The $45B DeepSeek Shake-Up I 7th May

A Chinese AI lab just hit a $45 billion valuation with their first investment round after proving they can build models cheaper and faster than OpenAI. Meanwhile, Anthropic is partnering with SpaceX in the weirdest AI deal we've ever seen, and we're getting leaked messages about Elon's 2017 plan to poach Sam Altman for Tesla. Plus, Apple just agreed to pay $250 million for overpromising Siri's AI features. The AI landscape is shifting fast and getting weird.

Transcript

SPEAKER_00 0:00

Genuine question. If you found out a company could build AI models just as good as open AIs, but using a tenth of the computing power and cost, would you immediately assume they were either lying or had found some kind of cheat code?

SPEAKER_01 0:13

Honestly. Yeah, I'd be skeptical as hell. Because that's basically saying they solved the efficiency problem that every major lab has been throwing billions at. It sounds too good to be true.

SPEAKER_00 0:25

Right. Well, that's exactly what DeepSeek just proved they can do. And now they're sitting on a $45 billion valuation from their first investment round.

SPEAKER_01 0:34

Wait, DeepSeek? The Chinese lab? Oh man, this is about to completely upend everything we think we know about the AI race.

SPEAKER_00 0:42

You're listening to Build by AI. I'm Alex Shannon. And today we're diving into a story that might just change how we think about AI development forever.

SPEAKER_01 0:52

And I'm Sam Hinton. We've also got some absolutely wild partnership news between Anthropic and SpaceX, plus leaked messages showing Elon's master plan to steal Sam Altman back in 2017.

SPEAKER_00 1:05

It's May 7th, 2026, and honestly, the AI world keeps getting weirder by the day.

SPEAKER_01 1:10

Let's jump right into this DeepSeek story, because it's kind of blowing my mind.

SPEAKER_00 1:31

Now that's already huge, but here's the kicker. They achieved this valuation because their large language model requires significantly less compute power and cost compared to what OpenAI, Anthropic, and the other major players are using.

SPEAKER_01 1:45

Yeah, and that's not just impressive, it's potentially game-changing. Think about it. If they can really deliver the same performance with a fraction of the resources, that that completely flips the economics of IE development. Suddenly you don't need to be backed by Microsoft or Google to compete.

SPEAKER_00 2:07

Right, but I have to ask, how is this even possible? I mean, we've been told for years that scaling up compute is basically the only way to improve these models. Are they using some fundamentally different approach?

SPEAKER_01 2:18

That's the million-dollar question, literally. I mean, it could be better algorithms, more efficient training methods, or they found some optimization that Western labs missed. But here's what's really interesting. This isn't just about the technology. A $40 billion valuation for a first round means investors are betting this efficiency advantage is real and sustainable.

SPEAKER_00 2:44

Okay, but let me play devil's advocate here. We're talking about a Chinese lab, and there's been ongoing concerns about transparency and verification of AI capabilities. How do we know their claims about efficiency are actually legitimate and not just marketing?

SPEAKER_01 2:59

Yeah, that's a fair point, but investors putting $45 billion on the table suggests they've done serious due diligence. Plus, if DeepSeq was overstating their capabilities, that would come out pretty quickly in the current competitive landscape. The real question is whether this represents a fundamental breakthrough or if it's more about smart engineering optimization.

SPEAKER_00 3:22

You know what's really wild about this? We're talking about a company that launched in early 2025, so they've been around for barely over a year, and they're already commanding a valuation that puts them in the same league as some of the most established tech companies. That's unprecedented, even in the AI boom.

SPEAKER_01 3:39

Absolutely. And it raises some serious questions about what the established players have been doing. If DeepSeat can achieve similar results with a fraction of the compute, were open AI and others just burning money inefficiently, or is there something we're not seeing here?

SPEAKER_00 3:57

That's what I keep coming back to. Like the big labs have some of the smartest people in the world working on this stuff. It seems unlikely they just miss an obvious efficiency gain. Maybe DeepSeek is making different trade-offs that aren't immediately apparent.

SPEAKER_01 4:10

Right.

SPEAKER_00 4:25

What does this mean for the big players like OpenAI and Anthropic? If DeepSeek can really deliver similar performance at a fraction of the cost, that's got to be keeping them up at night.

SPEAKER_01 4:36

Absolutely. It forces everyone to rethink their approach. If compute efficiency becomes the new battleground instead of just raw scaling, we might see a major shift in how AI development happens. Companies will need to focus more on algorithmic innovation rather than just throwing more GPUs at the problem.

