Why I Spent $70M to buy AI.com (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) Chris, good to meet you. (00:00:01) >> What's happening? (00:00:02) >> Nice to meet you guys. How are you? (00:00:04) Thanks for (00:00:04) >> I love the background. The logo is (00:00:07) looking great. (00:00:07) >> Is that real or is that AI? (00:00:10) >> It's totally made up. [laughter] (00:00:13) >> Well, (00:00:13) >> I was like, it feels feels pretty quick (00:00:15) to have, you know, the actual (00:00:17) >> design manufacturing. It takes time. (00:00:19) Well, take us through the thesis for (00:00:20) >> AI.com. Wait, before we Yeah, before we (00:00:22) get into the Super Bowl, like let's (00:00:25) let's maybe rewind to maybe probably a (00:00:28) year ago. you see a domain on the (00:00:30) market. I think they had been trying to (00:00:33) sell it for a while, I imagine, and uh (00:00:36) you came in as a buyer, but walk us (00:00:38) through that whole journey. (00:00:40) >> Yeah, I bought another domain and the (00:00:43) agent who was brokering this um told me (00:00:45) about AI.com being (00:00:48) um you know, in a process of of being (00:00:50) sold, if you will. So, I immediately (00:00:52) recognized the importance of it and just (00:00:54) jumped on it. got on the phone the same (00:00:56) day. Um, got the deal done. We shook (00:01:00) hands. Um, there were some ups and downs (00:01:02) through through the through the process, (00:01:05) but we managed to uh to get this done. (00:01:08) >> Uh, you've had some good success buying (00:01:11) uh buying iconic uh domains. What did (00:01:14) you pay for Crypto.com? Have you ever (00:01:16) disclosed that? (00:01:18) >> I don't think we ever did, but we paid (00:01:20) $12 million. Wow. And I I would argue (00:01:24) that that was um a more difficult (00:01:27) decision if you will. We were a small (00:01:29) company back then. (00:01:30) >> Um $12 million was about a third of our (00:01:33) capital. (00:01:34) >> Wow. (00:01:35) >> And it was bang in the middle of the (00:01:39) 2018 bare market. So people were (00:01:41) discussing whether crypto is going to (00:01:42) survive or not. (00:01:44) >> Yeah. The person selling it to you was (00:01:46) like probably like this guy is (00:01:47) >> I'm so glad to be washing [laughter] my (00:01:48) hands. (00:01:49) >> Of course. Of course, you uh ended up (00:01:51) ended up looking (00:01:52) >> on the Staples Center. I drive by it all (00:01:54) the time. (00:01:54) >> Um, okay. So, yeah, you you see this (00:01:57) domain is on the market. Do you have any (00:01:59) idea what you wanted to do with it at (00:02:01) the time of acquisition or do you just (00:02:03) think that, hey, this AI thing is (00:02:05) probably pretty important. Maybe I (00:02:07) should own AI.com. (00:02:09) Oh, look we um we were building products (00:02:12) ever since u chat GPT launched and uh (00:02:16) playing with uh consumerf facing side (00:02:19) but also internal tools. So constant (00:02:21) experimentation (00:02:23) um and the vision from day one was uh (00:02:28) you know we want to build a consumer (00:02:29) product. We believe that you will have 4 (00:02:33) billion people having personal (00:02:35) assistance that should play the role of (00:02:37) kind of a chief of staff for your entire (00:02:39) life with uh with great context um being (00:02:44) proactive getting things done for you (00:02:45) rather than just chatting. So the vision (00:02:47) didn't change from day one. (00:02:50) >> Uh and then yeah I guess fast forward (00:02:53) how did you how did you proc before we (00:02:55) get into the kind of Super Bowl and that (00:02:57) whole strategy? How did you how did you (00:02:59) process kind of the open claw launch? It (00:03:02) sounds like you guys are integrating the (00:03:05) uh are you building (00:03:06) >> some of those patterns at least? (00:03:07) >> Yeah, at least some of those patterns. (00:03:09) Are you kind of forking the project? (00:03:11) Talk about that. (00:03:13) >> So, as we started building this product (00:03:17) about mid 2025, it was never just (00:03:22) it just didn't click. It didn't have (00:03:23) this you you couldn't pass on Cany (00:03:26) Valley. uh and I think the pivotal (00:03:30) moment uh where we started