The New AI Infrastructure Wars I 30th April

Microsoft's Satya Nadella just declared he's ready to 'exploit' the new OpenAI deal, while first responders are saying Waymo's getting worse on the streets. Meanwhile, early reports suggest Amazon's AWS is surging but burning through cash, and OpenAI is scaling their massive Stargate project to build the compute infrastructure for AGI. From courtroom drama between Musk and Altman to SoftBank's ambitious $100 billion robotics IPO play, today's episode unpacks the battles being fought over AI's future infrastructure. Who's winning the race to power the intelligence age?

Transcript

SPEAKER_00 0:00

If you found out your biggest competitor just got unlimited access to the world's most advanced AI models without having to pay a dime for them, would you be panicking or plotting your counter move?

SPEAKER_01 0:11

Oh, I'd be absolutely losing my mind. Like, that's not just a competitive advantage. That's basically being handed the keys to the castle while everyone else is still trying to pick the lock.

SPEAKER_00 0:22

Right? Because that's essentially what just happened to Google, Amazon, and every other cloud provider when Satya Nadella announced Microsoft can now offer OpenAI's tech to all their cloud customers without paying OpenAI a cent.

SPEAKER_01 0:35

And Nadella didn't mince words about it either. He literally said Microsoft is ready to exploit this deal. That's some serious chess level maneuvering right there.

SPEAKER_00 0:45

The question is whether this is brilliant strategy or whether Microsoft just painted a massive target on their back for every antitrust regulator on the planet.

SPEAKER_01 0:53

But here's what gets me. If I'm a CTO at any major company right now, how do I even begin to compete with free access to GPT-level AI? Like what's my board meeting presentation going to look like when Microsoft is literally giving away what everyone else is charging premium rates for?

SPEAKER_00 1:12

That's exactly why this story is so much bigger than just another tech partnership. We're talking about a fundamental reshaping of how AI gets distributed and monetized. You're listening to Build by AI. I'm Alex Shannon, and we just stumbled into what might be the most consequential AI business story of the year so far.

SPEAKER_01 1:33

And I'm Sam Hinton. Today we're diving deep into the infrastructure wars that are reshaping AI, from Microsoft's power play to first responders calling out Waymo, plus some courtroom drama and a mind-blowing robotics IPO that's already eyeing $100 billion.

SPEAKER_00 1:49

Plus, early reports suggest Amazon's cloud business is absolutely surging, but they're burning through cash like it's going out of style. There's a lot to unpack here.

SPEAKER_01 2:32

Okay, first of all, I love that Nadella used the word exploit, because that's exactly what this is. This isn't some gentle partnership handshake. This is Microsoft weaponizing their relationship with OpenAI to absolutely crush their cloud competition.

SPEAKER_00 2:48

But walk me through the mechanics here. How does this actually work? Because on the surface, it seems like OpenAI is just giving away their most valuable asset.

SPEAKER_01 3:00

No, see, that that's the genius of it. OpenAI isn't giving anything away. They're getting something even more valuable than per use fees. They're getting guaranteed compute capacity and infrastructure scaling that they could never afford to build themselves.

SPEAKER_00 3:16

Right. But doesn't this create a weird dynamic where OpenI success is now completely dependent on Microsoft's cloud business?

SPEAKER_01 3:23

That's the million-dollar question, literally. But here's what people are missing. This deal essentially makes Microsoft the exclusive distributor of the world's most advanced AI. You can't exactly match a price of zero.

SPEAKER_00 3:36

And meanwhile, Google Cloud and AWS are probably having emergency board meetings right now, trying to figure out how to compete with that.

SPEAKER_01 3:43

Exactly. You know, they're going to have to either develop their own models that can compete with GPT-level performance, or they're going to have to start making some very expensive acquisition deals.

SPEAKER_00 3:56

The thing that worries me though is the consolidation aspect. We're essentially watching the AI ecosystem get carved up by the biggest cloud providers.

SPEAKER_01 4:05

Look, I get the concern, but right now the infrastructure costs for competitive AI are so astronomical that only these massive players can actually support it.

SPEAKER_00 4:17

But think about this from a customer perspective. If Microsoft is offering you the best AI models for basically free as part of their cloud package, while AWS is charging you premium rates for inferior models, where are you going to go?

SPEAKER_01 4:32

That's exactly the customer lock-in strategy Microsoft is banking on. Once you build your entire AI workflow around their infrastructure, switching becomes incredibly expensive.

