저자: 미상 · 외국인
요약 보기
AI 기업들이 막대한 데이터센터 투자 비용을 감당하지 못해 정부 구제금융(지분 참여)을 논의 중이라고 주장. 샘 알트먼 발언, 트럼프 정부의 AI 기업 지분 언급, Anthropic의 개발 일시중지 제안을 근거로 든다. 기업들의 토큰 비용 부담 증가로 OpenAI·Anthropic 수익 압박이 커진다고 분석. 반도체·AI 관련 주식이 최근 매도세를 보였다고 언급하나 원인은 AI 단정 불가.
다듬은 글
Okay, things are getting crazy now. It looks like we're getting pretty close to a potential government bailout of AI companies. Now, this would be a massive problem for the global economy. I'll explain in this video, but we've got a lot to cover. So, the big TLDR is this. Sam Altman is saying, "Oh, actually, the cost of AI is becoming a huge problem." The stock market, if you guys follow those things, well, yesterday, if you're watching the video today, the date is the 6th of the 6th, 2000 and uh uh 26.
Uh yesterday, there was a huge stock market sell-off in a bunch of technology, semiconductor, and AI-adjacent stocks. So, what is going on there? Um however, there could be loads of things going on there as well. So, I don't want to be too like it's AI that's causing that problem, but it's something to to be aware of because it leads into the next thing of Donald Trump saying, "Yes, the government might take a stake in leading AI companies." Which, by the way, are all being invited to the White House uh to well, next week, according to Trump.
Hmm. Oh, and Anthropic say, "Yeah, yeah, we should all pause AI development now." Which is kind of convenient, right? Because they've got the best AI models uh by a lot of metrics at the moment, and they're rushing towards an IPO. So, it would really not be in their interest if a competitor was to get a model as powerful or maybe better than them as they're going for their IPO uh because, obviously, they want to retain their lead.
Like, what is going on? Okay, so, you know by now I'm not going to play the videos because I'll probably get copyright striked, and it'll be the end of the channel. So, this is Trump, and it's Trump saying, you can see here from the tweet from Reuters, um Donald Trump told reporters uh that his team might buy uh US stakes in artificial intelligence companies and said he would host a meeting with AI executives as soon as next week.
Hold on to that thought. Actually, Reuters also reporting uh Trump says US stake in AI giants could be a beautiful thing, which is very typical Trumpian language. Okay, so, remember, OpenAI has got an an 1.5 trillion, or maybe they've revised that down closer to 1 trillion, which is still an insane amount of money.
Um of capex spend on their data center, their compute build out, right? So, they need to find that money from somewhere. So, they need to sell services to recoup that money and essentially pay for it, right? Because obviously they're getting a ton of investment from external investors who they're obviously going to want to see a return.
And when this goes on the stock market, because they are going to IPO, although that date hasn't been confirmed and I presume Anthropic will hit before them, and obviously SpaceX hits before all of this. Um which again, SpaceX could be um an issue for the market sell-off that we saw the other day. It may not just be 100% AI focused.
However, Sam Altman, this is over on Business Insider. He says, um "OpenAI's top token spenders use 100 billion tokens a month, and they're not even the world leader." So, that's a lot of tokens, right? Now, this article, again, there's a link to this in the video description below. I'm going to scroll down and I'm going to show you the bit that I've highlighted, because this is this is where we're getting all of this from.
So, he was basically asked, right? So, he said, "He also said that the cost question came up quite suddenly. At the beginning of 2026, the issue never came up." Altman said. "People were totally happy with the amount they were spending." he said. "Now, AI costs are a huge issue." he said. So, why could they be a huge issue?
Now, this is where things get really interesting, right? So, you've probably been following what's been going on in the news. You've been watching a bunch of YouTube videos. You've probably been like me, right? You can't not consume this stuff. It is absolutely everywhere. The companies all over the world were racing to deploy AI in whatever shape or form they believed they needed it, right?
