Latest / In the Loop AI / This Week in AI: Nvidia "Acquires" Groq, Meta Buys Manus, Gemini is the Biggest App

Transcript

This Week in AI: Nvidia "Acquires" Groq, Meta Buys Manus, Gemini is the Biggest App · In the Loop AI

0:00
Ejaaz: Nvidia just spent $20 billion to acquire a company no one's ever heard of,
0:04
Ejaaz: and it's their biggest acquisition to date.
0:06
Ejaaz: Grok specializes in a very unique type of GPU, which will help Nvidia gain an
0:12
Ejaaz: even larger monopoly on the GPU stack.
0:15
Ejaaz: But they're not the only ones making acquisitions over the last week.
0:18
Ejaaz: Meta announced their 20th acquisition of 2025, Manus AI, which specializes in
0:24
Ejaaz: their very own unique AI agent.
0:26
Ejaaz: But I'm feeling pretty bearish about this. I think that this is one of many
0:31
Ejaaz: pitfalls that Meta is going to face over the course of 2025 and 2026.
0:36
Ejaaz: And in other news, Google has finally dethroned OpenAI at the number one spot for AI app being used.
0:44
Ejaaz: It is top of the charts for Apple. We're going to be getting into all of these
0:47
Ejaaz: and much more in this episode.
0:49
Josh: Yeah, so let's get into Grok, which is not to be confused with Grok, the other Grok.
0:53
Josh: One is with a K, which is the one that we frequently talk about, which
0:56
Josh: is made by xai this grok that we're talking about today is
0:59
Josh: g-r-o-q which is a totally different
1:02
Josh: company that was just recently acquired by nvidia but
1:04
Josh: we need to make an important clarification about the acquisition it wasn't actually
1:08
Josh: an acquisition of the company nvidia does not own grok what it was is a non-exclusive
1:13
Josh: licensing deal for grok's inference technology and a big acqui-hire of grok's
1:18
Josh: top leadership and engineers including the founder and ceo jonathan ross who
1:22
Josh: we're going to get into why he is so important in a little while.
1:25
Josh: Basically, it's this non-exclusive license for the technology,
1:28
Josh: the inference technology. And I think that's what we're going to focus on a lot this episode is that
1:32
Josh: NVIDIA has the GPU, which we all know is the powerhouse of AI.
1:37
Josh: But one of the weak points that we're missing is this inference time,
1:41
Josh: this test time, this thing that needs to be quick and very cheap.
1:45
Josh: And what we've seen with Google and other companies is as they've started to
1:48
Josh: introduce TPUs onto the scene, well, things get a little complicated for NVIDIA's bulky GPUs.
1:54
Josh: They're not as efficient. They're not as effective. They're not as cost effective.
1:57
Josh: And it left them with this blind spot, which seemingly they filled with this Grok acquisition.
2:02
Ejaaz: Let's get into what that blind spot actually is, right?
2:05
Ejaaz: So to your point, Josh, NVIDIA makes these GPUs that are really good at helping
2:10
Ejaaz: AI models train, and it requires a different type of architecture to be able to do that.
2:15
Ejaaz: But when it comes to AI applications, what you really want to rely on is how
2:19
Ejaaz: quick the AI responds to you.
2:21
Ejaaz: I don't know if you've come across this, Josh, but when you write a prompt to
2:24
Ejaaz: your AI model, I sometimes get really impatient when it hasn't responded over like 20 seconds.
2:30
Ejaaz: I'm like, come on, like, can you think a little quicker and just give me the answer now?
2:34
Ejaaz: And this comes down to like that test time compute that you mentioned.
2:37
Ejaaz: And the way that you enable a quicker response is if you reduce the latency.
2:41
Ejaaz: And there's many different architectural choices that you can make to do that.
2:45
Ejaaz: The point is, NVIDIA didn't specialize in that, but this company Grok has.
2:49
Ejaaz: And so NVIDIA had a few options to make about this, Josh, which is highlighted in this tweet here.