SPEAKER_00 4:57

And let's talk about the geopolitical implications here. This isn't just a business story, it's potentially a national competitiveness story. If a Chinese lab can achieve these efficiency gains, what does that mean for the US lead in AI?

SPEAKER_01 5:10

That's huge. The US strategy has largely been based on having better access to cutting-edge chips and more compute resources. If efficiency can level that playing field, it completely changes the strategic landscape. Countries with less access to top-tier hardware might suddenly be able to compete.

SPEAKER_00 5:30

And for businesses looking to integrate AI, this could be huge. Lower costs mean AI becomes accessible to way more companies that couldn't afford the current pricing models.

SPEAKER_01 5:41

Exactly.

SPEAKER_00 5:54

But here's another angle. If this efficiency breakthrough is real, why is DeepSeq taking investment now? If you had technology that was this much better than everyone else's, wouldn't you want to keep it proprietary and dominate the market yourself?

SPEAKER_01 6:09

That's a really good point. Maybe they need the capital to scale up quickly before competitors figure out their approach. Or maybe they're not as confident in their moat as the valuation suggests. Taking investment could be a hedge against that uncertainty.

SPEAKER_00 6:23

Or maybe they know something we don't about what's coming next in AI development. If there's another major shift on the horizon, having $45 billion in the bank could be crucial for surviving the transition.

SPEAKER_01 6:35

Whatever the case, that this is definitely a story to watch closely. The next six months are going to be really telling. If DeepSeek's approach proves replicable, we're looking at a complete democratization of AI capabilities.

SPEAKER_00 6:58

Speaking of unexpected developments, let's talk about what might be the weirdest partnership announcement of the year. Anthropic and SpaceX just announced they're working together, with Anthropic using computing resources from Elon Musk's XAI. I had to read that three times to make sure I understood it correctly.

SPEAKER_01 7:14

Dude, um, same here. This is wild because Anthropic has been positioning itself as the thoughtful, safety-first alternative to more aggressive AI development. And now they're partnering with Elon, who's been pretty vocal about racing ahead in AI development.

SPEAKER_00 7:33

Right. And let's not forget that Elon has had some pretty public disputes with other AI labs, especially OpenAI. So what's anthropic thinking here? Are they just pragmatically going where the compute power is?

SPEAKER_01 7:45

That's probably part of it. Computing resources are still a major bottleneck for AI development. And if XAI has capacity available, it makes business sense. But it also signals something interesting about how the competitive landscape is evolving. Maybe the old rivalries are less important than access to infrastructure.

SPEAKER_00 8:07

But wait, doesn't this create some weird conflicts of interest? I mean, XAI is supposedly competing with anthropic in the AI space. And now they're providing the computing power for their competitor.

SPEAKER_01 8:19

Yeah, it's definitely unconventional. But think about it like this: maybe Elon sees more value in being the infrastructure provider than in direct competition. If XAI can monetize their computing resources while also getting insights into how other labs operate, that might be a smarter long-term play.

SPEAKER_00 8:42

That's actually pretty clever if that's the strategy. You get revenue from your competitors while potentially learning from their approaches. But from Anthropic's perspective, are they comfortable having Elon's company potentially peek under the hood?

SPEAKER_01 8:56

I'm sure there are strict contractual agreements about data privacy and access, but you're right to be skeptical. This partnership suggests that the practical challenges of AI development are starting to outweigh some of the philosophical differences between companies.

SPEAKER_00 9:13

You know what's really interesting about this timing? This comes right after we're hearing about DeepSeek's efficiency breakthrough. Maybe Anthropic is looking at the landscape and realizing that they need to optimize their compute usage. And this partnership with XII is part of that strategy.

SPEAKER_01 9:29

That's a great connection. If efficiency is becoming the new competitive advantage, then partnerships like this make even more sense. Anthropic gets access to potentially better or cheaper compute, and XII gets to demonstrate the value of their infrastructure.

SPEAKER_00 9:47

But let's talk about the cultural fits here. Anthropic has been very focused on AI safety and careful deployment. Elon's approach has been more, let's say, aggressive. How do those philosophies mesh in a partnership like this?

SPEAKER_01 10:01

That's the million-dollar question. Like maybe they're compartmentalizing. Anthropic handles the AI development and safety considerations while XAI just provides the raw computing power. But you have to imagine there are going to be some interesting conversations behind closed doors.

