RunPod's AI Fork Success  ## Document Notes ## Highlights - **Basement Rig To $1M In Nine Months** - Jaeden Schafer recounts RunPod's origin from two Comcast coworkers turning an Ethereum mining rig into AI servers in a New Jersey basement. - They bootstrapped, posted on Reddit for beta users, and hit $1M revenue within nine months, then quit their jobs. Transcript: Jaeden Schafer So they actually started the company in 2021. They were both co-workers at Comcast, and they decided that they wanted to start a hobby, a side project, which is essentially that they built a custom GPU rig. This is in New Jersey, in a basement, and they were mining Ethereum with it. They made some money, apparently, but they didn't really make enough money to actually pay back the equipment, so they weren't recouping their investment or anything. And so they were kind of coming to the end of the mining hype. Ethereum's network upgrade was about to happen, if anyone remembers the merge. So that was about to happen. Anyways, they said that they got kind of bored of the project. Between them, they'd convinced their wives to let them buy about $50,000 worth of hardware. And so if they wanted to keep their wives happy, apparently they had to find a better use for the GPU so because of that um they both kind of worked on machine learning projects and they Decided to turn their mining rig into ai servers it was like way before chadgpt and even before dolly too so when they were rebuilding the system they ran into a big problem and they they Said quote we were seeing just how awful the software stack was for working with gpus um and so they decided that this is what they wanted to go and fix. And so that's kind of when they built RunPod. That's when it came together. Lou, the co-founder, said, quote, the experience of developing software for top GPUs was just hot garbage. So then by 2022, they were ready to launch it. They kind of positioned themselves as a fast developer focused AI app. They were a hosting platform. They had kind of this configurable hardware. They also had a serverless option and then a bunch of like APIs and integrations, all that kind of stuff. At the time, though, he said that the integrations were really minimal and it was quite challenging for them to actually go find users. So they said, as first time founders, we had no idea how to market, so I thought, let's just post on Reddit. So they started just spamming Reddit, basically, and a bunch of different AI-focused subreddits. They were giving free access in exchange for feedback. And actually, this worked. They got a bunch of beta users that turned into paying customers, and within nine months, they quit their jobs, and they'd hit $1 million in revenue. ([Time 0:01:57](https://share.snipd.com/snip/af795b2e-3d03-474a-81de-588402e13cea)) - **Growth Started On Reddit And Support Chat** - RunPod found early users by spamming AI subreddits and offering free access for feedback. - Reddit mentions and a support chat message later attracted investors like Radke Malik and Julian Chomond. Transcript: Jaeden Schafer Different AI-focused subreddits. They were giving free access in exchange for feedback. And actually, this worked. They got a bunch of beta users that turned into paying customers, and within nine months, they quit their jobs, and they'd hit $1 million in revenue. So with all of that, this is a funny quote from them. They said six months in customers started saying I want to run real business workloads. I can't do that on servers in someone's basement. And so that's when they decided that, you know, maybe they need to look at raising money, raising venture capital wasn't the initial plan. They had, you know, kind of partnered with some data centers on a revenue share deal to expand. Um, and that actually worked. They said it was a little bit tense, but it was working. They weren't, they didn't have to go raise money. Um, they said, if we don't have GPU availability, user sentiment flips immediately when people don't see capacity, they leave. So they said that their community was growing on Reddit and discord, especially after chat GPT launch. And at the same time, they said that's when investors started reaching out to them. So Malik reached out after seeing them mentioned on Reddit. And of course, they didn't really know how to pitch to VCs, but they walked him through their software. And for about two years, RunPod ran without any outside funding. ([Time 0:03:49](https://share.snipd.com/snip/deddfd6d-3c04-4208-b6f8-eccdab6ed091)) - **Bootstrapping Built Durable Momentum** - RunPod intentionally avoided outside funding for two years and avoided taking on debt unlike many crypto-to-GPU clouds. - That bootstrapped discipline positioned them to scale when AI demand surged and investors returned. Transcript: Jaeden Schafer So that also meant that there was no free tier, and they had to basically at least pay for themselves. So unlike a lot of other GPU clouds that kind of evolved out of crypto mining, they avoided taking on any debt, which helped them. And by 2024, all of that paid off. There