Nobody is Noticing Nvidia's Next Move

This guy created 800 fake doctors to generate $1.8 billion in revenue with one employee. Okay, so this, this guy, this guy, Matthew Gallagher. Okay, so he made 800+ Facebook accounts for fake doctors to advertise on Facebook and went on to build a GLP-1 telehealth company with just $20,000 to start with, AI, and only one full-time teammate, his brother. So it generated $401 million in 2025 and could reach $1.8 billion in 2026. So this is the guy over here and then these are all the ads that he's running. I don't think so. I mean, he, this GLP-1 that he's doing, a lot of people are doing it. They're basically like affiliates for this GLP-1 company and he's just going hard on the doctors. And he was responding to this. He was saying like a lot of people do this. So I don't know if that's okay. He says a lot of people do it. I don't know if it's okay.

I don't think that's okay. Creating fake doctor profiles. Typically when you can make money in marketing that fast, something's wrong.

Easy come, easy go. Yes.

And it's going to get shut down. So, but if you made your money really quickly, Great for you, but me personally, I wouldn't want to do something unethical like create fake doctor profiles.

I think we're going to see more, maybe not exactly like this business model, but I think we are going to start to see more one-person billion-dollar companies with one employee, like just because he used AI to write the code, which isn't really that sophisticated, produce the website copy, generate the— I really think his main wedge was he pushed hard on AI ads piece because someone else on X responded saying, I'm in this space as well. Everyone knows this guy, he's just super aggressive with the ads. And that's what you need to do as an affiliate. You need to be hardcore. That's how the biggest affiliates win.

Yes, but you also have to be careful 'cause if you're super aggressive, you can get sued too 'cause you're probably making claims and promises that you shouldn't be.

Yes. And so we're not saying you should do that, but I think we're saying be aggressive and you'll be able to do well.

Or let me rephrase what Eric is really saying. He's not saying be aggressive from the aspect of creating false promises in ads.

Yes, don't do that.

He's saying be aggressive with AI and have it help you scale up faster. Sorry for interrupting.

No, no, no, you're fine. And speaking of, because we're talking about AI right now, have you seen— I've been reading these Kotu charts. I think it's worth it for— I'm like, I really like how they do their charts. They have their blog is just all these. You've seen these, right?

Yeah.

So, okay, this Kotu chart over here, those of you that can't see it, I'm going to explain it. Neil can re-explain it as well. So if you look from 1900 to 1960, this is the adoption of the tractor. Okay, so the tractor decimated farm employment, but led to the greatest expansion in US history. Okay. Here, you have 12 million agricultural workers in the beginning of 1900, and then manufacturing starts to take off. And then this guy Keynes, who's the economist, we are being afflicted with a new disease of which some readers may not have heard the name, but which they will hear a great deal in the years to come, namely technological unemployment. So the tractor becomes widespread in 1917. And then what happens is you see agriculture, the green line over here continues to go down. All the way to 1970. So it goes from 12 million to 3.5 million, which is 2% of the workforce, from 41% to 2%. But what ends up happening, Neil, is employment actually goes up. So job displacement happens, okay, from this period, 1970 to 1940-ish, but it goes up over time, right? And that's the same thing with the bank, the ATM. The ATM, when it came out, people were like, oh my God, everyone's going to lose their jobs, blah, blah, blah. There's a period of time where like, sure, there's some job displacement, but after what happened? More banks opened, which created more employment. So I think short-term, yes, fear. Long-term, everything's going to be okay.

I'm with you. And I think Marc Andreessen talked about how he believes that we're going to see a massive boom because of AI. And I agree with that. We're seeing it actually create more employment opportunities than it did before. Yes, it's also reducing employment, but you know, it's about how can you shift people from a job that can easily be replaced, retrain them, and put them into something else.