seeing this (00:03:32) change is uh Opus 4.5 release (00:03:35) >> it started working much much better and (00:03:38) uh we obviously saw cloudbot going live (00:03:40) and adding some elements of their (00:03:42) architecture to it (00:03:44) >> I think it gets it done in terms of how (00:03:48) it feels as a product you just need to (00:03:51) solve a whole litany of issues to make (00:03:53) it consumer friendly like how do you set (00:03:54) it up without being technical you know (00:03:56) the security issues around your data. (00:03:58) This those are serious serious issues (00:03:59) when you want to bring something to the (00:04:01) mass market. So uh I think we've uh (00:04:06) we're combining everything that we've (00:04:07) built over the last say 8 months into (00:04:11) something that we can roll out to the (00:04:12) end users and we're starting doing this (00:04:14) tomorrow. (00:04:15) >> Wow. (00:04:16) >> Tomorrow. Okay. Uh before we get into (00:04:19) the product and kind of more of the (00:04:20) vision, let's fast forward again to the (00:04:22) uh to the Super Bowl specifically. How (00:04:25) did that how did that all come together? (00:04:27) It it felt like it was coming together (00:04:29) quickly, but we know we ran a much much (00:04:31) smaller ad. We ran a regional ad. Uh you (00:04:33) do have to lock these things in ahead of (00:04:35) time, but uh walk us through the process (00:04:38) of kind of preparing uh and then (00:04:41) experiencing uh the the Super Bowl (00:04:44) Sunday. (00:04:46) >> So I bought the domain in April. (00:04:48) >> Um the deal closed and we got the domain (00:04:52) successfully. like, okay, we need to (00:04:54) launch this and it deserves um a global (00:04:57) stage. And in May, we were one of the (00:05:00) first companies to actually buy the (00:05:03) spot. (00:05:03) >> Oh, no way. (00:05:04) >> Wow. (00:05:04) >> At that at that time, (00:05:07) we had just the domain and idea. What do (00:05:10) we want to do with it? But the product (00:05:11) didn't exist. (00:05:12) >> Mhm. Um, and you know, I I know that we (00:05:17) only have one shot to get this done uh (00:05:19) correctly, and I didn't want to release (00:05:22) the product until I felt that it's there (00:05:26) uh for the end user. (00:05:28) >> Uh, you know, these things uh (00:05:32) you need to be able to develop an (00:05:33) emotional connection with the product in (00:05:35) order for this to be sticky and (00:05:37) retentive. Uh, so (00:05:41) [clears throat] I only made a decision (00:05:42) that we actually gonna pull the trigger (00:05:44) on this um a couple weeks ago and that's (00:05:48) why the ad feel felt like it was, you (00:05:51) know, quickly put together. It was (00:05:52) quickly put together. (00:05:54) >> Wait, so you bought So you bought the (00:05:56) Super Bowl spot on Crypto.com if you (00:06:00) wanted to. (00:06:00) >> Yeah. What was the idea? Like, hey, if (00:06:02) if we don't run it for AI.com, we have (00:06:05) the Crypto.com ad ready. We'll just run (00:06:07) that. (00:06:08) We have the crypto uh business, we have (00:06:11) a prediction markets business, you know, (00:06:12) we could do there's always some level of (00:06:14) optionality, right? Um (00:06:17) but this is the moment to run an AI ad (00:06:20) as you guys have seen. It is (00:06:22) >> and and and and timing is really (00:06:24) important in life. You know, scale, (00:06:25) timing, and combination of these things. (00:06:28) >> It's a good reference. (00:06:29) >> Uh okay, so you put the ad together in (00:06:31) effectively two weeks and then you run (00:06:34) it and what happens then? Because I (00:06:36) think you got you got the the attention (00:06:39) from buying the world's most expensive (00:06:41) domain ever. Then you got the attention (00:06:43) of like, "Hey, there's this new AI (00:06:45) product I've never heard of with a crazy (00:06:46) domain running a Super Bowl ad. Should (00:06:48) pay attention to it." And then you got a (00:06:50) whole another kind of amount of uh of (00:06:52) attention from people being like, "Wait, (00:06:54) I just got an ad for AI.com and I landed (00:06:56) on the website and it's not and and the (00:06:58) website's down." So, what what what what (00:07:00) kind of