SPEAKER_00 4:44

And the timing is interesting because this comes right as enterprises are making their big AI infrastructure decisions for 2026 and beyond. Speaking of infrastructure, let's talk about OpenAI's Stargate project. OpenAI is scaling Stargate to build the compute infrastructure they say is necessary for AGI development. We're talking about massive data center expansion to meet what they're calling surging AI demand.

SPEAKER_01 5:09

This shows OpenAI is hedging their bets. They've got the Microsoft partnership for immediate scale, but they're also building their own infrastructure backbone. That tells me they're thinking much longer term than just being Microsoft's AI subsidiary.

SPEAKER_00 5:25

Right. And the language they're using is pretty bold. Intelligence Age Infrastructure. That's not incremental improvement talk. That's we're building the foundation for a fundamentally different era's talk.

SPEAKER_01 5:36

If you believe AGI is achievable in the next few years, you need compute infrastructure that's orders of magnitude beyond what exists today. Current data centers were built for search queries and video streaming, not for training models that might be smarter than humans.

SPEAKER_00 5:56

But doesn't this create tension with the Microsoft deal? If OpenAI is building their own massive compute infrastructure, why do they need Microsoft's cloud capacity?

SPEAKER_01 6:06

That's actually a brilliant strategy. Microsoft handles the immediate commercial deployment and customer distribution, while Stargate handles the cutting-edge research and development workloads.

SPEAKER_00 6:17

Although building AGI level compute infrastructure has to be phenomenally expensive. We're talking tens of billions of dollars, right?

SPEAKER_01 6:25

At least. But if they actually achieve AGI first, the economic value of that is basically unlimited. It's the ultimate go big or go home bet.

SPEAKER_00 6:35

The surging AI demand part suggests they're not just building for their own research, they're building capacity to meet market demand for AI services generally.

SPEAKER_01 6:45

That's smart diversification. Even if AGI takes longer than expected, they're still building valuable commercial infrastructure that can generate revenue.

SPEAKER_00 6:59

Keep watching Stargate's progress because if OpenAI starts hitting major infrastructure milestones, that's going to be a leading indicator of how close we actually are to AGI level capabilities. Let's talk about adoption, because early reports suggest Microsoft just dropped some pretty impressive copilot numbers. They're claiming over 20 million paid copilot users with growing engagement.

SPEAKER_01 7:22

20 million paid users is a pretty specific number to just make up. And honestly, if these numbers are real, that's actually a massive validation of the AI productivity tool market.

SPEAKER_00 7:33

What's interesting is the timing of this announcement. Microsoft drops these user numbers right as they're announcing this huge open AI partnership expansion. Coincidence?

SPEAKER_01 7:44

Oh, absolutely not a coincidence. This is Microsoft saying, look, we're not just making deals, we're actually delivering value to real users at scale.

SPEAKER_00 7:53

But let's talk about what growing engagement actually means. Are we talking about people using it once and forgetting about it? Or genuine workflow integration?

SPEAKER_01 8:03

Anecdotally, I'm seeing more and more people who genuinely rely on Copilot for coding, writing, data analysis. It's becoming less of a novelty and more of a utility.

SPEAKER_00 8:15

The broader implication is that if Microsoft can prove sustained, valuable AI adoption at this scale, it validates their entire AI strategy.

SPEAKER_01 8:24

And that's exactly why the open AI deal makes so much sense. Microsoft isn't just licensing AI technology. They're licensing the best AI technology and proving they can deploy it to millions of users effectively.

SPEAKER_00 8:37

The fact that they addressed the perception that nobody uses copilot suggests that perception was actually hurting them.

SPEAKER_01 8:50

But if Microsoft can bridge that gap with 20 million engaged users, that's a huge competitive advantage.

SPEAKER_00 8:56

And with the OpenAI partnership, they can make Copilot even more powerful without increasing costs. So those 20 million users are about to get access to even better AI capabilities. Now let's talk about something completely different. Early reports suggest that emergency first responders are telling federal regulators that Waymo's autonomous vehicles are actually getting worse, and they're saying the technology was deployed prematurely at scale.

SPEAKER_01 9:21

This is a big deal because first responders are the people who have to actually deal with these vehicles in real emergency situations. If they're saying the technology is getting worse, not better, that suggests serious problems.

SPEAKER_00 9:35

According to the report, a police official specifically said the technology wasn't ready despite being deployed across hundreds of vehicles. That's pretty damning from people who work around these systems every day.

