Latest / eMotors: Electric Revolution / Tesla's Proposed xAI Investment: Pros and Cons
Transcript
- 0:09Welcome to the deep dive. Today we're looking at something
- 0:12our sources are calling pretty explicitly a global war.
- 0:16But this isn't about tanks or jets.
- 0:19No, not at all. This is fought in massive data
- 0:21centers with, well, billions of billions of dollars worth of
- 0:25GPU's. We're talking about the race for
- 0:27artificial general intelligence, AGI.
- 0:29And it's really shifted, hasn't it?
- 0:30It's not science fiction anymore.
- 0:31It's the big, high stakes, multibillion dollar battlefield
- 0:36for tech in 2025. Absolutely.
- 0:38AGI, you know, achieving intelligence that can reason,
- 0:41learn, adapt, maybe even better than humans.
- 0:44It's become this huge corporate goal and it's pushing these
- 0:46giants into making some really big, maybe even risky, strategic
- 0:50moves. And that specific kind of
- 0:51strategic move is exactly what we're digging into today.
- 0:54We've gone through a lot of documents detailing 1 a
- 0:56particularly controversial plan, Elon Musk proposing Tesla invest
- 1:00$5 billion into his own private company, XAI, right?
- 1:04And that's all hanging on a really critical shareholder vote
- 1:06coming up in November 2025. South, our mission in this deep
- 1:11dive is to give you real clarity on this.
- 1:13We want to unpack the connections, the synergy they're
- 1:16aiming for, the frankly huge financial risks and the legal
- 1:21mess that's already brewing around this idea.
- 1:23Yeah, We need to get past the headlines and figure out, does
- 1:26this move actually change what Tesla is?
- 1:29Is it shifting from just an EV leader to, I don't know, a
- 1:32fundamental AI platform? Or is it maybe an overreach,
- 1:36something that could just drain money and focus away from their
- 1:39core business, especially in what's probably the most
- 1:42expensive tech race ever? Let's start with the Y.
- 1:45What's the core argument for this deal?
- 1:47Why merge Tesla's resources with XAI?
- 1:49Well, the big picture, the vision Musk is selling is
- 1:52creating this kind of accelerated closed loop AI
- 1:55system, some thing nobody else can copy easily.
- 1:57Exactly. He wants to leverage Tesla's
- 1:59existing hardware. Think the Dojo supercomputer and
- 2:04maybe more importantly, it's enormous pool of real world
- 2:07data. And that data is key, isn't it?
- 2:09We're not just talking about stuff scraped off the Internet
- 2:11here. No, absolutely not.
- 2:13That's the unique angle they're pushing to justify this massive
- 2:17investment. Tesla has what, over half a
- 2:19million cars out there? Yeah, something like that,
- 2:22acting as sensors basically. Right, Dynamic real time sensors
- 2:25all over the world. World caturing tricky
- 2:27situations, edge cases, weird road conditions, how cars
- 2:31actually interact. Stuff that you know a standard
- 2:35language model trained on web text just wouldn't encounter.
- 2:39So the idea is to feed all that complex real world data into
- 2:43XAI. To accelerate grok their
- 2:45language model to give it a better grasp of the physical
- 2:47world, how things actually work out there.
- 2:49And it flows back the other way too, right?
- 2:51Advancements at XAI are meant to improve Tesla cars.
- 2:54Precisely, that's the closed loop.
- 2:56Any breakthroughs XAI makes in AI potentially AGI could
- 2:59theoretically be plugged straight back into Tesla's self
- 3:02driving systems. Maybe improve the in car
- 3:04experience customer service via. X formally Twitter, right?
- 3:08And don't forget Tesla's own computing power, especially
- 3:11Dojo. That gives XAI a massive head
- 3:14start in terms of the raw processing power needed without
- 3:16having to rely so much on outside cloud providers.
- 3:19OK, that sounds like a neat package on paper, but this is
- 3:23where things get kind of wild. The money involved.
- 3:25The numbers our sources show are.
- 3:28Dizzying. Yeah, the risk profile is
- 3:29extraordinary. Yeah, XAI is apparently
- 3:32projected to burn through something like $13 billion just
- 3:35in 2025. $13 billion in one year.
- 3:39Mostly on compute, yeah, heavy investment in servers, GPUs,
- 3:43that kind of thing. And Despite that burn rate,
- 3:46they're reportedly looking for a $200 billion valuation in their
- 3:50next funding round. Whoa.
- 3:51OK, 200 billion. How does that stack up?
- 3:54Well, that's the gamble, isn't it?
- 3:56To give you some context, our sources mentioned market leader
- 3:58Open AI is valued much higher, potentially $500 billion.
- 4:02But look at their user base. Right, Open AI has huge reach.