2:55
Ejaaz: Option A, they just kind of continue using NVIDIA GPUs. It costs a lot more
3:00
Ejaaz: and it's just inefficient.
3:02
Ejaaz: Or option B, they acquire or partner with Grok, which is the leader in these
3:07
Ejaaz: specialized chips that allow for lower latency.
3:10
Ejaaz: Or option C, they start creating their own specialized
3:14
Ejaaz: chips from scratch but it'll cost them a lot of time a lot
3:17
Ejaaz: of money at which point grok will already be the number
3:19
Ejaaz: one leader so i think this is a strategic position they spent 20 billion dollars
3:24
Ejaaz: which is a lot of money for them but um they're able to acquire the number one
3:29
Ejaaz: leader in that case and own that monopoly for them and not be afraid of grok
3:34
Ejaaz: kind of taking over nvidia or providing an alternative to nvidia in the future yeah.
3:39
Josh: And if you remember like google recently we made so many episodes about the
3:42
Josh: TPU and how it's becoming a legitimate threat to NVIDIA. Why?
3:45
Josh: Well, it's because the TPU is good, really good at something that NVIDIA is
3:49
Josh: not, which is this very deterministic computation that happens locally and rapidly
3:53
Josh: when it's actually delivering AI to the user.
3:56
Josh: So this won't interfere with the core business of NVIDIA, which is largely the
4:00
Josh: GPUs and this pre-training part of the world.
4:02
Josh: And we have a post that breaks this down really well, Ejas, that you shared
4:05
Josh: with me earlier that I kind of want to take a second to go through if you don't on sharing
4:08
Ejaaz: Yeah so it's this post right here um and it
4:10
Ejaaz: describes something called lpu so
4:13
Ejaaz: we've heard of gpu graphics processing unit lpu stands for language processing
4:17
Ejaaz: unit and as it as its name implies it's specialized for llm specifically and
4:23
Ejaaz: efficiently processing tokens into whatever output you would like from your
4:28
Ejaaz: ai model right and there's a few things that make this specifically unique,
4:33
Ejaaz: architecturally. One, it uses this thing called static random access memory.
4:39
Ejaaz: So we've all heard of RAM.
4:41
Ejaaz: We've all heard of DRAM, dynamic random access memory.
4:45
Ejaaz: SRAM is an alternative to DRAM. So for context here, DRAM is like the most sought
4:51
Ejaaz: after type of memory that people want to put into their AI chips to train their models.
4:55
Ejaaz: And Josh, you and I have been discussing this a lot.
4:57
Ejaaz: The price of DRAM has like 5X'd over the last couple of months.
5:03
Josh: It's kind of outrageous.
5:04
Ejaaz: Because we just don't have enough. In fact, like the cost of a, what was this?
5:08
Ejaaz: The cost of like a PC memory card now, Josh, is the equivalent of like the latest MacBook.
5:14
Ejaaz: It is just insane out there, right?
5:17
Ejaaz: And so Grok has this alternative memory board called SRAM, which they integrate
5:21
Ejaaz: into this LPU chip, which is what makes it so unique.
5:24
Ejaaz: Well, what's so unique about it? In a typical GPU, so take, for example,
5:28
Ejaaz: NVIDIA's GPU, in order to process an LLM request, you need to take information
5:34
Ejaaz: from the memory and take it to the processing unit, the processing part of the chip.
5:40
Ejaaz: With SRAM, you don't have to do that as much. You can just store the entire
5:43
Ejaaz: model weight on that part of the memory component, and it's able to process
5:48
Ejaaz: way more efficiently. How much more efficiently?
5:51
Ejaaz: 10 times more. So for a singular chip, you can store 10 times more data, or more memory rather,
5:59
Ejaaz: and it'll require one-tenth of the energy cost, which makes it an extremely
6:03
Ejaaz: efficient chip to get fast inference between your AI model and the person that's using the AI model.
6:09
Ejaaz: That's the unique unlock here, and it's just super cool to see.