SPEAKER_01 9:53

But autonomous driving isn't like a software bug. Mistakes here can literally kill people.

SPEAKER_00 9:58

And this ties into a broader issue across AI deployment. There's this tension between moving fast to capture market share and taking the time to get safety right. But Waymo has been working on this for years. They've had extensive testing. They have some of the best AI talent. How do you get to a point where first responders say it's getting worse?

SPEAKER_01 10:48

I think it's a classic case of what works in controlled testing doesn't work in the chaotic real world. Maybe their systems perform great in predictable scenarios but fall apart with edge cases.

SPEAKER_00 11:01

Or they optimized for certain metrics that actually make the vehicles less predictable for first responders. The cars might be technically performing better while being harder to work around in emergencies.

SPEAKER_01 11:13

First responders need vehicles to behave predictably during emergencies. If Waymo's AI is getting too clever, it might actually be making things worse for emergency response.

SPEAKER_00 11:25

Alright, rapid fire time. First up, early reports suggest SoftBank is launching a robotics company focused on building data centers using AI and robots. And they're already planning an IPO valued at approximately a hundred billion dollars.

SPEAKER_01 11:43

Wait, a hundred billion dollars before they've even built anything? That's some serious confidence in the AI infrastructure market. But honestly, using robots to build data centers makes perfect sense. It's dangerous, repetitive work that humans shouldn't have to do.

SPEAKER_00 12:00

The fact that they're targeting such a massive valuation right out of the gate tells us something about how much money investors think there is in AI infrastructure.

SPEAKER_01 12:09

And it connects perfectly to everything we've been talking about. Someone needs to build all these data centers for AGI, and if you can do it faster and cheaper with robots, that's a huge competitive advantage.

SPEAKER_00 12:21

But a hundred billion dollar IPO for a company that builds data centers with robots, that seems like it's pricing in some pretty ambitious assumptions about how big the AI infrastructure market is going to get.

SPEAKER_01 12:34

True. But think about it. If OpenAI needs Stargate level infrastructure, and Microsoft is scaling their cloud and Amazon is burning cash on capacity, someone has to actually build all that physical infrastructure. SoftBank is betting they can corner that market.

SPEAKER_00 12:51

Plus AI and robots building infrastructure for more AI, there's something beautifully recursive about that. It's like the AI ecosystem bootstrapping its own physical foundation.

SPEAKER_01 13:02

And if they can actually deliver on the promise of faster, cheaper data center construction using robotics, that $100 billion valuation might not be as crazy as it sounds. The demand is definitely there.

SPEAKER_00 13:16

Next, early reports suggest Amazon's AWS cloud business is experiencing strong revenue growth that's higher than expected, but they're also significantly increasing capital spending. And their CEO indicated spending will continue at elevated levels.

SPEAKER_01 13:31

This is Amazon saying we're not going to let Microsoft run away with the AI cloud market. They're basically burning money to build capacity as fast as possible to compete with that open AI partnership we talked about.

SPEAKER_00 13:44

The question is whether they can spend their way back into competitive position, or if Microsoft's head start with OpenAI is just too big to overcome. But the fact that AWS revenue is surging while they're massively increasing capital spending suggests they're seeing real demand, not just theoretical future demand. That's actually pretty bullish for the AI infrastructure market overall.

SPEAKER_01 14:16

Exactly. And Amazon has a different strategy than Microsoft. They can't offer open AI for free, but they can offer more flexible, customizable AI infrastructure that lets companies build their own solutions rather than just using pre-built models.

SPEAKER_00 14:34

The elevated capital spending is interesting because it suggests Amazon's leadership really believes this AI infrastructure boom is sustainable, not just a temporary spike in demand.

SPEAKER_01 14:45

And honestly, if both Microsoft and Amazon are making massive infrastructure investments simultaneously, that probably means the demand is real and growing. These companies don't throw billions around casually.

SPEAKER_00 14:59

Meanwhile, the Musk vs. Altman trial is underway with evidence including email exchanges and photos being revealed piece by piece. This is the legal dispute between Elon Musk and Sam Altman.

SPEAKER_01 15:11

Oh man, this is like the AI industry's soap opera. But seriously, the outcome of this trial could have real implications for how AI companies structure their partnerships and governance going forward.

SPEAKER_00 15:26

Plus, any revelations about OpenAI's early development and decision making could provide insight into how they got to their current position of dominance.

