Inside The Trillion Dollar Ai Buildout Dylan Patel Interview

TITLE: Inside the Trillion-Dollar AI Buildout | Dylan Patel Interview CHANNEL: Invest Like The Best DATE: 2025-09-30 ---TRANSCRIPT--- If the models don’t improve, we’re absolutely screwed. And in fact, the US economy will go into a recession.

It’s about the highest stakes like capitalism game of all time.

Godsend in terms of like how much efficiency and value can be created and it doesn’t ever have to get to like digital god level. Now, I do believe we’re going to get to digital god level eventually. Eventually, if I could have an intelligence as smart as a Google senior engineer, that’s $2 trillion of software value.

Is that the main like bottleneck to be attacked?

We’re popping the bubble right now cuz the limit of AI is infinite.

I was going to lay out this idea of going through the past, present, and future of compute as like the big big idea for our conversation, but since it just happened, I don’t think you’ve heard you talk about it anywhere. I’d love to start by asking about this whole OpenAI Nvidia thing, which sounds exciting, seems vague, not really sure what’s going on. And maybe you could explain it to us as you see it and what the strategic implications are of the big announcement.

All right. So, I think it’s very very simple, right? You’ve got OpenAI paying Oracle lots of money. You’ve got Oracle paying Nvidia lots of money. You’ve got Nvidia paying open lots of money meme. We’ve got the infinite money glitch here. No, no, no. That’s not actually what’s happening, right? What’s really happening is OpenAI has an insatiable demand for compute. The compute precedes the buildup of business. You have to have the cluster before you can rent it out for inference, right? Or rather run models on it for inference. You have to have the cluster to train the model that’s good enough that it unlocks new use cases which then can be adopted and there’s an adoption curve there for any new use case. So you have to have all these things like sequenced given this is a game of the richest people in the world or rather the biggest tech giants in the world, right? It’s Zuck. It’s Google, you know, Larry and Sergey or Sergey is like constantly in the business now again, right? It’s all the biggest people in the world. It’s Elon, right?

There’s very much a risk of OpenAI being too small to matter, right? You know, which is crazy to say because they’ve got 800 million users, but like where’s the revenue? Where’s the compute? They could easily get swamped in terms of how much compute they have. If they don’t move fast enough and if they don’t have the most compute or like among the most compute they will get beaten. The magic of OpenAI was that they just spent way more compute on a single model run on GP3 and four and they had the foresight and the vision and the execution. But they made that bet and they were able to secure it. And at the time it was like meh, right? It was a few hundred million, whatever. But now it’s sort of like, well, Mark Zuckerberg sees how much compute he’s going to have to get even though he has this insane cash flow that he’s like, “Oh, wait. I need to go sign a deal with Apollo for $30 billion on this data center in Louisiana, this mega data center I’m going to build.” It’s like, “Wait, why didn’t you just fund this with cash flows? You have so much cash flow.” It’s like, “Because my plans, that’s just the physical data center. Now, what am I going to put in it?” Is like so much money. The amount of capital that people are going to have and are dumping into this is insane, right?

Google was slow to wake up and then they were slow to pivot their data center operations. They were slow to do everything and so while they could have way more compute than anyone by a humongous degree they haven’t been able to deploy as fast. So OpenAI has still been on the curve. And then they have like how much they allocate to search and generative search is not really necessarily competing with OpenAI, it’s the mega models. So if you have this tremendous vision of what’s going to happen with AI, you know that it takes a ton of compute to build them, you know pretty much the amount of compute you could dedicate to these models is limitless and they will get better. Now it’s a log log scale, right? I.e. you need 10x more compute to get to the next tier of performance. You might think of it as diminishing returns, but what if the next tier of performance is like a six-year-old versus a 16-year-old. Like child labor is quite effective versus a six-year-old you can’t get to do much, right?

This is the conundrum that OpenAI is in. They have to get more compute than anyone or at least among there. They have to race with the giants. These giants are trillion dollar businesses. So, how does OpenAI get there? It’s partnering with Microsoft. Well, that soured some. It’s partnering with Oracle. Well, Oracle can do a lot, but Oracle doesn’t even have a balance sheet like Google and Microsoft and Amazon. So it’s meta, Elon, sport of kings. This is very much like the Pascalian wager nature of all of this with the tech giants. Oracle can be part of it, but OpenAI needs allies. They need people to effectively spend the capex ahead of the curve and trust that they’ll be able to pay the rental income because that’s what it is at the end of the day. OpenAI is committing to five-year deals. These five-year deals cost X amount of money. It’s 10 to 15 billion dollars per gigawatt of data center capacity that you pay a year. And then that 10 to 15 billion dollars for a gigawatt of data center capacity. You’re paying that for five years. That’s 50 to 75 billion of cash that goes out the door for one gigawatt of capacity. And you talk about what Sam’s saying is like, hey, I need 10 gigawatts. I need more than 10 gigawatts.