6:12
Josh: Yeah. So in conclusion, it's not that NVIDIA bought Grok.
6:16
Josh: It's closer to NVIDIA bought the parts of Grok that actually matter the most.
6:19
Josh: And a big part of that equation is the talent, the people that were involved
6:23
Josh: in Grok. And where did I find out about the people?
6:25
Josh: Well, on Chamath's memo from 2016, nine years ago. So for people who don't know
6:32
Josh: Chamath Paliapati, I mean, you probably do.
6:34
Josh: He's a very, just very influential person in the world of technology.
6:38
Josh: And he published his memo from 2016 very early on about this guy named Jonathan
6:43
Josh: and about his decision to be the lead investor in the Series A of this small
6:48
Josh: little company called Grok.
6:50
Josh: And nine years later, fast forward, this tiny little company worth $10 million sold for
6:56
Josh: got a license acquisition for $20 billion. And it was fascinating to read this
7:00
Josh: memo and kind of see what the world looked like in 2016. It's this funny snapshot.
7:05
Josh: It's a very forward-looking and accurate representation of where we would have
7:09
Josh: ended up at a time when this wasn't clear.
7:11
Josh: So if you remember in 2016, the transformer paper hadn't even come out yet.
7:16
Josh: There was no such thing as a transformer.
7:18
Ejaaz: Josh, I was at university in 2016. I had no idea that was happening either.
7:23
Josh: No one did. And yet this memo signaled like, hey, this thing called a TPU is
7:28
Josh: going to be very important towards machine learning, which is also going to
7:31
Josh: be this very important thing.
7:32
Josh: And what's really funny about the memo is the header. It includes a subsection,
7:37
Josh: fundamental human need, which says understanding and creation.
7:40
Josh: Then it says, is this going to affect 25% of the population?
7:43
Josh: Yes, but indirectly. $100 billion by 2045?
7:46
Josh: Possibly. A special person? Yes. And then as you read more through the memo,
7:50
Josh: you start to understand that he invested in this company because the founder
7:53
Josh: and CEO was, or the founder and CTO at the time, who later became the CEO,
7:58
Josh: he was the founder and the creator of the TPU, which Google now has.
8:03
Josh: This is the dude, the dude running rock is the dude who created the TPU,
8:06
Josh: which is now the biggest thing in the world.
8:07
Josh: So clearly they saw something early.
8:10
Josh: Clearly it was the correct bet. And I mean, good for them. They deserve it.
8:14
Josh: I mean, we certainly were not thinking about these things nine years ago when really mattered
8:17
Ejaaz: Isn't it crazy that he had the foresight to identify both the trend of machine
8:24
Ejaaz: learning becoming the most popular form of computation,
8:28
Ejaaz: and coupling that with like, we need a new chip.
8:32
Ejaaz: So at the time, GPUs were super popular. You and I were discussing this before
8:35
Ejaaz: we started recording, Josh.
8:37
Ejaaz: You know, we were using GPUs for gaming. We were then using it for like crypto
8:40
Ejaaz: mining around that time. It was becoming more popular.
8:42
Ejaaz: But he had the foresight to be like, this is probably going to be used for some
8:46
Ejaaz: kind of AI ML inference in the future.
8:49
Ejaaz: And he saw it a decade ago and he waited it out. So if you look here,
8:52
Ejaaz: it says it would allow them to own around 28.57% of the company.
8:58
Ejaaz: Which is just a massive bet.
9:00
Ejaaz: I mean, yeah, great. His 10 million is now worth a hell of a lot more.
9:04
Ejaaz: I think there's rumors that he made over a billion dollars on this.
9:07
Ejaaz: It's just insane to kind of see at that time.
9:10
Ejaaz: I just want to like kind of put into context what this NVIDIA acquisition looks like, right?
9:15
Ejaaz: So there are specific types of chips or kind of like hardware that you need
9:20
Ejaaz: to make a successful AI model or an AI product in general.