happened? uh did you uh you (00:07:03) spent the the 70 million on the domain, (00:07:05) the 8 million on on these spot and then (00:07:07) you didn't have enough for to host it or (00:07:10) [laughter] (00:07:10) what uh what happened? I'm assuming a (00:07:13) lot of people got through but certainly (00:07:14) a lot of people got uh got stuck as (00:07:16) well. (00:07:18) >> Yeah, I think we are happy with the (00:07:20) outcome. (00:07:20) >> Okay, (00:07:21) >> we had about 300,000 people signing up. (00:07:24) >> Wow. (00:07:24) >> Wow. Let's go. (00:07:25) >> That's a lot. (00:07:26) >> Can we get the gong? Can we get the (00:07:27) gong,000 (00:07:28) signups? Yeah, there's been a lot of big (00:07:31) numbers. (00:07:32) >> That's a big number for a one day for (00:07:34) one day. Um, (00:07:37) but uh on a on a on a more serious note, (00:07:41) how did you uh how do you even prepare (00:07:44) for that that amount of traffic? Like (00:07:46) how do you how do you what was going on (00:07:47) in the war room? (00:07:50) >> Guys, you know, we've uh we're on a (00:07:53) platform that is used to spikes, right? (00:07:56) Um, and we've got a great DevOps team (00:07:59) and we've got all the stuff that you (00:08:01) usually would expect, autoscaling and (00:08:03) whatnot. So, there were intermittent (00:08:06) problems for some people, but largely it (00:08:09) held up. (00:08:10) >> Uh, so I I think fundamentally it's the (00:08:13) name and the fact that there's a certain (00:08:16) element of curiosity there and we (00:08:18) designed it in a it was a very simple (00:08:20) call to action. (00:08:21) >> Go and um and sign up. Yeah. (00:08:24) >> Um, so I think it worked. Yeah, talk to (00:08:26) me about consumer AI. Chachi PT has (00:08:30) broken through Gemini and Google, (00:08:32) they've been leveraging, you know, the (00:08:34) Google platform and the network to (00:08:36) onboard consumers. Nano Banana was a big (00:08:39) moment. Sora and the Meta Vibes app sort (00:08:43) of made a splash but haven't been super (00:08:45) sticky. Uh where do you see the gap in (00:08:48) breaking through with consumers in a new (00:08:50) way or just doing what consumers already (00:08:53) expect but better, cheaper, faster? Like (00:08:55) where is the consumer AI opportunity now (00:08:59) that we're three years into the chat GPT (00:09:01) boom? (00:09:03) >> I think fundamentally you're able to (00:09:05) actually get stuff done uh right now. So (00:09:08) that's a big differentiating factor for (00:09:10) the user experience. Um and we don't (00:09:14) really know where this experiment is (00:09:16) going to take us takes us uh given the (00:09:19) how the domain is um is resonating with (00:09:22) people you know we can introduce uh (00:09:25) social network elements to it. I think (00:09:27) the fact that every single person um on (00:09:31) earth is gonna have an assistant of this (00:09:33) sort (00:09:34) >> can unlock new type of interactions and (00:09:36) and (00:09:38) you know make our lives just better (00:09:40) through serendipity advice staying on (00:09:43) top of things and being proactive really (00:09:45) understanding us. So it's (00:09:47) >> I'm pretty pretty excited about the (00:09:50) wandering aspect of it. We try to keep (00:09:52) an open mind and not really be set on (00:09:56) one thing. We will uh we now have (00:09:58) 300,000 people waiting for us uh to give (00:10:01) them the product. Uh we're going to very (00:10:03) quickly iterate on it. I'm a huge (00:10:05) believer in moving quickly and (00:10:07) >> and and listening to actual customers (00:10:09) and we will see where the journey takes (00:10:11) us. I take a very long-term view. you (00:10:13) know what can we do with this in 10 (00:10:15) years it's I think it worked in the (00:10:17) crypto space and (00:10:19) >> the opportunity here the size is much (00:10:20) much bigger (00:10:21) >> yeah so much of what happened in crypto (00:10:23) was sort of permissionless you know uh (00:10:27) the bankless uh this opensource these (00:10:30) networks anyone can set up a node and (00:10:33) part of this latest uh open claw clawbot (00:10:36) you know hype cycle is driven by the (00:10:38) fact you like you get a Mac mini it can (00:10:41) talk to iMessage