Here's what I'll say, Neil. This is just my observations. I will say that everyone on my team is a good person, right? I will say the interns are completely cracked out right now. Okay, so the interns are extremely hardworking, all in on AI. Okay, that's one piece. The second piece is the engineers and the product people all in. Okay, when you talk to them about this stuff, they're just working around the clock. They are working until their eyes hurt, right? And so I actually think, so I have 2 NVIDIA DGX Sparks arriving today. So I'm gonna be working on that over the weekend. I think my token costs, Even though I'm reducing it with this infrastructure, I think it's going to go way higher because everyone's starting to use it like this. And so what's going to happen is I was, I was doing the math. I'm like, man, maybe even 36 months out, I'm going to need the power of 336 H100s from NVIDIA. And by that time it's going to be something else. But I'm just— this thing's called Jevons paradox. When electricity came out, people started to use it more. Internet came out, people started to use it more and more. AI came out, people are going to want to use it more and more. So our costs, it's, I think it's going to skyrocket. It might even be equal to or more than our headcount costs.

Yeah.

Have you mapped that out yet for your company?

No.

I'm sure your people have.

They may have. I have no idea, honestly, if they mapped it out or not.

Yeah.

I just, dude, it's just fascinating on how many things are changing so fast. Like you're, we're coming to inflection point where AI is really useful in corporations because of agents. Everyone talked about agents and how they're going to change consumers and, oh, they're going to shop for you and do all this stuff. I think the big revolution with agents is not going to be the consumer side. I actually think it's going to be the corporate side. On the consumer side, yeah, people want an agent to be like their little mini assistant.

Mm-hmm.

But on the corporation side, I think it's just going to change the world on how businesses do work. But the hard part that we're seeing, Everywhere we go, dude, everyone wants to move 10 times faster now. We're not seeing people talk too much about like, hey, how do we cut costs? They're talking about we need to do 10x more. And that's the mentality that I think is the correct approach.

That's the vibe. So like one of my friends messaged me, he's like, so I wasn't even telling him to buy anything. I was just telling him what we're working on. And all of a sudden he's like, the answer is yes, we need this. And so whoever, like, I think if you're working right now, you're like, man, there's not going to be that much opportunity. I'm telling you the opportunity is going to be insane for everybody. I think a lot of people are going to win. And I just, by the way, I've been mapping out these charts, Neil, check this out. So here's the infrastructure scale, check this out, Neil. So this is the infrastructure requirements for single brain. If we experience hypergrowth, it goes like this in terms of how much infrastructure we need in the next 36 months. Aggressive is like this blue line over here. conservative is like, it still goes up, but this shows how many H100 nodes I'm going to need. And by that time, it's going to be something else because the chips are advancing so quickly. We're probably going to have to rent this infrastructure and you're probably going to have to rent, which is why I'm like, oh man, maybe we should buy CoreWeave stock, not financial advice, but they have all the infrastructure. I think a lot of people are going to be renting this stuff.

Uh, yes. Sorry, my bad. Someone was texting me. They're like, you jacked a client from me. And I was like, I'm like, I don't know what client I jacked I'm sorry, we have like a ton of people in our company.

That's— do I know this person?

Yeah, you know them and I know them and they're a good friend.

Oh, are they in Orange County?

Yeah, they're in Orange County. And I'm like, I didn't jack a client from you. I'm like, dude, I don't know who the heck reaches out to me, you know?

Yep.

So, but yeah, I was like, what can I do?

What can you do? I can't.

It's like, this is what happens when you have a company and people reach out and then like I get blamed for jacking a client when I didn't even—

Oh, that means it's okay for me to poach from Neil. Just kidding. Just kidding.

No, dude, do whatever works for your business.

No, I don't do that. That's, that's screwed up. Remember, we have, we know someone that has done that to you multiple times, and that's not cool because this person came to you for advice and was your friend. That was employees. Yeah.

No, no, but this person didn't. He's pissed off that I took his client.

I know, but I kind of look at them in similar veins, right? Like you just don't do that. But like in your case, like this is not something you were like aware of. Like if you were aware of it, you wouldn't have done it. This person, willingly did it to you on the employee side, right?

Yes.

So yeah, but by the way, this is maybe rosy, but it shows how revenue scales and infrastructure scales. The revenue is the blue, so it scales a lot faster than the infrastructure cost. So I'll pay for this all day. So yeah. Anyway, you can pick.

I'm just hoping all these costs start going down because we're starting to see our financial bills or, you know, we had an internal meeting about reducing our AI costs. Like it's getting outta hand.

But why would you reduce it if it's, you're paying for intelligence that can do more for you and the costs are gonna continue to come down?