9:24
Ejaaz: You need GPUs, which everyone's heard of. Now we We have these LPUs,
9:27
Ejaaz: which now, I'm here on this tweet, NVIDIA now also owns Viagroc,
9:33
Ejaaz: the number one leader of this.
9:34
Ejaaz: And then there's CPUs. And we
9:36
Ejaaz: know that companies that are specializing in CPUs include Intel and AMD.
9:40
Ejaaz: Earlier on in this year, NVIDIA made very big investments in both of the leading
9:46
Ejaaz: companies, AMD and Intel.
9:47
Ejaaz: So I just see this kind of like this Thanos grip that tightens with every single
9:53
Ejaaz: acquisition that Jensen makes. And the craziest part about this,
9:57
Ejaaz: Josh, is that it's all pretty much free.
9:59
Ejaaz: When the news broke that NVIDIA had acquired Grok, their market cap pumped $35
10:05
Ejaaz: billion, which paid for the acquisition of more. That's so nuts.
10:09
Ejaaz: It's just crazy. It's just crazy.
10:11
Josh: So that's the big news on NVIDIA this week in an acquisition that feels very
10:16
Josh: powerful and meaningful, like it will actually move the needle.
10:18
Josh: And now here's another acquisition where I'm not sure the same thing can be said.
10:22
Josh: Just this week, Meta announced another acquisition after the seemingly hundred
10:27
Josh: that they've done in terms of personnel and companies this year in a company named Manus.
10:33
Josh: Now, Manus is a Singapore-based, Chinese-founded company, and the price hasn't
10:37
Josh: been disclosed yet, although it's rumored to be somewhere between $2 and $5 billion.
10:41
Josh: Manus is less than a year old, EJS, I believe, and it was valued at half a billion
10:48
Josh: dollars just earlier this year, and they raised $75 million. dollars.
10:52
Josh: And then in nine months, they went from zero to a hundred million dollars.
10:55
Josh: Oh yeah, here's the post. In eight months, they went from zero to a hundred
10:58
Josh: million annual recurring income, which is just this unbelievable hockey stick growth.
11:04
Josh: So why would Meta want to acquire that?
11:06
Josh: Manus is different than other companies. Manus is more of an agent than a chatbot.
11:11
Josh: And if you'll remember, Meta's had a tough time building a chatbot,
11:15
Josh: but has a lot more use for an actual agent. So Ijaz, maybe you want to explain
11:19
Josh: kind of what this acquisition was,
11:21
Josh: why Meta made it, and what their hopes are in terms of what type of value this
11:26
Josh: can add to a company that's kind of having a pretty tough time deploying AI
11:30
Josh: in an effective and meaningful way.
11:32
Ejaaz: So I'm reminded of something specific that Zuckerberg said in his interview
11:37
Ejaaz: with Dwarkesh earlier this year.
11:39
Ejaaz: Dwarkesh was like, why are you investing tens of billions of dollars into AI?
11:43
Ejaaz: You guys are a consumer app product. And he goes, yeah, I don't want to build the best AI model.
11:48
Ejaaz: I want to build the best consumer AI app.
11:52
Ejaaz: Meta specializes in consumer apps. We think we're the best at it.
11:55
Ejaaz: And we think we can be pioneers in this new AI world.
11:58
Ejaaz: We're not trying to get the best model. We just want to build the best consumer experience.
12:01
Ejaaz: I think this is another step towards that. So if we kind of like track back
12:06
Ejaaz: to like the age of the internet, which we currently kind of live in right now, right?
12:10
Ejaaz: We use websites. We download apps. It's a very kind of archaic thing when you
12:16
Ejaaz: consider that AI is probably going to automate a lot of this.
12:18
Ejaaz: How is that going to be automated? It's probably going to be in the form of
12:21
Ejaaz: agents, agents that can automate a bunch of work for you, but also make your
12:26
Ejaaz: entertainment and lived experience way, way cooler, right?
