So a $38 billion deal with AWS just landed. That's huge, securing massive compute power. And for a company looking at, what, over $1 .4 trillion on infrastructure over the next decade? Yeah, $1 .4 trillion. It sounds absolutely staggering. It is. And that's not just spending money. It feels like they're building the literal foundations for whatever AI becomes next. It really is the price of admission these days, isn't it? Welcome,
everyone, to the Deep Dive. Our whole mission here is to get you past those splashy headlines and really dive into the core strategy. What's really going on in this, let's call it the Great Compute War. Okay. So today, yeah, we're going to unpack these massive deals, look at some of the flashpoints, you know, layoffs blamed on AI, where the legal lines are being drawn, and then we'll get to the really wild stuff like data centers in space. Seriously. Seriously.
So, yeah, this is going to be your shortcut to really getting this strategic. landscape of AI right now. All right, let's start with that $38 billion AWS deal then. Because for a while, it really felt like Microsoft kind of had open AI locked down, didn't it? It did seem that way. They put in $13 billion early on, had that first dibs thing on infrastructure spent. Exactly. But well, that old agreement, it's basically done, gone. What's really interesting, though,
is the type of money involved. How so? Well, yes, they still have this enormous... commitment with azure we're talking like 250 billion dollars huge number quarter trillion yeah but A lot of that is in cloud credits, and it's spread out over quite a few years. Okay, so it's not like a giant cash payment up front that stops them looking elsewhere. Precisely. That's the key. So this AWS deal, plus the fact they're working with Google and Oracle and even CoreWeave, it
shows they're absolutely multi -cloud now. They're free to shop around. They can look for the best GPUs, the lowest latency, specific network setups, whatever they need. It's smart. Stops them getting locked into one vendor. It almost feels like... Like a geopolitical strategy, not just business. Spreading the risk, spreading the power. Making sure no single partner controls their most vital resource, which is compute. That's exactly the strategic play. And, you know, this makes Amazon's
position fascinating. How is that? Well, Amazon is the main backer of Anthropic, right? One of OpenAI's biggest rivals in the LLM space. Yeah, a direct competitor. And yet AWS is now hosting both of them. On their servers. Whoa. So Amazon is trying to be like the neutral ground, the Switzerland of AI clouds hosting everyone, even the company competing with their own investment.
That's bold. It's a massive gamble because controlling that underlying compute infrastructure, you know, where the actual models run, that's rapidly becoming the new critical global infrastructure, maybe even like a sovereign resource down the line. And Amazon just wants to own the highway. Doesn't matter who's driving the trucks. Pretty much. And the spending. It's just ramping up this $38 billion to AWS. It's just one more step towards that massive $1 .4 trillion goal over the next
decade. That number. It's still hard to really wrap your head around. How do you even plan for needing over a trillion dollars worth of servers and power? Softly. Whoa. Imagine scaling that infrastructure to handle like a billion queries. Consistently. So why is this big shift to multi -cloud actually matter for future innovation? Well, thinking about it, it means less reliance on just one partner, right? And way more freedom to shop around for pure speed and capability.