SPEAKER_00 10:19

And what does this say about the broader consolidation in the AI space? Are we moving toward a world where there are infrastructure providers and model developers rather than everyone trying to do everything?

SPEAKER_01 10:30

I think we might be. It's like the early days of cloud computing all over again. Some companies specialized in infrastructure, others focused on applications. Maybe AI AI is heading toward a similar division of labor.

SPEAKER_00 10:46

That would actually make a lot of sense. Building and maintaining massive compute infrastructure is a different skill set than developing AI models. Maybe it's more efficient for companies to specialize rather than trying to do everything in-house.

SPEAKER_01 11:00

Exactly. And it could lower barriers to entry for new AI companies. If you don't need to build your own data centers and buy your own chips, you can focus all your resources on the actual AI development.

SPEAKER_00 11:15

What do you think this means for the broader AI ecosystem? Are we going to see more of these unexpected alliances?

SPEAKER_01 11:22

I think so. As the infrastructure costs keep climbing and the technical challenges get harder, companies might have to choose between ideological purity and practical progress.

SPEAKER_00 11:40

It's definitely something to watch. The AI space is clearly evolving from a simple competition between labs to a more complex ecosystem of partnerships and resource sharing.

SPEAKER_01 11:50

And honestly, that might be healthier for innovation in the long run. When companies can focus on their core strengths rather than trying to do everything, then we might see faster progress overall.

SPEAKER_00 12:01

Speaking of Elon and strategic thinking, we're getting some fascinating behind-the-scenes insight into his 2017 plans for AI dominance. Messages between Shivon Zillis and Tesla executives reveal that Musk tried to recruit Sam Altman to Tesla and establish a rival AI lab there. They even considered Demis Hassabes as a potential leader.

SPEAKER_01 12:22

Okay, this is like reading the alternate history version of AI development. Imagine if Sam Altman had gone to Tesla instead of staying with OpenAI. We might be living in a completely different AI landscape right now. Tesla could have been the dominant AI player instead of open AI.

SPEAKER_00 12:43

Right. And the timing is crucial here. This was 2000 seventeentinium. So before OpenAI really took off with GPT, and before the current AI boom, Elon was clearly seeing something coming that maybe others weren't fully grasping yet.

SPEAKER_01 13:00

Absolutely. And and it shows how strategic Elon was thinking about talent acquisition, going after both Sam Altman and potentially Demise Hesebies. Those are probably the two biggest names in AI leadership. If he had pulled that off, Tesla wouldn't just be a car company with some AI features.

SPEAKER_00 13:21

But here's what I find interesting. Why didn't it happen? Was it Sam who wasn't interested? Or did the Tesla board push back on the idea? Building an AI lab is expensive and risky, especially back in 2017.

SPEAKER_01 13:35

Oh, that's a great question. And um I suspect it was probably a combination of factors. Um Tesla was still figuring out Model 3 production at that point, so maybe taking on AI research felt like too much distraction. Plus, Sam already had momentum with OpenAI, so leaving might not have made sense for him.

SPEAKER_00 13:53

You have to wonder about the conversations that happened. Like what was Elon's pitch to Sam? Was he offering equity, control, unlimited resources? And what made Sam decide to stick with OpenAI instead?

SPEAKER_01 14:05

I'd love to be a fly on the wall for those discussions, but think about it from Sam's perspective in 2017. OpenAI was this ambitious nonprofit trying to democratize AI while Tesla was primarily a car company with big ambitions. The OpenAI mission might have been more appealing to someone focused on AI research.

SPEAKER_00 14:24

And let's talk about Demis Hasabis being considered too. That's interesting because he was already at DeepMind, which Google acquired. So Elon was basically trying to poach leadership from the existing AI giants.

SPEAKER_01 14:36

That shows just how aggressive his strategy was. He wasn't trying to build from scratch. He was trying to essentially transplant proven leadership and expertise directly into Tesla. It's actually a pretty smart approach if you can pull it off.

SPEAKER_00 14:50

What's really fascinating is how this connects to current events. Here we are in 2026, and Elon's XAI is now providing compute resources to Anthropic. It's like he found a different path to influence in the AI world.

SPEAKER_01 15:03

Exactly. He couldn't get the talent, so he's positioning himself as the infrastructure provider. It's actually pretty brilliant. Instead of competing directly on model development, he's making himself indispensable to the companies that are doing that work.

SPEAKER_00 15:19

But I wonder if this was always the backup plan, or if he pivoted to infrastructure after the talent acquisition strategy failed. The messages from Shivon Zillis might give us more insight into his original thinking.