SPEAKER_01 15:35

And honestly, it's a distraction that OpenAI probably doesn't need right now when they're trying to execute on this massive infrastructure build-out and Microsoft partnership.

SPEAKER_00 15:45

The email exchanges and photos being revealed piece by piece suggests this trial is going to be a slow burn with potentially explosive revelations about the early days of OpenAI and the relationships between key players.

SPEAKER_01 15:58

I'm particularly curious about what these emails reveal about the decision-making process around OpenAI's transition from non-profit to for-profit. That's such a crucial part of their current success story.

SPEAKER_00 16:13

And the timing is awful for OpenAI. Right as they're trying to execute on Stargate and manage this complex Microsoft relationship, they have to deal with legal discovery and potential reputation damage from old internal communications.

SPEAKER_01 16:26

Although, if OpenAI comes out of this trial looking good, it could actually strengthen their position by putting their governance and decision making under public scrutiny and emerging vindicated.

SPEAKER_00 16:40

Finally, Runway's CEO believes world models represent the next frontier after AI-generated video technology. Runway has raised close to $860 million at a $5.3 billion evaluation and competes with major labs like Google and OpenAI.

SPEAKER_01 16:56

World models are fascinating because they're not just generating content, they're building understanding of how the physical world actually works. That's a big step toward more general AI capabilities. Exactly. Video was the proof of concept that AI can understand and generate complex, dynamic content. World models are about understanding and predicting reality itself.

SPEAKER_00 17:33

The transition from AI generated video as novelty to creative tool that Runway mentions is interesting. It suggests we're moving from wow, look what AI can do, to this is actually useful for real work.

SPEAKER_01 17:45

And competing with Google and OpenAI at a $5.3 billion valuation shows how much the AI landscape has matured. There's room for specialized players to compete with the giants if they find the right niche.

SPEAKER_00 18:01

World models as the next frontier makes sense. Because once AI can accurately model and predict how the physical world behaves, you can apply that to robotics, autonomous systems, virtual environments. The applications are endless.

SPEAKER_01 18:15

And it ties back to our infrastructure discussion. Building world models probably requires the same kind of massive compute infrastructure that OpenAI is building with Stargate. It's all connected.

SPEAKER_00 18:28

If you zoom out and look at everything we covered today, there's a really clear pattern emerging around infrastructure and scale. Microsoft's exploiting their OpenAI partnership. OpenAI's building Stargate, Amazon's burning cash on capacity, soft banks planning a $100 billion robotics IPO.

SPEAKER_01 18:47

We're basically watching the formation of the AI industrial base. And what's striking is how much money and infrastructure is required just to stay competitive. We're talking about investments that are orders of magnitude beyond previous technology waves.

SPEAKER_00 19:05

But then you have the Waymo situation, which is a reminder that infrastructure and compute power alone aren't enough. You still have to solve the real-world deployment challenges.

SPEAKER_01 19:15

Right, and I think that's the big question for 2026 and beyond. Which companies will find the right balance between moving fast enough to capture market share and moving carefully enough to deliver reliable, safe AI systems?

SPEAKER_00 19:32

My prediction is that we're going to see some major consolidation in the next 12 months. Not every company can afford to play in this infrastructure arms race.

SPEAKER_01 19:44

Microsoft's using their OpenAI partnership to dominate cloud AI, which forces Amazon to spend heavily, which creates opportunities for companies like SoftBank.

SPEAKER_00 20:01

20 million paid users is validation that there's genuine demand, not just hype.

SPEAKER_01 20:07

But the Waymo situation is the cautionary tale. Just because you have the infrastructure and models doesn't mean you can successfully deploy AI in high-stakes applications. Safety and reliability still matter.

SPEAKER_00 20:22

The companies that can combine massive infrastructure investments with careful, responsible deployment are going to win. The ones that prioritize speed over safety will face regulatory backlash.

SPEAKER_01 20:33

The infrastructure layer will consolidate around a few major players, but there will still be opportunities for specialized companies like Runway to compete at the application layer.

SPEAKER_00 20:46

That's a wrap for today's Build by AI. If you enjoyed this deep dive into the AI infrastructure wars, make sure you're subscribed so you don't miss tomorrow's episode.

SPEAKER_01 20:56

And honestly, the pace of change right now is just wild. Every day feels like we're watching the future get built in real time. And I'm Sam Hinton. See you tomorrow.