was obviously this huge AI developing surge, and RedPod had about 100,000 developers and close to about $20 million in a seed round, which was co-led by Dell Technology Capitals And Intel Capital. And then they had some participation from Nat Friedman and Chumond. ([Time 0:05:02](https://share.snipd.com/snip/39b33ec2-d8f1-4d40-9f3f-e13732b76658)) - **Rapid Scale After Product Market Fit** - By 2024 RunPod scaled to roughly 100,000 then 500,000 developers and worked with Fortune 500 customers across 31 regions. - Their revenue grew to $120M ARR after a $20M seed co-led by Dell Technology Capital and Intel Capital. Transcript: Jaeden Schafer There was obviously this huge AI developing surge, and RedPod had about 100,000 developers and close to about $20 million in a seed round, which was co-led by Dell Technology Capitals And Intel Capital. And then they had some participation from Nat Friedman and Chumond. And they have not raised since then, since back in May in 2024, almost two years ago. But they said that they're preparing for a series A and their company has grown a lot. Apparently they have about 500,000 developers. They got a lot of individual builders, Fortune 500 companies that spend millions annually with them. And their cloud is in about 31 different regions. And they have some big customers like Replit and Cursor and OpenAI and Perplexity, Wix and Zillow. So a lot of huge players are using them. ([Time 0:05:19](https://share.snipd.com/snip/18d7bdc6-61a8-42ab-a6af-87b1d76266cf)) - **Intense Competition In GPU Cloud Market** - The AI GPU cloud market is intensely competitive with hyperscalers and specialized providers like CoreWeave. - RunPod must differentiate as a developer-focused platform to compete against AWS, Microsoft, and Google. Transcript: Jaeden Schafer What I will say about this whole industry though is that the competition is very intense. There's like hyperscalers, AWS, Microsoft, Google, who are specializing in a lot of these GPU provider types of things. There's obviously CoreWeave and CoreScientific. I think with all of that, I'll be interested to see if RunPod is gonna continue to grow and thrive. ([Time 0:06:04](https://share.snipd.com/snip/2b41a863-940a-4c33-ae91-2a0745b34b64)) - **Developer Experience As A Long-Term Bet** - RunPod positions itself as the platform the next generation of developers will grow up on by focusing on developer experience. - They view programmers evolving into builders and operators of AI agents rather than disappearing. Transcript: Jaeden Schafer According to them, they see software development evolving and not disappearing. The programmers, they argue, are going to continue continuously become builders and operators of AI agents. According to Lou, he said, our goal is to be the platform the next generation of developers grows up on. ([Time 0:06:26](https://share.snipd.com/snip/76ed372c-6716-4ee4-bbbf-d4f8e83ad9ca)) - **Prove Traction Before Raising VC** - Jaeden Schafer suggests founders should bootstrap as long as possible to prove traction before taking VC money. - He argues VCs prefer investing in companies with clear product-market fit and user growth to accelerate existing momentum. Transcript: Jaeden Schafer I think we're going to see a lot of these companies where startup founders are going to bootstrap for as long as they possibly can before they go and try to raise VC money. And I think VCs honestly prefer a situation like that where they can see a company that has awesome traction. They've got a lot of users, they have an awesome product, and then they'll be able to jump in and make an investment and help grow the company. I think the older model that we saw in the past was more someone had an idea, they got funded, and then it was very risky whether it would actually be successful or not. Today, I think we're seeing more investments going into companies. Obviously, all of the huge hyperscalers, all the big companies, those ones are getting, you know, huge rounds. But there are still a lot of new startups that are getting money. And usually that's when they're able to show that they have strong product market fit growth and a lot of users right off the bat that are using them. Those are the ones that will get the VC dollars. And that just is, you know, essentially there to help expand and kind of expedite what's already happening, not like, hey, we need money so that we can make a company and grow. I think it's so easy to build software today. And I know so easy as a lot of people will be raging at me for saying that case, not so easy. It is way more accessible to build software today than it has been in the past. And because of that, we're seeing a lot more software come out, whether it's better or not. I mean, you guys can argue about that. We were seeing a lot more software come out today. And so if we can build that, you can validate the idea, see if users are going to come on board. And once they are, you get the VC dollars and just kind of expedite what you're already doing. ([Time 0:06:50](https://share.snipd.com/snip/ede01d9a-09e8-417f-8acc-4354a89b7a2c))