SPEAKER_01 15:32

That's a good point. Knowing Elon, he probably had multiple strategies running in parallel. Plan A was to recruit the top talent. Plan B was to build infrastructure that the top talent would need to use.

SPEAKER_00 15:46

And think about how different the competitive landscape would be if he had succeeded in 2017. We might not have had the OpenAI Microsoft partnership. ChatGPT might have been a Tesla product. And the whole AI race could have looked completely different.

SPEAKER_01 16:00

Yeah, it's wild to think about. Tesla, with an AI lab led by Sam Altman or Demise Hasabis, could have integrated AI into their vehicles way earlier and more deeply. They might have had a huge head start on autonomous driving.

SPEAKER_00 16:19

And it makes you wonder what other strategic moves we might see. If these leaked messages show us anything, it's that the big players are thinking several moves ahead in this AI chess game.

SPEAKER_01 16:29

For sure. And for people following the AI space, the these kinds of strategic partnerships and attempted talent acquisitions are probably better indicators of where things are heading than just looking at product announcements.

SPEAKER_00 16:44

You know what's also interesting? The fact that these messages are coming out now, almost ten years later, it makes you wonder what other behind-the-scenes maneuvering is happening right now that we won't hear about for years.

SPEAKER_01 16:55

Absolutely. The AI space moves so fast publicly, but there are probably all sorts of strategic discussions and attempted partnerships happening behind closed doors that could reshape everything.

SPEAKER_00 17:08

Keep watching the behind-the-scenes moves, because they're often more telling than the public-facing news. Let's shift gears to something that hits closer to home for a lot of people. According to early reports, Apple has agreed to pay $250 million to settle a class action lawsuit over over-promising AI features for Siri. Apparently they promised capabilities that were significantly delayed from their original timeline. So Apple eventually delivered, but apparently way later than they promised. Is that really worth a quarter billion dollar settlement?

SPEAKER_01 18:06

Well, think about it from the consumer perspective. People bought devices based on promised AI capabilities that didn't materialize on schedule. In the tech world, timing matters enormously. A feature that arrives two years late might as well not exist, especially in something as fast moving as AI.

SPEAKER_00 18:27

Right. And this probably speaks to a broader problem in the industry. Everyone's been racing to announce AI features to stay competitive, but the actual development timelines are proving to be much longer than anticipated.

SPEAKER_01 18:39

Exactly. And Apple's usually pretty conservative about their announcements compared to some other companies. If they're getting hit with lawsuits for overpromising, imagine what might be coming for companies that have been even more aggressive with their AI claims.

SPEAKER_00 18:56

That's what's kind of scary about this. Apple has a reputation for underpromising and over-delivering, at least compared to some of their competitors. If even they got caught up in the AI hype cycle, what does that say about the pressure everyone was feeling?

SPEAKER_01 19:10

It shows just how intense the competitive pressure has been. When even Apple feels like they need to make promises they can't keep, that tells you the whole industry was probably moving too fast and making commitments they couldn't back up.

SPEAKER_00 19:26

And let's be real, Siri has been a punching bag for years when it comes to AI capabilities. People have been frustrated with it long before this latest round of promised features. Maybe this lawsuit is as much about accumulated frustration as it is about specific delays.

SPEAKER_01 19:49

Apple's been promising to make Siri better for years, and people have been waiting for those improvements.

SPEAKER_00 19:56

That's a scary thought for the industry. Are we going to see more of these kinds of settlements as AI promises collide with technical reality?

SPEAKER_01 20:04

I think so, especially if this Apple settlement emboldens more class action lawyers to go after tech companies for AI-related promises. Companies are going to have to be much more careful about their timelines and capabilities claims.

SPEAKER_00 20:21

And it's not just about the money, right? This kind of settlement creates a legal precedent. Other companies are probably looking at this and thinking about their own AI promises and whether they're legally exposed.

SPEAKER_01 20:33

Absolutely. Legal teams at tech companies are probably having some very serious conversations right now about what constitutes a promise versus an aspiration when it comes to AI features.

SPEAKER_00 20:49

What does this mean for consumers? Should people be more skeptical when companies announce upcoming AI features?

SPEAKER_01 20:56

Definitely. The lesson here is to treat AI feature announcements like any other tech promise. Wait until it's actually available and working before making purchase decisions based on it. The hype cycle has gotten ahead of the development cycle.