Not necessarily. Um, just because AI is gonna be more efficient and just because the cost for using these platforms and technology's gonna come down, there's efficient ways to use it and there's inefficient ways to use it.

That you should reduce.

Yes. And when we look at how some of our team members are using it, mm-hmm, there's much more efficient ways where they can be using it and your costs go down drastically.

Yeah.

Like, dude, You know, there's a lot of things you can do on-premise on your own machines or buy machines, and it's just way cheaper than paying some of these guys money. So it's just like, okay, do we want to end up putting in all this effort to set up our own machines? And I'm like, yes, because we're spending so much money on a monthly basis. I'm like, the recuperation time is not years. It's really quick, like a month or two. This is like, just pay for this.

Quick break. Look, I know what you're thinking. Another AI content tool. More garbage content on the internet. And I thought the same thing. That's why we spent years building ClickFlow differently. Here's actual feedback from a user. It's sophisticated, grounded in real language, authoritative, but not academic. You've hit the sweet spot. That's not AI slop. That's content that you would actually publish. ClickFlow also helps with things such as internal linking, building FAQs, and reporting on the content performance. If you're skeptical, you can just go to clickflow.com and try it for free for 14 days. And if it sucks, just cancel, but I don't think you will. Back to the show. I'll explain it this way. Like, so here's how we save on AI right now. So the way we save on AI is 15% is Ferrari cost, okay? So you have Ferrari models, which is like the frontier models, and then the other 85% is Honda Accord budget, okay? So that's where you like, that's a daily beater that you use to go get your groceries or whatever. So most of the time your people are asking like normal questions, like basic queries, like going back and forth. Maybe sometimes I might ask stupid questions. That goes in 85%. The 15%, the Ferrari budget, that goes into you actually having the model think for you and code for you. Okay. And that's how our infrastructure is like, the cost is breaking down right now. And you might use tools like OpenRouter to route to the right model and that'll save you more on cost. They'll charge you like 5% or something. That's pretty good. But also, you know what you can do with your model? Here's the hack. If you tell your model to speak like a caveman, that saves you a lot on costs. So it's like, me code this now.

Yeah.

Me need this. I'm not even joking. It cuts down like 50 to 70% of costs. I'm not even joking. Yeah. It probably is.

I haven't actually looked at the breakdown, but it probably really is.

Yeah. Yeah. So like—

And then what do you think of the outputs? I don't think people are getting the outputs they really want, but they're burning the money.

Oh, here, I have an example. So there's this gal, Claire Vo, that shares this on her Twitter. So her open callout is full caveman right now. So here, check this out. Oh my God, she tweets a lot. Okay, here we go. Check this out. Need find files. Need maybe scripts in another dir hidden, directory hidden grep. Need locate script, maybe grep with find plus grep. Need likely there's another script for mention monitor generating worth engaging. Maybe system uses xsearch. Search worth engaging.

What we're starting to analyze on our end is how much are people using internally? What's the output that they're getting and how much of the output do they actually use? Because there's a big problem right now, at least what we're seeing in marketing departments. Yes, people are using AI. Yes, there are some efficiencies to be had and you can say you can save on employees or whatever, but you're paying these LLMs and it's not cheap and what you're finding is People are using them for whatever they want, and a lot of the stuff that they're using AI for, A, they may have not needed to use in the first place, and B, the outputs that they were created, they aren't using it because they weren't happy with it, but sometimes they are using it, sometimes they're not, so if you start looking at the wastage and how much money you're spending on that, it's adding up for like big companies to be millions of dollars a year. And you know me, I go through credit card statements line item by line item and expenses. And I'm like, dude, there's just waste. How do we cut the waste?

Dude, the good news now is we started getting paid for a sponsorship. So that's why you're not getting bills for a while. So for this stuff.

We have been for a while now.

No, they just started paying. We just got the first payment on that one.

Oh, but we were doing it for a while. We just started receiving the checks is what you're saying. But before then, dude, we had sponsors for a while. Did she ever pay up on all the other stuff? Or did the other company ever pay up on the other stuff?

We didn't get that much. It came over to cover our costs. We didn't get a lot. We got like, she would send us like payments for like $10 sometimes.

But I thought we were doing all these sponsorships and she's like, look at all these contracts that I got.