A lot of this is hype-driven. It's like, "Oh, we need AI. How can we insert AI into this project?" Like I said before, this is very similar to how can we insert blockchain? How can we insert crypto, NFTs, all of that kind of thing? And even like how can we insert dot com, if you go way, way back into the dot com boom era, where companies that had no business having any kind of online functionality were trying to add it into their business and were seeing no return.
Anyway, there's there's like literally thousands of examples of that as well. However, um if we bring it back forward to today and we look at AI, companies are racing to build up their AI um agentic systems, their AI whatever systems they're using. They end up removing members of staff. So, they get a period of actually this is great, right?
We can fire a few thousand people, bring on the AI and it doesn't cost that much more. This is great. In fact, it's miles cheaper. But, what's happening now is we're seeing a lot of people reporting and companies reporting that it's actually more expensive to run whatever workloads used to be run through people through the AI systems, especially the high-end models.
Now, this could actually break down into an entire different discussion and debate and I'm not going to go into too much detail in this video cuz we'll be here forever. But, there's this whole world now of you don't just want to use and this has kind of been in the world for a long time, but you don't want to just use the highest-end models, whether that's, you know, Opus 4.8 or, you know, moving into Mythos when that eventually launches.
Uh or any of the top-tier models, right? Cuz they cost a lot per token, especially the Anthropic models. They cost a lot. You can maybe use lesser models uh which are much cheaper for your workloads. So, you're not going to see I guess like what I'm trying to say is companies are probably trying to look now at like okay, the tasks that we need the high-end AI for, we're going to use that, that's fine, but we're going to reduce our spend at those companies, maybe spend it elsewhere or maybe use in-house models.
Now, the reason why I'm saying this is this is a big problem for Anthropic and OpenAI cuz they're the two leaders of the big high-end models. This will highly impact their revenue because they get their revenue from nowhere else. It is literally companies that are spending money with them. Consumers do not spend enough to make a dent in any of their services.
So, if companies start bleeding away, they've got a huge problem and guess what? They're then going to need to find funding from somewhere else or they're going to have a severe financing problem. Insider as well also saying the AI industry's profit problem is getting bigger, says Goldman Sachs equity research chief.
And this, like I said, is going to be an issue because it's not just about how much it costs per token of the model that you're using. It's about how much it costs to build the infrastructure to serve that model, how much it costs for the electricity, the build-out, all of this stuff, the hardware that's required.
And you know, we we can debate this forever because whatever direction AI is going to go in, a lot of these big companies, the likes of your Nvidia's of this world, your Amazon's of this world, your Facebook's or your Meta's, they want you to use their cloud-based AI systems. They don't really want you using local um models.
They don't want you downloading models. They don't want you using smaller, cheaper models. They want you using their models. Yeah, Nvidia's got some open-source models and things like that, but really they want you using the high-end stuff. Um but this is becoming a problem, right? Because I think companies are starting to realize, actually, we maybe don't need the best, highest-end model to run like a customer service bot.
We can maybe use something a bit cheaper. And again, just to drill this home, like this is Bloomberg and they're saying, "While AI labs promise uh return on investment, the reality is the technology costs more than ever, and more companies are deciding they can't afford it." Now, obviously notably as well, um Uber said they burnt through their token budget, their AI budget in like the first 3 months of the year, which is insane when you think about it.
But again, this all comes down to how is the tokens being spent? How is the AI budget being spent? Are they You know, like there's that meme, isn't there? It's like using It's like How can I explain this? It's like using something completely like imagine if you want to um I'm trying to think something off the top of my head.
Imagine if you've got a garden, right? Say you've got a really small garden. I don't have a garden. I've just got a concrete drive, right? But if you had a garden and it was really small, maybe 5 m squared, right? You're not going to use an absolutely massive like industrial ride-on mower to cut that, right?
It might do it in seconds, but it is absolutely massively expensive to use it and it's totally over the top. You're just going to use a little, you know, electric, you know, uh Flymo thing, right? Or something like that. Or you're going to use a little strimmer or something, right? Something that's really cheap, but it still gets the the job done.