SPEAKER_00 21:13

And maybe this is actually good for consumers in the long run. If companies face real financial consequences for overpromising, maybe we'll get more realistic timelines and better products when they actually ship.

SPEAKER_01 21:25

I think you're right. This could force companies to be more honest about development timelines and more conservative about what they promise. That might slow down the hype cycle, but it could lead to better, more reliable AI features when they do arrive.

SPEAKER_00 21:40

And for the industry, this might actually be healthy in the long run. If companies face real financial consequences for overpromising, maybe we'll get more realistic timelines and expectations around AI development.

SPEAKER_01 22:00

We'll have to see how the next round of AI announcements are worded.

SPEAKER_00 22:05

Alright, let's hit some rapid fire stories. Early reports from the Milken Global Conference suggest that five major figures in the AI supply chain got together to discuss where the industry is struggling. They talked about chip shortages, orbital data centers, and concerns that the fundamental architecture of current AI tech might be flawed.

SPEAKER_01 22:25

Orbital data centers, that's not something I expected to hear today. But the bigger concern is that fundamental architecture comet. If the people building the infrastructure are worried about the foundation, that's a red flag.

SPEAKER_00 22:38

Right. And chip shortages are still a major bottleneck. It sounds like the infrastructure side of AI is struggling to keep up with demand, which makes sense given how fast everything has been scaling.

SPEAKER_01 22:50

This ties back to that DeepSeek story perfectly. If the current approach is hitting fundamental limits, maybe efficiency innovations like theirs are exactly what the industry needs to break through those constraints. Probably a bit of both. When you're running up against physical limits on Earth, space starts to look pretty appealing.

SPEAKER_00 23:24

But the logistics would be insane. How do you maintain and upgrade hardware in orbit? It makes me think the terrestrial solutions aren't as exhausted as some people believe.

SPEAKER_01 23:34

Fair point, though. If these supply chain architects are seriously discussing it, maybe the ground-based options are more limited than we realize. This could be another sign that the industry needs to focus more on efficiency rather than just raw scale.

SPEAKER_00 23:51

Next up, early reports suggest Snap and Perplexity have amicably ended their $400 million partnership that was announced back in November. This would have integrated Perplexity's AI search directly into Snapchat.

SPEAKER_01 24:04

Wow. $400 million deals don't usually end amicably unless something major changed. Either the technology wasn't working as expected, or the business model didn't make sense anymore.

SPEAKER_00 24:16

It makes you wonder if AI search integration is harder than companies initially thought, or if user adoption wasn't there. Snapchat users might not have been looking for search functionality in the first place.

SPEAKER_01 24:27

That's a good point. Snapchat is primarily about messaging and social sharing. Adding search capabilities might have felt forced or unnecessary to users. Sometimes product market fit matters more than technical capabilities.

SPEAKER_00 24:42

And $400 million is a lot of money to walk away from, even for companies of this size. The fact that both sides agreed to end it suggests there were probably some fundamental issues that couldn't be resolved.

SPEAKER_01 25:00

Perplexity can focus on search. Sometimes the best partnerships are the ones that don't happen.

SPEAKER_00 25:06

It's also a reminder that not every AI integration makes sense, even if the technology is impressive. Context and user needs matter as much as capabilities.

SPEAKER_01 25:17

Exactly. This might actually be a sign of market maturity. Companies are getting more selective about AI partnerships rather than just chasing the latest trend.

SPEAKER_00 25:27

Genesis AI, the robotics startup backed by Kostler that raised $105 million, just unveiled their foundational AI model called GNA 26.5. If early reports are accurate, they demonstrated it with robotic hands performing complex tasks.

SPEAKER_01 25:47

A $105 million seed round for robotics, that's massive. And Full Stack suggests they're trying to control everything from the AI models to the hardware. That's ambitious, but could be really powerful if they pull it off.

SPEAKER_00 26:04

Robotic hands doing complex tasks is always impressive in demos, but the real test is consistency and reliability in real-world conditions. We've seen a lot of impressive robotics demos that didn't translate to practical applications.

SPEAKER_01 26:18

True. But with Kosla's backing and that funding level, they're clearly serious about making robotics AI more practical and deployable. Veno doesn't usually bet big on moonshots without seeing some real technical progress.

SPEAKER_00 26:34

The full stack approach is interesting because it means they're not relying on other companies' AI models or hardware. That could give them more control, but also means more complexity and risk.