We were declining a lot. We just didn't want to do these things. Yeah, it was stupid. So, but we did like, we did cover all of our costs going through her. We didn't make a lot though, which is why we came off of it.

It was a terrible deal because her people promised the world and they didn't deliver on much.

Here's what I'll say about that. Then I want to come back to what you're saying. And I'm starting to forget what you're going to say too, but like, what were you just talking about before I cover that?

AI costs.

Okay, so we'll come back to AI costs. So the partnership, so what had happened, Neil, was I was talking to a bunch of people and a lot of people had a lot of followers. And even when I talked to the HubSpot team, that team, like when they heard about this team, they're like, oh, like, I'm like, why are you guys reacting like that? It's like, it's not good to work with. And everybody I've talked to is like, it's not good to work with these people. We tried it, didn't work out. You know, we have our own experiences, but we'll leave it at that. So AI costs. So on the AI cost side, what we're doing is one of— because we're putting everything into Nemo Claw, right, which is the enterprise-grade version of OpenClaw. And we're saying, okay, we want it, we want the agents to help with performance intelligence, which is where it will coach people. Okay. It will see how they're using it, how they're asking questions, how often they're asking questions, how it's getting coaching and all that, and how productive they're being with this in general.

Right.

So You know, in my mind, I'm like, okay, if you have this world intelligence, this single brain, and I think every company needs a single brain, if you have the single brain, it understands the entire company, all of the goals, and it works with every single person, it coaches them up as well, then everyone should grow. And then also it's like, okay, based on how you're working with all these people, in what ways are we using this in a stupid way, and how do we save on costs? And then you can have the agent itself improve on saving you money. That's how I look at it because I'm constantly asking it. There's a command, Neil, called /cost. You would run this all the time. It just tells you how you're spending your money and how to optimize it. Yeah.

I didn't know Nemo Cloud was out already. The NVIDIA one. It's already out. When did it come out?

Like a month? When he released it, it was already coming out.

Oh, got it. I heard the announcement. I didn't know that it was already out.

Yeah. So that's why I have NVIDIA infrastructure. I'm like, oh crap, I need to buy a lot more of these DGXs.

Nvidia's just, their chips are so expensive.

Yeah. So I just, for 2 of them for myself, it costs 10 grand for, and here's the crazy thing, Neil, these DGX Sparks originally they were $3,900. Now they're $4,700. So I think the pricing on these is going to continue to go up over time.

Yes. I think they're going to go up over time. I think you're going to continually see costs go up.

And it's the weirdest thing to me, Neil, because when I was 9 years old, I bought an Nvidia chip. I bought a GeForce for my computer because This is what during the internet boom and during the gaming boom, okay? 30 years later, I find myself buying Nvidia chips again during another boom. So everything kind of comes full circle. Yeah. So—

What else do we have on? Oh, the 4 levels of AI marketing. I was curious on this one. I haven't read this one yet.

So this one is from Anthropic's single employee. Remember they had that single growth marketer? So he has mapped out the 4 levels of AI marketing use. So most people sit at level 1, automating what they already do. So level 1, Neil, automate what you already do. So reporting, copy, data pulls, like using it at a pretty rudimentary level. That's level 1.

Yeah.

Level 2 is you use AI as a thinking partner where it's better than you.

Okay.

So you might ask about tax situations or tax laws. I might ask about those as well, right? Or I might ask about, you know, certain implications around doing work, right? Level 3 is do work that was below the ROI threshold before. So an example of that might be, Performance intelligence, which I just mentioned. Like it's too much work to have to say, "Oh Neil, how did you? What questions do you ask? Say how'd you work? How'd you use your token? Say da da da da da. You couldn't do that before. This work is now all doable, right? A lot of these edge case scenarios you can do now. Level four: Build custom tools only you would ever build. Okay, so level three is work that never existed before. So stuff nobody did because the manual cost was never worth it. So mining negative keywords across every ad group, checking your full site for broken links daily. Same logic applies to content research, QA, competitor monitoring, and work that. existed in theory, but nobody had the hours for, which is kind of how we're using our thing for. Level 4, which we just talked about, is where the ROI compounds. So there are hundreds of AI marketing skills and plugins floating around GitHub right now. Most of them work in theory but fall apart in practice because they're built for general use cases, not your case. Your business has specific data, specific workflows, specific edge cases that no generic tool will ever cover. The people building custom tools around their own problems are the ones pulling ahead, which is similar to what Ramp is doing with Glass. They don't want to use any off-the-shelf software. They're building it for their solution. Their single-brain version is that we have our own single-brain version and then we're working on selling that. But I think every company has their own institutional AI that they need to have.