I think this is the problem with all of this AI spend. Companies are using that massive industrial uh grass cutter or mower for for tasks that it's not really required for. Okay, I just want to jump over to this on CNBC cuz I quite like the little key point breakdown here cuz it's got an interesting point.
So, as you can see here, Trump administration OpenAI discussing possible government stake, yada yada yada. Yada yada yada. So, this here, it says OpenAI uh CEO Sam Altman and the White House are in ongoing talks about possible government stake in the company. Uh the AI startup could donate equity to the US government to seed something like the public wealth fund that the company outlines in its April policy proposal, according to a source familiar with the discussions.
The talks have been in progress for more than a year as Altman first shared the idea with the Trump administration in 2025. So, I don't think it comes as any surprise. I'm pretty sure Sam Altman OpenAI were very aware that they could not possibly recover the cost of the build-out. They are in a in a way, they want to become too big to fail.
This is very similar to what happened back in, if you can remember this, in the 2008 financial crisis. For me specifically in the UK, there was a bank called Northern Rock which went under. That bank went down and it was a massive problem. But a bunch of the other banks also heavily suffered, but the government bailed them out because this term was thrown around, they are too big to fail.
This is what Sam Altman wants. Now, in his mental utopian view of AI and how it's going to function in the future and how it's going to, you know, we need it, it's we desperately need this and all that stuff, he sees it not as a financial thing, not as a business. He sees it as a it's almost like um it's almost like prophecy, right?
This thing needs to be functional. It doesn't matter what the cost is. We're going to make this a public service. We're going to make it so the government cannot allow this to fail, which will continue our mission and we will keep marching forward unstoppably. This is crazy, obviously, cuz this is just totally insane because I mean, what are they doing?
All we know right now that AI is doing, especially OpenAI, is they're spending tons of money on data centers. They're getting loads of investment money. They want to go onto the stock market to then take people's investment money cuz it's not just from, you know, me or you going on the stock market going, "I want to buy some OpenAI shares when the IPO hits." It's about the ETFs.
It's about the investment funds. It's about pension funds all all automatically being pulled into this cuz they will be as time goes on. And then that means they almost get self-funded. I mean, not self-funded, but they they force funding of their stock to keep it at a certain level. This is very similar to what's going to happen, unfortunately, with SpaceX very soon.
But yeah, what the point in at this section of the video I wanted to make is I don't think it's quite interesting that OpenAI was thinking about this way back last year cuz I think they knew the magical money pot was running out because they got a very very bad problem of actually generating money. cost too much to make what they're offering.
Now, you've seen this with pricing systems. All of it was heavily subsidized. They wanted you to use OpenAI, to use Anthropic, to use all these systems at massively reduced rates. And the moment they turn over to token-specific billing, the bill goes through the roof and companies are going to get hammered cuz they're using tons of tokens and suddenly they need to get more efficient.
They can't use as many tokens. Why will they use that service? Maybe they'll look to using an open-source model on their own hardware potentially. It is a very interesting dynamic which is starting to form. And I think OpenAI identified that a long time ago. Okay, so remember Anthropic? Uh actually, by the way, this graphic is really good.
It's probably generated with AI, but look at this graphic when it runs. I think it's already run, so it stopped. Maybe if I scroll down it'll start running again. Hang on. Yeah, here we go. Look at this. It's like it builds itself, then it like zooms out. It's really cool. Anyway, um this is a uh I'm going want this a dossier from Anthropic where they're saying their AIs are getting very close to just completely recursively So, I mean, that's really cool animation.
Recursively rebuilding themselves, right? Any fans of Terminator 2 out there will remember the scene where Arnie's like, "Yeah, it became self-aware." Self- Self- Skynet became self-aware. Self- What a crap name that would be. Skynet becomes self-aware and it learns and then it just nukes everyone nukes Russia then Russia nukes us and nukes everybody else and then it takes over the control.