- 4:04Estimated somewhere between 700 and 800 million weekly users.
- 4:08OK. And XAI.
- 4:09The estimate for XAI is closer to 30 million weekly users. 30
- 4:12million versus 7 or 800 million. And they want $200 billion.
- 4:17Exactly. So that valuation isn't really
- 4:19based on current market share or traction.
- 4:22It's almost entirely pinned on the hope, the speculation that
- 4:25Grok, fueled by that unique Tesla data, will eventually
- 4:29achieve some form of AGI and leapfrog everyone else.
- 4:33Wait a second, $200 billion based on 30 million users?
- 4:37Even with special data and a $13 billion burn rate, that sounds?
- 4:42I mean, isn't that valuation relying almost completely on
- 4:45hype and Musk's reputation rather than anything concrete
- 4:47yet? That's the absolute crux of it.
- 4:49The justification hinges entirely on this belief that
- 4:52true AGI, which would be an exponential game changer, needs
- 4:56that kind of integrated physical world data loop that only Tesla
- 5:00can provide. So if they pull it off, maybe
- 5:02the valuation makes sense in hindsight.
- 5:03If they don't. If they fail, or even just
- 5:05stumble, that $5 billion investment from Tesla gets
- 5:08chewed up pretty fast by the ongoing costs.
- 5:10And it's precisely the speculative nature that's
- 5:13feeding the legal fire. Yes, the ugly side, as you put
- 5:15it, the legal drama there are shareholders actively suing
- 5:18Musk, right? Which must complicate this vote.
- 5:20Very much so. They're claiming a breach of
- 5:23fiduciary duty. The core argument is that Musk
- 5:26basically took a massive business opportunity building a
- 5:29major AI company and diverted it to his private venture XAI
- 5:33instead of keeping it within Tesla.
- 5:35And that's particularly sensitive because of his history
- 5:38with open AI. Exactly.
- 5:39He left Open AI citing a potential conflict of interest
- 5:42back then. Now, the lawsuit argues he's
- 5:45created XAI, which directly competes with Tesla for scarce
- 5:49AI talent and potentially drains resources and focus from the
- 5:53core EV business he's supposed to be leading.
- 5:55So the shareholders have the final say because the board is
- 5:58sort of stuck. Pretty much.
- 6:00The board hasn't taken an official stance on the $5
- 6:02billion proposal. Most people read that as quiet
- 6:05approval, but because of the lawsuit, Musk and the board
- 6:08members have to abstain from voting on this specific
- 6:10investment. So for you, the listener may be
- 6:12an investor, the question is really stark.
- 6:14It boils down to this. Do you believe XA is highly
- 6:17speculative path justifies Tesla putting in $5 billion and
- 6:22potentially subsidizing that huge burn rate, especially while
- 6:25Tesla is also fighting tooth and nail to stay dominant in the
- 6:28increasingly crowded EV market? That really sets the stage
- 6:31perfectly for the broader AGI arms race.
- 6:34To really grasp what Tesla trying to buy into with this X
- 6:37AI bet, we need to look at the wider battlefield.
- 6:40What does AGI even mean to these companies right now?
- 6:43In this context, AGI isn't just slightly smarter AI, it's about
- 6:47achieving a general level of intelligence, the ability to
- 6:49form a huge range of tasks, reason abstractly, learn
- 6:53flexibly, basically operate out or even beyond human levels.
- 6:56And the key difference from today's AI like ChatGPT?
- 7:00It's the ability to continuously learn and integrate new stuff
- 7:03after it's been deployed without needing massive retraining every
- 7:06time. That's the the Holy Grail
- 7:08they're all chasing. OK, so in that race, Open AI,
- 7:10the Pioneer is still kind of setting the pace even if they
- 7:13haven't hit full AGI yet. What was so significant about
- 7:16their GVT 5 launch back in August 2020?
- 7:19Well, the sources we looked at call GPT 5A really major step
- 7:23forward. They describe this kind of
- 7:25sophisticated 3 part structure. You've got the standard model
- 7:28for everyday tasks, then an advanced reasoning part they
- 7:32actually call GPT 5 thinking, and then there's a real time
- 7:35router that figures out which part of the brain, so to speak,
- 7:38is best for the specific thing you're asking it.
- 7:40It's more modular, more dynamic. And how good is that thinking
- 7:43part? How far did they get with
- 7:45reasoning? They're hitting levels our
- 7:47documents describe as PhD level reasoning.
- 7:50I mean, look at the math tests on the AE in 2025 benchmarks.
- 7:54These aren't simple sums, they're complex multi step
- 7:57problems. GPT 5 apparently scored an
- 8:00amazing 94.6% accuracy. Wow.
