understand why I didn't have them because we, it's really hard. It turns out a lot of this stuff is really hard, but to be able, you know, the vision for oxide dates back so long. It dates back, you know, for aspects of it over 20 years. And the, and it's not just, you know, it's a vision for a lot of people have had. And watching that come to fruition in so many different dimensions is extraordinarily satisfying. ([Time 1:29:31](https://share.snipd.com/snip/627bd172-836a-4de7-b434-00c717786685))