SPEAKER_01 26:45

It's a classic make versus buy decision. Going full stack means higher upfront costs and complexity, but potentially better integration and performance. It'll be interesting to see if that pays off.

SPEAKER_00 26:59

And robotics is one area where AI efficiency really matters. You need models that can run unlimited compute power in real time. Maybe the efficiency innovations we're seeing elsewhere will finally make practical robotics AI possible.

SPEAKER_01 27:14

That's a great point. If robotics companies can benefit from the kind of efficiency breakthroughs that DeepSeek is claiming, we might see a real acceleration in practical robotics applications.

SPEAKER_00 27:29

Finally, early reports suggest Google has updated their AI search to include quotes from Reddit and other online forums. This helps with niche queries, but could also introduce some chaos and quality concerns.

SPEAKER_01 27:42

Oh boy, Reddit quotes in search results. That's either going to be incredibly helpful or absolutely chaotic. Reddit has great niche knowledge, but also a lot of uh questionable content.

SPEAKER_00 28:00

Right. It's like opening Pandora's box. You get access to really specific community knowledge, but you also get all the internet weirdness that comes with it. Google's going to need some serious filtering.

SPEAKER_01 28:12

Google's going to need some serious filtering and verification systems, but for niche technical questions, Reddit discussions are often more useful than official documentation. The community knowledge is incredibly valuable.

SPEAKER_00 28:25

It's actually a smart move competitively. Other search engines and AI assistants are pulling from similar sources, so Google needs to stay current with where people actually discuss and solve problems online.

SPEAKER_01 28:37

And it acknowledges the reality that a lot of real-world problem solving happens in forums and discussion boards, not just in official documentation or academic papers. Google's trying to capture that conversational knowledge.

SPEAKER_00 28:53

The challenge will be maintaining quality while expanding sources. Reddit can be amazingly insightful or completely wrong, sometimes in the same thread. Context is going to be everything.

SPEAKER_01 29:06

That's where Google's ranking algorithms will really matter. They'll need to figure out how to surface the genuinely helpful Reddit content while filtering out the noise. It's a fascinating technical and editorial challenge.

SPEAKER_00 29:20

If you zoom out and look at everything we covered today, there's a really interesting theme emerging. We've got DeepSeq proving that efficiency might matter more than raw scale, weird partnerships between former rivals, leaked strategic plans from years ago, and companies facing real financial consequences for overpromising AI capabilities.

SPEAKER_01 29:46

The deep seek valuation rewards actual efficiency. Apple's paying for overpromising. And these strategic partnerships suggest companies are getting more pragmatic about what they can actually achieve. Exactly. I think 2026 might be remembered as the year AI development got more realistic and more strategic, less hype, more focus on actual capabilities and sustainable business models. The question is whether this maturation helps or hurts innovation.

SPEAKER_00 30:31

I think it's probably healthy, honestly. The hype cycle was getting unsustainable, and companies were making promises they couldn't keep. If we're moving toward more realistic expectations and better actual products, that's good for everyone.

SPEAKER_01 30:45

And the partnerships we're seeing, like Anthropic and XAI, suggest that companies are realizing they can't do everything in-house. Specialization might lead to better outcomes overall, even if it means less dramatic headlines.

SPEAKER_00 31:00

What's really interesting is how the competitive landscape is shifting. It's not just about having the biggest model anymore, it's about efficiency, practical deployment, and sustainable business models. That opens up opportunities for more players.

SPEAKER_01 31:15

Right. And if if DeepSeek's approach is replicable, it could completely democratize AI capabilities. Suddenly, you don't need billion-dollar compute budgets to build competitive models. That could lead to a lot more innovation from unexpected sources.

SPEAKER_00 31:34

The geopolitical implications are huge too. If efficiency can level the playing field between companies and countries with different access to cutting-edge hardware, that reshapes global AI competition completely.

SPEAKER_01 31:47

And the legal consequences, like Apple's settlement, create real accountability for AI promises. Companies can't just promise the moon anymore without facing potential lawsuits if they don't deliver.

SPEAKER_00 32:01

What should people be watching for as this trend continues?

SPEAKER_01 32:05

Keep an eye on efficiency metrics, not just raw performance. Watch for more unexpected partnerships as companies focus on practical results, and definitely pay attention to legal and regulatory responses to AI promises that Apple settlement might just be the beginning. And infrastructure decisions. One way or another, the AI industry is definitely maturing. Whether that's good or bad for innovation remains to be seen, but it's probably good for actual users who want reliable, practical AI tools rather than just impressive demos.