Anyway, this is Anthropic saying their models are getting very close to building the next model. And this is almost like one of those critical mass moments of when this happens, there'll be exponential growth. Now, exponential growth sounds great, infinite growth, but remember all of this is backed by electricity.
It's backed by well, hardware, data centers. If you don't have that, you can't just exponentially use compute. That doesn't exist, which is the problem. Like, and remember Anthropic have just done a massive deal with SpaceX to rent their OpenAI not OpenAI, the XAI Colossus. It's something crazy like a billion a month.
And I think Google have done the same thing as well, which again is mental. This is a massive article. I'm not going to go over all of this in the video. I'm going to link to it in the description below. But, I'm going to scroll down and you can see like it's just showing you how they think this is working or whatever.
I'll be here for all day if we're just going through this. But, there is a section that I've highlighted which I really hope has stayed highlighted. Cuz if it hasn't, we're going to be like, "Uh-oh." There we go. I'll do it. So, yeah, this is the Anthropic we should all slow down AI. Now, listen to this.
We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. The Anthropic and remember, they are going around with Mythos citing this as being cited and credited as the most powerful AI that's ever existed.
We believe this is going to be available publicly fairly soon, um, but we know this has been used by Anthropic obviously for a long time beforehand. Governments have had access to this. Um, there is discussion whether the US has been using this to conduct, um, cyber attacks on other countries. And, you know, banks in the UK have been like desperately trying to get access to this cuz they're all worried that their systems will be attacked and all of this stuff.
Um, so they say the Anthropic Institute will conduct research in collaboration with many others to take actions to help build the systems uh that a credible slow down or pause would require. These systems would enable frontier AI developers to verify that others globally have actually stopped or slowed and that a bad actor could not use the I I mean, we could just go on about this, right?
But this is like I this to me, it's fine if you are the leader and you want everybody else to slow down because you're rushing to an IPO. So, you want your company to be at the top the the biggest chance of it getting the most investment when it IPOs. The other thing as well is they can say this for well knowing it's never going to happen.
Why would anybody slow down? This is like a race to try and develop you know, general artificial intelligence or artificial general intelligence or artificial super intelligence, right? This is not going to slow down. So, to me, when I read this, it's like again, this is the Anthropic thing of we're on our high horse and we're oh, we want to do what's best for society while also relentlessly chasing this and relentlessly doing whatever we can to have the fastest models and the best models by benchmarks or whatever.
Um, and also remember that that same old news blitz of oh god, this is um a moment of absolute inflection where everything is just going to this AI is going to take over the world. And they say this for all of their models, right? And obviously it's been the biggest thing uh with Mythos saying it's this massive cybersecurity threat.
Nothing's going to be the same. All software's just going to get obliterated by it. And this is, you know, yet to be seen. Maybe it will happen. Maybe it won't. So, what do you think about this? Um, I this is all over the place, right? I think we're starting to see cracks in financing now, cracks in the economy.
We're starting to see the first um sort of public statements by the US government that they're considering doing bailouts for AI companies, um which is essentially what a direct government investment would be in whatever form that takes. AI companies wanted to become too big to fail. Um where do we go from there?
Because if the technology is not working, if the technology is not generating revenue, if it's not actually doing anything worthwhile, why are they spending so much money on the technology? It is mental. So, talking of worthwhile stuff with the technology, this has always been the reason why I've been pro AI, right?
From the very first moment I realized what AI was and what it was capable of. Yeah, I've said this before, but I've got ulcerative colitis, and I thought, "Holy they're going to be able to use this to very quickly design probably very DNA-specific treatments for certain people. That's going to be fantastic.
Um that's not happened just yet, but I think this is a a good bit of news that I'll I'll leave you guys with, that the world's first vaccine designed by artificial intelligence uh is on the way, and this apparently can target every single type of COVID. All right, guys, thanks for watching today's video.
I hope you enjoyed that. If you did, do let me know about more good news stories in the AI world. I do like to read good news. Um but yeah, I hope you enjoyed that, and I will catch you lovely lot on the next one. See you soon.
Comments 0