Exactly. Less reliance, more freedom to chase performance. But, you know, scrambling for the hardware for the compute, that's only one side of this. The other side is how society, how the law, is trying to keep up with what this power means when it actually gets deployed. Let's talk about those flashpoints happening right now. And that deployment is moving fast. OpenAI's video tool, Sora, it's now live on Android across
the US, Canada, Japan, Asia. Big rollout. And we're seeing a whole new flood of AI -generated videos hitting places like TikTok. It's gone viral again. But at the same time, you see these headlines about layoffs. Big numbers. Over 60 ,000 jobs cut at huge companies, Amazon, UPS, Target. Yep. And the official reason often given is AI automation. Right. But the experts, they often point to something else, something they call AI washing. AI washing. Can you break that
down for us? Yeah, sure. AI washing is basically, well, it's a convenient story for corporations. It means blaming massive job cuts on AI, you know, this futuristic force, even when the reality is it's not purely because of automation. It might be standard cost cutting or market changes or restructuring. But AI is an easy scapegoat. Yeah. Makes it sound inevitable almost. Yeah, I can see that. You know, I still wrestle with
prompt drift myself. Like you tweak one word in a prompt and suddenly the AI gives you something totally. different. It's unpredictable sometimes. And this AI watching thing just highlights how much confusion there still is. Confusion about what the tech can actually do versus, you know, the hype around it. And that confusion. It's right at the heart of the whole public trust issue. Why do some people love AI and others
really fear it? It often comes down to how our brains handle risk and, frankly, how much we trust something we don't fully understand. If it feels confusing, it feels riskier. Makes sense. And the legal system is definitely being forced to react now. We're already seeing some clear boundaries emerge. Like what? Well, like you can ask GBT for health questions, although seriously, be careful. Double check every. Absolutely. Disclaimer
needed there. But you cannot currently build a regulated business that offers, say, legal advice. using that model. That's a really important line in the sand for commercial uses right now. And the legal heat is definitely rising. I mean, talk about drama. Did you see that moment when a guy actually walked up and served Sam Altman a real legal subpoena? When was this? During a live talk Altman was having with Steve Kerr, the basketball coach. It was apparently related
to some anti -AI protest case. Wow. That's intense. It shows this isn't just academic anymore. It's hitting the real world legal system and all. For sure. But even with all that tension, the money keeps pouring in for specific applications. Look at Hippocratic AI. That's a health care focused model. Right. I saw that. They just raised one hundred and twenty six million dollars. Big names backing them. Andreessen Horowitz, Google Ventures. They're valued at what? Three point
five billion dollars now. Shows there's still a huge appetite for specialized AI. OK, so given all this legal stuff, this tension. How quickly is the legal system actually adapting to these non -negotiable AI limits you mentioned? Honestly, slowly. It seems like it requires really clear boundaries, like avoiding using AI for regulated stuff like health or legal businesses directly. Yeah, slow and steady seems to be the pace for
law. Oh. Okay, let's shift gears a bit. From the serious stuff in boardrooms and courtrooms to the culture around AI, it feels like we've reached peak AI meme season. Oh yeah, definitely. The vibe check is strong right now. AI memes are everywhere. And honestly, the collected roasting of the big tech players, OpenAI, Anthropic, Google, Apple, it feels kind of healthy, you know? A way to process all the competition and maybe some of the frustration. It is pretty funny sometimes.
But then, then there's the weird stuff. The spine check trends, as the source called it. Apparently, there's this bizarre new wave of... AI chiropractors fueled by video tools like Sora generating realistic looking adjustments. Wait, what? AI chiropractors? Yeah. And we really need to say this clearly. Please, please do not use AI for medical advice and especially not for anything like spinal adjustments based on a generated video. Just don't. Okay. Yeah. That's definitely a do not try this at
home situation. Wow. But that absurdity, it actually highlights something important, doesn't it? What's that? When the outputs look this convincing visually. Some people will try anything, which is maybe why we're seeing creators push back in other ways. Oh, so? Like sharing really detailed AI performance charts, adding their own analysis of what these models can really do. It feels like a push for more transparency, for community driven benchmarks, not just corporate hype. That
makes sense. That need for reliable info, for real metrics. Yeah. It connects right into all the new tools popping up, too. This feels like where the democratization starts to kick in. Totally. We're seeing this explosion of really
focused. tools like a lay its whole purpose is making high quality presentations using ai takes away a lot of that painful slide building time yeah i could use that and loosen for video right you just chat with the ai tell it what you want and it generates edited video that cuts out a ton of manual editing exactly then there's skyvern that one sounds interesting automating workflows you do in your web browser think of all those repetitive clicks and data entry tasks and offices