- 8:03And crucially, they also claimed to have cut down on those
- 8:06factual errors, the hallucinations, by about 45%
- 8:09compared to the last version. That's a big deal for
- 8:12reliability. So Open AI is pushing the
- 8:14boundaries on pure digital smarts, but Google and DeepMind
- 8:18seem to be taking a different tack, focusing more on
- 8:21connecting AI to the physical world.
- 8:23That definitely seems to be the core of their Gemini strategy.
- 8:26Yes, they're working on text and images, but they're really
- 8:29emphasizing logical thinking with something called deep
- 8:32think. Deep think.
- 8:33Yeah, it's basically a tool that lets the AI simulate complex
- 8:37multi step reasoning internally before it gives you an answer.
- 8:41Try and make it think things through more carefully.
- 8:44But the really potentially disruptive part is Gemini
- 8:47Robotics. Gemini Robotics.
- 8:49That sounds less like a chatbot and more like, well, robots.
- 8:52It is. They call it a Vision Language
- 8:54Action model. So the AI doesn't just
- 8:57understand text or images, it can actually do things, execute
- 9:01physical actions. In the real world.
- 9:03That's the goal they're aiming at squarely at industrial
- 9:05robots, advanced manufacturing, automating physical tasks,
- 9:08trying to move AI from being just a digital helper to a real
- 9:12physical collaborator. And that ambition, covering both
- 9:15digital and physical, is maybe why there are these rumors about
- 9:18Apple. It seems plausible the sources
- 9:20suggest Apple is seriously negotiating to integrate Gemini
- 9:24into the next version of Siri that could launch maybe as early
- 9:27as August 2025. Which would be.
- 9:28Huge. Massive.
- 9:30It would instantly put Gemini's capabilities on the hundreds of
- 9:33millions of Apple devices. That puts enormous pressure on
- 9:36rivals like Microsoft, Amazon, on everyone fighting in that
- 9:40integrated AI ecosystem battle. OK, so we have the pioneer, the
- 9:44high stakes gambler with TESLAX AI and the ecosystem player like
- 9:48Google. But there are others pushing
- 9:50hard too, right? Like Meta, they seem determined
- 9:53to catch up. Oh, absolutely.
- 9:55Meta is throwing serious money at this.
- 9:57The figure mentioned is around $15 billion just on data
- 10:00infrastructure and training hardware.
- 10:02But what's really fascinating is the focus on talent.
- 10:04The people involved. Yeah, Zuckerberg himself is
- 10:07apparently leading this elite team called the Fantastic Fifty,
- 10:10and the report suggests these individuals are getting massive
- 10:12salaries, like 9 figures. It sounds less like R&D and more
- 10:16like, yeah, well, an AI talent arms race trying to hoard the
- 10:19best minds. Wow.
- 10:21OK, In Amazon, they're taking yet another different path,
- 10:24leveraging their strength. Completely different.
- 10:26Amazon seems laser focused on AGI agents through their
- 10:29platform called Nova Act. AGI agents meaning.
- 10:32These are AI systems designed to handle really complex sequences
- 10:36of tasks, mostly in simulated environments for now.
- 10:40Think super intricate supply chain management, optimizing
- 10:43logistics, automating huge chunks of e-commerce operations.
- 10:47So less about general conversation, more about
- 10:49specific complex jobs. Exactly.
- 10:52Amazon's vision seems to be creating these advanced virtual
- 10:55employees that can manage large scale operations with minimal
- 10:59human input, leveraging their expertise in logistics and
- 11:02scale. It's incredible how diverse the
- 11:05approaches are, but this whole competition, this AGI race, it's
- 11:08not just happening in Silicon Valley, is it?
- 11:10It's truly global. Absolutely.
- 11:12You can't ignore the global dimension.
- 11:14The documents make it clear that while U.S. companies get a lot
- 11:16of the headlines, China is rapidly becoming a major
- 11:19independent force in AI. How are they approaching it?
- 11:23They're definitely leveraging their huge scale of data, but
- 11:25significantly, they're also pushing hard to rely more on
- 11:29their own domestic semiconductor chips, esecially Huawei's Ascend
- 11:32series. Ah, so that's a strategic move
- 11:34likely related to Western chip restrictions.
- 11:37Recisely, it's about building technological self-reliance,
- 11:41insulating their AI development from potential sanctions or
- 11:45supply chain issues. And which Chinese companies are
- 11:48really making waves globally using these domestic chips?
- 11:51We're seeing 3 standouts in the reports.
- 11:53First, there's DeepSeek. They seem to have global
- 11:55ambitions and are really focused on making their AI cost
- 11:57effective. OK.
- 11:58Then there's Z dot AI. They used to be Z Poo AI.