it can talk to (00:10:43) WhatsApp, it can talk to Telegram, it (00:10:45) can go sort of wherever you go as a (00:10:48) person. And that feels unique because (00:10:50) maybe OpenAI can't go over to WhatsApp (00:10:53) because Mark Zuckerberg doesn't want to (00:10:55) let him. And so I'm wondering about how (00:10:57) you're thinking about the trade-off (00:10:58) between certain things that are only (00:11:00) possible with an open-source AI system (00:11:04) that is sort of acting as an impostor as (00:11:07) a human versus you're a company. If you (00:11:11) want to integrate with WhatsApp, you (00:11:12) might have to give Meta a call. So, how (00:11:14) do you think about delivering the vision (00:11:16) of like a truly universal AI agent that (00:11:19) can do things with the realities of the (00:11:22) business community? (00:11:24) >> Yeah, there are a lot of um business and (00:11:26) UX challenges here. (00:11:28) >> Yeah. (00:11:28) >> So, um we'll have to resolve them one by (00:11:30) one. Sure. And our our view is we want (00:11:34) to stand on the side of the consumer and (00:11:37) help them make the uh these you know (00:11:39) technical choices uh make sure that (00:11:42) their data is safe make sure that they (00:11:43) can do what they want to do without (00:11:46) putting themselves at risk and solving (00:11:47) these issues with uh with access to data (00:11:50) with with user experience it cannot feel (00:11:54) like a chore. (00:11:55) >> Mh. (00:11:55) >> Uh you know today you need to be really (00:11:57) technical to get value out of it. (00:11:59) >> Mh. Uh so there's plenty of work and (00:12:01) it's difficult and that's part of the (00:12:03) opportunity. (00:12:04) >> Yeah. One of the uh uh some unrequested (00:12:08) feedback for me I had kind of heard that (00:12:10) AI.com was like potentially something (00:12:13) like some type of uh uh rel you know (00:12:16) leveraging some open claw technology and (00:12:18) then I got hit with a Google login and I (00:12:21) was like I don't have time to read (00:12:23) through kind of the the full terms of (00:12:26) service and privacy policy and really (00:12:28) understand. So I would love to see I (00:12:29) mean maybe there's plans for it but I'd (00:12:31) love to see just like being able to (00:12:32) create account an account on the (00:12:34) platform totally to play around with it (00:12:36) just because I was looking at my Gmail (00:12:38) which my life is on my work email which (00:12:40) obviously uh you know has its own (00:12:43) >> privacy concerns all that stuff. Um what (00:12:46) you're you said you're rolling out the (00:12:47) product tomorrow. What's the first thing (00:12:49) that you want people to do with it? (00:12:53) >> I think this is the big part of the (00:12:55) product. uh figuring out how do you (00:12:57) onboard people to it and get them to do (00:12:59) a couple of things so that they can see (00:13:01) value very quickly and connect with it. (00:13:03) I think (00:13:04) >> uh I think today it's pretty hard to um (00:13:09) to get the feeling for what it can do (00:13:11) without really connecting your email and (00:13:12) calendar. (00:13:13) >> Yeah. (00:13:14) >> Um we will see. I I feel that there's (00:13:16) going to be a lot of experimentation (00:13:17) there and we will look for user feedback (00:13:20) and and truth in data in numbers of what (00:13:22) really works and what doesn't. uh to a (00:13:24) certain degree you need to gify it until (00:13:27) such point where uh where users are deep (00:13:31) enough that it actually gives them the (00:13:34) feeling of like wow this is special this (00:13:36) is different (00:13:37) >> yeah yeah that gamification is so (00:13:39) important like you see the studio giblly (00:13:41) moment where you know the latest images (00:13:43) in chat GP launches and you don't even (00:13:46) have to think you just have something in (00:13:47) your camera roll and you're going to (00:13:49) type studio gibible and you're going to (00:13:50) get the value prop and then a couple (00:13:52) months later you're still going to be go (00:13:53) in there for slide inspiration and stock (00:13:56) photography and all the other things (00:13:58) that you can do, but there's a killer (00:13:59) app