oh yeah automating the boring stuff. Huge potential there. And my favorite from this list has to be Atom, text to 3D models. Yeah. And apparently with sliders, so you can tweak them easily after generation. Like stacking Lego blocks of data almost. You just describe what you want and poof, a 3D object appears that you can actually adjust. Really cool. So thinking about the culture, these tools. What do those weird trends like the AI chiropractors really tell us about the risk when
this stuff gets deployed? I mean, it tells me people will try absolutely anything, right? So any regulation, any safety measures, they really need to focus super hard on user safety first and foremost. Because assumption of sensible use. Probably not safe. Good point. User safety first. And this brings us kind of full circle back to infrastructure, but on a way bigger scale, a cosmic scale maybe. We know these massive AI models are already straining Earth's power grids,
and crucially, water. They need so much water for cooling those giant server farms. That strain, that resource pressure, it's forcing the big players to think, well, radically differently. Which brings us to Google's Project Suncatcher. Possibly their wildest plan yet. Which is? Launching data centers into space. Okay. You mentioned this earlier. Data centers in orbit. Tell us more. Where exactly? When? So the plan is pretty
ambitious. They want to deploy over 80 satellites, each one packed with AI processors powered by solar panels. And they're aiming for low Earth orbit about 400 miles up the target date by 2027. 2027. That's soon. But why space? It just sounds incredibly expensive and maybe environmentally tricky, too. Launching rockets isn't exactly green. Right. There are definitely costs, but the efficiency gains they're chasing are potentially enormous. That's the justification. Efficiency
how? Solar panels. In space, they're roughly eight times more efficient than on Earth. Eight times. Wow. Why? Simple. No clouds, no atmosphere scattering the light, no nighttime cutting off power, no shade from buildings or trees, just pure constant sunlight. Okay, that's a big energy advantage. Huge. Plus, the vacuum of space means you don't need vast amounts of land for server farms. And critically, you don't need millions of gallons of water for cooling. Space is its
own cooling system, basically. So you bypass the biggest resource bottlenecks we have down here, power generation, consistency, and water usage. That's right. Instantly. That's the idea. extreme efficiency. It's actually making the economics look surprisingly possible. Launch costs are dropping dramatically. Yeah, SpaceX
and others are making it cheaper. Exactly. So some projections suggest that by the 2030s, the actual operational cost of a space data center might match or maybe even beat the cost of running one on Earth. Still, that counterpoint you raised, launching rockets does create CO2. A lot of it. Are we just trading one environmental headache for another? That's the big question, isn't it? Launching is polluting. And astronomers are really
worried, too. They're concerned about all these new satellites cluttering up the sky, interfering with telescopes. They call it the bugs on a windshield effect. Bugs on a windshield. Vivid. And Google's not the only one thinking about this, right? This space compute race. Oh, definitely not. Elon Musk with Starlink and SpaceX is obviously a massive player in space infrastructure already. And there are startups jumping in, too, like
StarCloud. Apparently, they're planning to send NVIDIA chips up this month just for testing. This month. Wow. OK, so the race is already underway. It seems so. So final thought on this space angle. Will the environmental hit from all those launches end up canceling out the long term energy savings they're promising? That's the trillion dollar question. Maybe literally it might. But the company
is pushing this. They're focusing hard on those huge long term efficiency gains and getting rid of the strain on Earth's water and power grids. Right. It's a long term bet. OK. So we've covered a lot of ground in this deep dive today. What's the big picture here? What should we take away? I think the core findings are pretty clear. First, that compute war. It's leading to power decentralizing, moving away from just one single infrastructure partner. It's a strategic necessity for companies
like OpenAI. Okay. Decentralization. Yes. Second, public trust in AI is really fragile. It's under constant pressure from things like that AI washing we talked about and all these legal challenges popping up. Trust is a major issue. Right. Fragile trust. And the third thing. The third thing is just the sheer, almost unbelievable demand for compute power. It's so intense, it's literally pushing the industry to look off planet, to space, to orbit, to figure out how to power its future.
Decentralization, fragile trust, and looking to space for power. Got it. And knowing this stuff, it's really vital for you listening now to understand the strategies behind these huge tech companies. Their decisions today are genuinely going to shape how pretty much every digital service works over the next 10 years or more. Absolutely. Okay, so here's a final thought for you to chew on, something to consider after we
wrap up. Will any of these big cloud providers, AWS, Azure, Google Cloud, will any of them truly stay neutral? Switzerland. Once hosting, the core AI models become seen as the new essential global infrastructure. Ooh, that's a great question. Can neutrality hold when the stakes are that high? Something to think about in a world that seems to be changing at, well, orbital speed. Thanks so much for joining us for this deep dive into the sources today. We really hope you found
it useful. Keep exploring. Keep questioning. Yeah. Thanks, everyone. We'll get you next time.