12:30
Ejaaz: So it can help you with research, but it can also help you kind of discover
12:33
Ejaaz: new things, new shopping experiences, stuff like that.
12:36
Ejaaz: And Manus AI is one of the top agent producers out there.
12:40
Ejaaz: And like you said, Josh, this hockey stick growth to 100 million AIR,
12:43
Ejaaz: even though I think that's a pretty vague metric and I think a lot of that can
12:48
Ejaaz: be gamed it's still very very impressive and I know a bunch of my friends that
12:51
Ejaaz: use Manus in many different ways can I just put my bear hat on for a second
12:56
Ejaaz: Josh I know we don't like the bears on the show but my skin is tingling for this one particularly,
13:01
Ejaaz: why can't like a Claude code or a Claude in general just kind of.
13:06
Ejaaz: Recreate this entire product. I don't understand the unique capacity for here.
13:11
Ejaaz: And maybe it's because Meta didn't have the props or the talent to create this
13:16
Ejaaz: themselves. So they just saw this as an easy acquisition for a cheap $2 to $5 billion.
13:21
Ejaaz: And now they can kind of just put this ready-made product out to their millions of billions of users.
13:27
Ejaaz: That's also still bearish for me because they've spent $35 billion this year
13:31
Ejaaz: alone to acquire like a hundred different people and they've got nothing to
13:35
Ejaaz: show for it. I'm just honestly confused, dude.
13:38
Josh: Yeah, it's aside from the actual strategic thing, you could kind of look at
13:42
Josh: what Manus does and then I guess back yourself into the conclusion from there.
13:46
Josh: And you mentioned Cloud Code. I would say Manus is kind of similar to Cloud
13:49
Josh: Code in the sense that it's kind of a harness for a base model.
13:53
Josh: There are these base models that are interchangeable that Manus accepts because
13:56
Josh: Manus does not actually have a base model.
13:59
Josh: But it creates this harness around the model that allows the AI to do more interesting
14:04
Josh: things than it would if you were to just go to ChatGPT or Gemini.
14:08
Josh: A few of those things include tool use and orchestration. So a big thing when
14:12
Josh: you're using AI agents is tool use.
14:15
Josh: What types of tools are they able to use in order to do things for you?
14:19
Josh: And if you think of CloudCode as the agent for building code,
14:23
Josh: you could think of Manus as the agent for controlling your computer,
14:27
Josh: for doing deep research, for managing files.
14:29
Josh: It's more of the consumer-facing version of CloudCode, where it's really great
14:34
Josh: at tool use. It's really great at running virtual computers.
14:36
Josh: It's really good at analyzing real-world competitiveness through deep research
14:40
Josh: of many, many different variables. It's really good at UX and design.
14:44
Josh: And it has this unique skill set that doesn't require a core base model in order to use.
14:50
Josh: And I think that's important because Meta does not have this core base model yet.
14:53
Josh: In fact, I suspect if they're going to continue using Manus,
14:55
Josh: they're going to have to use some third-party application until they could deploy
14:59
Josh: their new supposed model named Avocado, which is their first closed-source model
15:04
Josh: that will hopefully release sometime soon.
15:07
Ejaaz: The way you describe that, like, you know, it doesn't use a base model.
15:10
Ejaaz: It kind of sounds like a rapper, right? And, you know, there are a bunch of
15:14
Ejaaz: critics against rappers. I actually think rappers are great and fine.
15:17
Ejaaz: But isn't that just cursor? I guess it's the same as what cursor is for coding.
15:23
Ejaaz: Manus is for agents in a way, and it doesn't matter what model it uses.
15:27
Ejaaz: I don't know if that convinces me enough behind this acquisition.
15:32
Ejaaz: And I'm purely coming at that from the fact that they have acquired so much
15:36
Ejaaz: supposedly good talent and teams over the last six months, and they've had nothing to show for it.
15:42
Ejaaz: And I'm kind of confused at meta strategy in general. They say that they want
15:46
Ejaaz: to go into consumer AI, but the Vibes app, which is basically AI TikTok that
15:51
Ejaaz: they released, got completely swept under the rug. No one cares about it.