I believe the level 1, the automate what you already do, reporting, copying, data pulls, et cetera. I don't think people will be using the LLMs directly for that. I believe they're going to start using most of the software companies for that because it'll just be cheaper and more efficient.

I think they're going to be using a lot of these Chinese models, I'll tell you that much. So the Chinese models are really good and they're really cost-effective. Quinn, Kimmy, they're all amazing. Gemma, even from Google, Gemma's amazing. You can run it on like a crap computer and it's amazing.

Yeah, but I really do believe people are just going to pay for the software solutions.

Yeah.

Because some of these software solutions are like $10, $15 a month or even freemium. And it's just cheaper to do that because they'll perfect it. for that specific task or those marketing functions that you have. And it's just easier than you having someone internally using AI to do it and maintaining it and making sure it's doing it accurately. Did you see the Alibaba news, how they're funding a real-world version of AI?

No.

So Alibaba, I believe, funded $200-something million round with a Chinese AI company where it's deciphering what's happening in real world and videos and all of that kind of stuff to make AI way more sophisticated. So instead of just text space is trying to analyze real-world stuff based on how we interact as humans. Because the way you read text is very different than how a human interacts with each other in person or through video.

Yeah. Dude, this is funny. Okay. So you're talking about this, right? Like you just got me to think about how crazy people— I'm Chinese, right? So look, so people in China, they're all lining up in public for OpenCLAW to learn OpenCLAW. I don't see any of that happening in America. Where's that happening right now?

Wow.

Like, this is like embedded into our culture. It's like, oh, if you're gonna make money for it, we're gonna line up for this stuff.

What's interesting is the narrative in the US is there's gonna be a lot of job cuts, AI's gonna create a lot of displacement.

AI's bad, it's so negative.

Yes.

Yeah.

In China, I've seen, and all the articles I've read, it's actually a very positive sentiment. People don't worry about job displacement, they worry about, hey, how can I learn all this stuff and adapt so then that way I can be better for the future?

So let me give you the numbers here, Neil, 'cause we like numbers, right? So Chinese public sentiment towards AI is significantly more positive more optimistic and trusting compared to the US. What do you think the number is percentage-wise, optimism in China?

90-plus percent.

Close, 80%. That's like having pocket aces, okay? You're good, right? Now, what do you think it is in the US?

10%.

35%. 'Cause look, you and I are pretty optimistic. I think Noah's pretty optimistic about it too. I am optimistic about it. I'm like short-term doomer, long-term optimistic, right? Which is like, a lot of it's politicized here. It shouldn't be politicized. This is like intelligence in your hands.

Wait, you know, Noah, this just made me think. You have a business where you pay people to help you create content. Why don't you just use AI?

How could you not have thought about this already?

And AI just create the content? No, no, for him, for his business.

I know, I know, but how could— this is obvious to me, like in the beginning.

Yeah, but he's not doing it.

The difference is that the customers don't want that. They don't want AI.

Really?

Micah, we had to get a whole new set of customers. See, this is like our customers that want manual content only. Human-generated content, human-generated content.

Yeah. So it's just, it's a future path, but it's—

I think it depends on the user because there are customers, there's a .ai for that with the word. Okay. And they do well. Right. Yeah. I think it depends, Neil.

Yeah.

Yeah.

Yeah. I can see it. We work with a lot of companies and even my own company, I don't want AI content for my own business.

Dude, look at this. And then I'm going to move over to the next thing. So look, They're literally in the park. Okay. You have people in China in the park. They're all like huddled over like these laptops. People are on their phones. They're all trying to learn OpenCL. I bet you this is during the weekend too. How do you beat this?

Society-wise, it's really tough.

Yeah.

They're just, majority of the people here that I've met are not that hungry, sadly.

I bet you they're like this in India too. Your, your people. Yeah. So that is it for today. Please don't forget to rate, review, subscribe, and we'll see See you tomorrow.