on day one that you come in and you (00:14:01) and you get joy out of. (00:14:03) >> Uh, did you ever talk to the original (00:14:04) owner of ai.com? I heard he sat on it (00:14:07) for 30 years because his initials, his (00:14:10) first and last name, his first name (00:14:11) starts with A, last name starts with I, (00:14:13) so he had bought the domain. Wow. (00:14:15) >> I'm shocked that he held on to it for so (00:14:17) long. You would think like IBM Watson (00:14:19) IBM Watson in like 2010 would like we'll (00:14:22) give you a million bucks for it. But uh (00:14:24) >> he held on. (00:14:26) >> Look uh we spoke um on the day when we (00:14:29) closed the transaction because he had a (00:14:31) it was a little bit of a bidding war. He (00:14:33) had a very serious bidder on the other (00:14:35) side and uh it required um (00:14:39) >> connecting in order to get it done. (00:14:41) >> That's good. your deals. (00:14:42) >> And by the way, and and by the way, um (00:14:45) right after um we closed the deal, (00:14:51) >> I got approached for the from the other (00:14:53) side offering 500 million plus, not for (00:14:56) 500 million for the domain. I think I (00:14:59) could have pushed it to a billion uh if (00:15:02) I wanted to, but I didn't want to. (00:15:04) >> So I think you guys need to you guys (00:15:06) need to understand I am pot committed. (00:15:09) >> I love it. I love it. We are committed. (00:15:11) I love I love that you're just thinking (00:15:12) you're you're viewing this like (00:15:14) obviously you're taking it very (00:15:15) seriously but you're also taking the (00:15:16) approach of like it's very early days (00:15:18) >> in you know uh what will be a long (00:15:21) journey for for the project but also the (00:15:23) industry and (00:15:24) >> uh you're just going to you know listen (00:15:26) to your users and figure it out but the (00:15:28) conviction to turn down what would have (00:15:29) been turning 70 million into uh 500 or a (00:15:33) billion in in 24 hours is uh is (00:15:36) admirable. Will you train a foundation (00:15:39) model? (00:15:43) I (00:15:44) >> think I'm more focused on getting this (00:15:46) to scale and getting the data flywheel (00:15:49) going so that we can deliver for our (00:15:51) users. Our users don't really care about (00:15:54) which model runs in the background as (00:15:57) long as the job gets done and their data (00:15:59) is safe. Uh but once you get to a (00:16:01) certain scale, (00:16:04) who knows? (00:16:05) >> Anything's on the table. I like it. (00:16:07) >> Well, I'm sure you'll be back on. (00:16:09) >> That's very (00:16:10) >> I'm excited to see this roll out and uh (00:16:13) >> I'm signing up tomorrow. I might be (00:16:14) using a dummy Google account, but I will (00:16:17) be signing up and testing this out. I'm (00:16:18) excited. And then I'll slowly forward (00:16:20) myself data for my real account to give (00:16:23) you a little bit more a little bit more (00:16:25) to see what it can do. But I'm excited (00:16:27) for for uh for uh for the launch (00:16:29) tomorrow. And congratulations. Uh I mean (00:16:32) a fantastic career but also uh this (00:16:34) particular project really really fun (00:16:36) execution and a wonderful story. So (00:16:37) thank you for coming to (00:16:38) >> Do you come you you're basically the (00:16:40) mayor of of Los Angeles through the (00:16:42) through the Crypto.com arena. Do you do (00:16:44) you come through much? (00:16:45) >> You're a Lakers guy. (00:16:47) >> Um I've been uh in DC last week and then (00:16:50) I stopped over in Silicon Valley. I have (00:16:53) never been to the arena. (00:16:54) >> Never been to the arena. (00:16:56) >> Wow. You got to come sometime. Catch a (00:16:59) game. May maybe we should uh catch a (00:17:01) game. (00:17:01) >> Yeah, they also do monster truck rallies (00:17:04) there. Underrated Crypto.com Arena (00:17:06) experience, especially if you have kids. (00:17:08) Big monster truck guy. Big monster truck (00:17:10) guy. I don't really follow basketball (00:17:11) that much, but I will be watching (00:17:13) Gravedigger live on the Crypto.com (00:17:15) Arena. Anyway, thank you so much for (00:17:17) taking the time. Have a great rest of (00:17:19) your day. and we'll talk to you