15:54
Ejaaz: No one uses their AI chatbot. In fact, their kind of gaming metrics by allowing
15:59
Ejaaz: any Instagram search is now an AI model search.
16:03
Ejaaz: I don't know if you've seen that, Josh, but like it comes up with this really
16:05
Ejaaz: annoying kind of summary when you're on Instagram.
16:07
Ejaaz: I'm just like, why are you doing this? I don't really care about this.
16:09
Ejaaz: So I'm not convinced, but let's see in Q1 whether they release this new avocado
16:15
Ejaaz: model and whether it's actually something that people want to use.
16:18
Ejaaz: I'm going to remain kind of like uncertain on this for now, I think.
16:21
Josh: Yeah, I guess in terms of the agentic use case, it does make sense because like
16:25
Josh: Meta has, they have billions of people generating real world data all the time
16:28
Josh: to improve this. Then they have the glasses and the Quest headsets that they want agents for.
16:33
Josh: They have, like you mentioned, Instagram, WhatsApp, Messenger as places to use.
16:38
Ejaaz: They have the distribution.
16:39
Josh: They have the distribution. So if you can apply this agent, this agentic force
16:43
Josh: on top of it, you can generate a lot of value for the user.
16:46
Josh: So maybe, I mean, again, we'll see. And then one last point is there's this
16:50
Josh: funny little chart that I like to reference, which is Yahoo.
16:52
Josh: It shows their acquisitions by year up until 2014 because, well,
16:56
Josh: if you remember after that, they didn't do so well and then had to get acquired.
16:59
Josh: But there's an interesting chart that follows this. And this is kind of reflective
17:02
Josh: of Meta, which shows their market cap over time. And you could see that,
17:07
Josh: Their market cap exploded and was at the highest it's ever, or just about the
17:11
Josh: highest it's ever been as they were making all these acquisitions.
17:13
Josh: They were desperately trying to acquire the talent needed in order to be competitive in this new world.
17:18
Josh: Unfortunately, that didn't work.
17:19
Josh: And just two years later, they were acquired by Verizon for 90% less.
17:24
Ejaaz: They acquired Tumblr? That's crazy.
17:26
Josh: They acquired everybody. They made so many acquisitions two years and one year
17:30
Josh: prior to them going under.
17:31
Josh: And what we're seeing with Meta, not to draw a direct tie, but there's this
17:36
Josh: kind of almost irrational acquisition spree where they're spending tens of billions
17:41
Josh: of dollars on individual people.
17:43
Josh: They're spending billions of dollars on companies and they're just trying to
17:46
Josh: get talent in-house, but they still have nothing to show for it.
17:49
Josh: They're not able to generate revenue. They're not able to make the products better.
17:51
Josh: So I don't want to directly compare it to, because I don't see them going down
17:55
Josh: 90% like Yahoo did, but it's just a testament to show that normally when people
17:59
Josh: start, or companies start acquiring at this rate.
18:02
Josh: It's just something to be aware of is all. But we should probably move on to
18:06
Josh: the next bit of news for this week because there's a lot going on and that is
18:09
Josh: the Google Gemini app back in the news, right?
18:12
Ejaaz: Josh, my favorite company is
18:14
Ejaaz: officially number one. And it's a massive release being able to say that.
18:18
Ejaaz: So what are we talking about? So Google Gemini is obviously Google's leading AI model,
18:25
Ejaaz: but it's also the name of their AI app, which is a dedicated app like the ChatGPT
18:29
Ejaaz: app that you can go on and you can speak to Gemini 3 Pro and all the latest models.
18:34
Ejaaz: And for a while now, it's been creeping up the rankings.
18:39
Ejaaz: And it was in the top 10 for a while. and then Gemini 3 Pro released and they
18:43
Ejaaz: shot to number two and it's been at number two for a while. They are finally number one, Josh.