- 12:01They've released a model called GLM 4.5 that's specifically
- 12:06tuned to run really well on those homegrown Ascend ships.
- 12:09Interesting. And the third.
- 12:11Moon shot AI. They seem quite technically
- 12:13sophisticated. Their big model is called Kimi
- 12:15K2. Kimi K2, what's special about
- 12:18that? You mentioned something about an
- 12:19OE architecture. Right, Moe stands for Mixture of
- 12:22Experts. Yeah, Kimi, K2 is apparently a
- 12:24massive 1000 billion parameter model, but it uses this Moe
- 12:28approach. What does that mean?
- 12:29Mixture of experts. Think of it like having a huge
- 12:32team of specialized mini AI's inside the main model.
- 12:36When you ask it something, it doesn't need to activate the
- 12:38entire enormous 1000 billion parameters, it just activates
- 12:42the relevant expert or combination of experts for that
- 12:44specific task. So it's more efficient.
- 12:47Much more efficient. It gives them incredible speed
- 12:49and performance while potentially using less
- 12:52computational power overall. This allows them to compete
- 12:55really aggressively on cost and capability, even against the
- 12:59giant US models. So we have this relentless
- 13:02global race, billions pouring in, incredible tech
- 13:05breakthroughs. But it leads to a fundamental
- 13:07question, doesn't it? Is AGI actually here or close or
- 13:11some of this just marketing? Yeah, there's definitely a
- 13:14strong vein of skepticism running through the source
- 13:16material. A lot of analysts, international
- 13:18media too, are asking if AGI is maybe being used more as a
- 13:23marketing buzzword, right? Now to attract investments.
- 13:25Exactly. You know, corporate AI spending
- 13:27is projected to be enormous, maybe $400 billion in 2025 in
- 13:32the US alone. Calling your next AI model a
- 13:35step towards AGI is a very powerful way to capture a slice
- 13:40of that investment pie, even if the underlying tech is more
- 13:43incremental. But the ethical concerns are
- 13:45real. Regardless of the marketing
- 13:47hype. We're seeing warnings from top
- 13:49executives like Microsoft CEO talking about the possibility of
- 13:52conscious AI. Yeah.
- 13:54And experts flagging risks like, you know, mass manipulation or
- 13:57losing control. For sure, and the documents are
- 13:59pretty clear on this point. Well, true, continuously
- 14:02learning AGI hasn't quite been achieved yet.
- 14:05The pace of development is faster than ever.
- 14:07It's that acceleration that makes the ethical and
- 14:09philosophical questions so urgent, moving them beyond just
- 14:13technical debates, which. Brings us to this really
- 14:15interesting, almost philosophical point from our
- 14:17sources. Something that tries to
- 14:19distinguish raw super intelligence from what makes
- 14:21human intellect unique. Right, the material talks about
- 14:24this difference between the Loinoi, which translates roughly
- 14:27to the human core, and the loin rawing AI, the AI hollow core.
- 14:31That's a powerful metaphor, the human core versus the hollow
- 14:34core. What's the argument there?
- 14:36The argument presented is that AGI is biggest weakness.
- 14:39Perhaps it's fundamental right now is its inability to
- 14:43genuinely achieve what the source calls empathy and
- 14:46compassion over super intelligence when dealing with
- 14:50really significant human issues. So AI can be incredibly smart,
- 14:55logical, powerful. But it lacks that core human
- 14:58capacity for genuine empathy, for compassion.
- 15:02That human core, the source argues, is our unique strength.
- 15:05It's what prevents pure intelligence from becoming just
- 15:08cold, hard calculation. It's seen as the really
- 15:11essential part of human development and decision making.
- 15:14It really underscores just how complex this whole landscape is.
- 15:17I mean, this deep dive has touched on everything from
- 15:20modular AI structures and massive financial bets like
- 15:23Tesla's. To the intense global race
- 15:25involving the US and China, the legal fights, the marketing
- 15:28versus reality debate, and ultimately, these deep questions
- 15:31about what intelligence even means.
- 15:33The AGI race is definitely not just one thing.
- 15:35It's technology, finance, risk, global strategy, even philosophy
- 15:38all tangled together. And hopefully you now have a
- 15:42much clearer picture of the key subjects at the heart of this
- 15:46global competition. It's a competition that really
- 15:48does touch the future of, well, human development itself.
- 15:52We really hope this deep dive provided that clarity for you
- 15:55today. Engaging with these complex high
- 15:58stakes topics is crucial, so if you found this useful, please do
- 16:01follow our channel. That way, you won't miss out on
- 16:04new content as this incredible global AI story continues to
- 16:08unfold. Thank you so much for joining
- 16:09us. We'll see you next time.