18:48
Ejaaz: And this comes after learning a few weeks ago that more people.
18:54
Ejaaz: Or rather people are spending more time using Gemini models than the average
18:59
Ejaaz: ChatGPT user, which is just crazy.
19:02
Ejaaz: And the number of daily, weekly, and monthly active users have been creeping
19:05
Ejaaz: up to the point where it's almost, not quite yet, but almost at parity with
19:10
Ejaaz: OpenAI's ChatGPT users as well.
19:13
Ejaaz: This is a major milestone, in my opinion, because for a while,
19:16
Ejaaz: I mean, you and I know this very well, Google's just been on a rollercoaster of a ride through AI.
19:21
Ejaaz: They created the Transformer paper, so they were supposedly in the lead,
19:25
Ejaaz: but then they didn't do anything with that.
19:27
Ejaaz: ChatGPT and OpenAI ate their lunch, and then they made this massive 180-degree
19:31
Ejaaz: turn. Sergey Brin came back, and they've made this crazy comeback. And it's just
19:35
Ejaaz: Kind of cool to see them take the crown of this position, both at the consumer
19:39
Ejaaz: level and at the chip layer with their TPUs.
19:43
Ejaaz: And now everyone's using Google Search in AI mode. It's just crazy to see this
19:46
Ejaaz: type of comeback from Google.
19:49
Josh: And they forced the $20 billion acquisition for NVIDIA and Grok.
19:52
Josh: I mean, I assume if GPUs didn't become big, that would not have been a pressing issue.
19:56
Josh: But this is not the only time that Google was in the news this week.
19:59
Josh: There's already more news coming out of Google, which is as it relates to energy
20:02
Josh: and infrastructure, EJS, what's going on with this? I didn't get a chance to read it.
20:05
Ejaaz: So the headline here is Alphabet's acquired this company called Intersect.
20:10
Ejaaz: And I think they acquired it for $5 billion.
20:13
Ejaaz: So what does this company do? Intersect is an energy supplier and producer.
20:19
Ejaaz: And so you might be like, well, why is this important? Well,
20:22
Ejaaz: Google's building out all these crazy AI data centers to help house their TPUs
20:28
Ejaaz: and train and inference their Gemini models.
20:31
Ejaaz: But there's one big issue, Josh. there's not enough power to power these GPUs
20:36
Ejaaz: and TPUs up in the first place. There's never enough power. There's not enough power.
20:39
Ejaaz: America's electric grid is constrained at the moment.
20:43
Ejaaz: And it is a big, big worry going into 2026 because everyone's spending...
20:48
Ejaaz: Tens of billions of dollars, hundreds of billions of dollars to create these GPUs.
20:52
Ejaaz: But if we don't have any power to supply them in the first place,
20:55
Ejaaz: then these become kind of defunct pieces of metal that just collect dust.
21:00
Ejaaz: And Satya Nadella earlier this year actually spoke about that,
21:02
Ejaaz: saying he has a warehouse that has hundreds of millions of dollars worth of
21:05
Ejaaz: NVIDIA GPUs, and he can't do anything with that because he can't get the electrical supply to it.
21:10
Ejaaz: So I see this as a very strategic move. What sounds boring on the surface is
21:14
Ejaaz: actually a really important move.
21:16
Ejaaz: And And Google obviously sees meaning here, and it's kind of making its next
21:20
Ejaaz: steps towards solving that crisis.
21:21
Josh: Yeah, part of being an AI behemoth now requires owning the entire stack from
21:26
Josh: electron generation through token output.
21:29
Josh: And it's becoming increasingly clear that if you want to be a serious competitor,
21:32
Josh: you must compete across every facet of the spectrum, including actual energy
21:37
Josh: generation, which is what we're seeing from Google.
21:39
Josh: And it's a signal that what we predicted is kind of coming true,
21:43
Josh: where you can't really compete unless you are one of these giant behemoths.
21:46
Josh: And the race is now getting shrunken down, it's tighter and tighter and tighter
21:50
Josh: by the people who could actually afford to compete at this level.
21:53
Josh: Google has won, OpenAI has won, Microsoft's XAI, but there's not that many left.
21:58
Josh: So it seems to me that if you really want to be a big player, you need the money.
22:03
Josh: Google's doing it and they are making big strides in this.
22:06
Ejaaz: And the final topic of this episode, Josh,
22:10
Ejaaz: Daddy has officially spoken. Andre Carpathy has given us his 2025 LLM year in review.
22:17
Ejaaz: We also did a year in review. That episode is out. You should definitely go check it out.
22:21
Ejaaz: But I was really keen to see what Andre would talk about because his history
22:25
Ejaaz: has been across all the major AI model providers.
22:29
Ejaaz: He's worked at Google before. He was actually one of the founding engineers there.
22:33
Ejaaz: When it comes to the AI world, he was involved in the research side of things
22:36
Ejaaz: and OpenAI as well, right?
22:39
Ejaaz: And so I was curious, like, you know, where does his allegiance lie?
22:42
Ejaaz: And some of his takeaways were super fun. One of my favorite takeaways,
22:45
Ejaaz: Josh, I think you would agree with this. I'm scrolling right to the bottom here. One second.
22:49
Ejaaz: Is Nano Banana. He loved Google's text to image and image editing generation
22:58
Ejaaz: model because he sees it as the future of graphic user interface.
23:03
Ejaaz: So if you have to navigate to a website or look at a screen to interact with
23:07
Ejaaz: the internet today, the future of AI will be generative.
23:10
Ejaaz: And he believes you don't have to go to a website. The website will just be
23:13
Ejaaz: custom made and created for you.
23:15
Ejaaz: And he thinks NanoBanana is the first step to it. Do you agree with that,
23:18
Ejaaz: Josh? I know you're super excited with this.
23:20
Josh: Oh, dude, I love NanoBanana. It's one of my favorite models.
23:22
Josh: I think it won my favorite model of the year in 2025.
23:25
Josh: It is a miracle that that thing works. And I love Andre's framing of how he
23:29
Josh: thinks it is most impactful when applied to building user interfaces.
23:34
Josh: And that seems really cool because as we go to this dynamic world where everything
23:38
Josh: is generated in real time, being able to generate really powerful interfaces
23:42
Josh: or just visual elements in general is so important.
23:45
Josh: And Google, again, is leading the charge. So it's cool to see Andre...
23:48
Josh: On board with that and agreeing in a way that is very thoughtful.
23:53
Josh: This is a really good post. I would encourage everyone to go and read the full
23:56
Josh: thing if you get a chance, because I really enjoyed it.
23:58
Ejaaz: And that is it, folks. A crazy week for a week where supposedly not much is
24:02
Ejaaz: meant to be happening. You know, we've got a quiet Christmas week.
24:06
Ejaaz: New Year's is around the corner. And we have a casual $20 billion acquisition
24:10
Ejaaz: or licensing acquisition, if you want to be technical about it before the year's end.
24:14
Ejaaz: I'm excited for 2026. We have an episode releasing soon, which is going to be
24:18
Ejaaz: our predictions for 2026.
24:19
Ejaaz: Josh and I have thought long and hard about that to the extent where Josh brought out his glasses.
24:24
Ejaaz: So please stay tuned for that episode. It's going to be a very good one.
24:27
Ejaaz: And we have some really, really good takes.
24:29
Ejaaz: But aside from that, if you're not subscribed, if you haven't turned on notifications,
24:33
Ejaaz: if you haven't subscribed to our newsletter where we drop alpha twice a week,
24:38
Ejaaz: please do all of those things and we will see you in the next episode.
24:41
Josh: Yeah, it's the guys, it's the holiday season. If you do anything for us for
24:44
Josh: Christmas, just hit the subscribe button. That'd be pretty cool.
24:47
Josh: We got a great episode that came out earlier this week. We have the predictions coming out on Friday.