Then you end up with this really challenging aspect of like how do you pay for that? And hey, that’s only the rental price. It becomes who is the balance sheet for this. That’s the reason these deals are coming about. Oracle is making a massive bet. He’s getting good margin off of it, but he’s making a massive bet that this capex that he’s going to pay for OpenAI will actually be paid because he signed a $300 billion deal with OpenAI. Where’s that going to come from? Your revenue is like 15 billion ARR this month maybe right on a run rate basis it’ll get to 20 by the end of the year. How do you pay $300 billion?

If the bet works out they’ve just made 100 billion of profit. Pure cash profit. But if it doesn’t work out they’ve got this huge… and they’re starting to raise debt. They’re going to start raising more and more debt. So this game… now Nvidia’s kind of got the same conundrum. Google and Amazon are doing these deals whether it’s to TPUs or for Tranium, whether it’s Anthropic or others, they’re trying to court OpenAI, they’re trying to court other companies. How do I get into this game? Okay, fine, I can rely on Microsoft somewhat, I can rely on Oracle somewhat, but at the end of the day, GPUs, if I want GPUs to be king, part of it is just like my chip is the best but part of it is also who’s going to pay the capex upfront. Google and Amazon will pay the capex up front if it’s for TPUs. They won’t pay the capex up front necessarily for that same capacity of GPUs. So you’ve got this challenging aspect and so that’s where this Nvidia and OpenAI deal comes from.

I want to dig into the underlying assumptions driving this on the training and inference side. I don’t think it’s a diminishing return. I think that’s important to recognize. Given it’s a log log chart, scaling laws are right. Given there’s no model architecture improvements you just throw more compute, data, model size at it, it gets better at this pace. Everything has shown that it will continue.

GP5 is not necessarily that much bigger than 4 and 4 is smaller than four. What’s changing is sort of paradigm of how you spend the compute and also like if they made a bigger model, could they even serve it? No. They did 4.5 and it was terrible. No one could serve it. It was actually quite a bit smarter but they couldn’t actually serve it at any reasonable cost and speed. This is why Anthropic has the same issue. All of their revenue comes from Sonnet, doesn’t come from 4.1 Opus, which is the better model. It’s bigger, but it’s slow because the hardware is not caught up in terms of inference speed. And so no one wants to use a slow model. The user experience sucks.

But as far as like if the model gets better at each scale of hardware spend, I would say all the tech giants believe it. I believe it. I think a lot of people in the financial community are like this is freaking scary. Because the moment it stops, wherever you were on the rung, if we went from $50 billion spend to $500 billion spend, well, that $500 billion spend is never going to have ROI. It was one thing if 50 billion didn’t have ROI, but now this 500 doesn’t have ROI. It’s a big problem.

One could think of it as diminishing returns because when you go from $50 billion of spend to $500 billion of spend, you only move up one tier of model capabilities in absence of major algorithmic improvements. But that iterative performance improvement in the model is like a six-year-old versus a 13-year-old. The amount of work you can get a 13-year-old to do is actually quite valuable relative to a six-year-old. And the same applies to like a college intern versus someone who graduated and has even one year of work experience.

Where do you think we are today? Depends on the domain. For software developers I think we’re really pretty good. That’s where we’re seeing the most value creation happen. Where you see Anthropic have gone from like a billion or less of revenue to seven to eight. It’s the fastest revenue ramp we’ve ever seen for anything. And it’s basically all code related. Some of it’s their own Cloud Code product, some of it’s Cursor, some of it’s GitHub Copilot which also offers Anthropic models. It’s Windsurf. It’s all these different avenues to access the same thing.

Eventually, if I could have an intelligence as smart as a Google senior engineer, that’s $2 trillion of software value. Because that’s how much the world pays to software engineers today. You could augment them and build five times as much or 10 times as much because these things don’t just run on their own. It’s more of like a force multiplier to the existing person.

So sort of this draws back to the OpenAI Nvidia deal. Most people in the market don’t quite get it. They’re like, “Oh, this is just like round tripping.” It is to some extent. If OpenAI builds 10 gigawatts of capacity, Nvidia will do hundred billion dollars of equity investment into OpenAI in the form of cash. The first chunk of the deal in the press release is one gigawatt, $10 billion. So pretty straight line but one gigawatt to build is like $50 billion. So Nvidia is paying 10 billion, OpenAI still has to come up with other 40 somehow. They can go to the markets, get a loan or get someone else to put a loan. There’s infrastructure funds trying to get into this. There’s all these commercial real estate people trying to get in this.

The nice thing for Nvidia is they sell… of that 50 billion they capture maybe 35 billion of that is capex that goes directly to Nvidia. Nvidia’s gross margin is 75%. So effectively like half their gross profit from that deal is going directly to OpenAI in the form of an equity investment. The 25% that’s COGS is staying. And then they keep the other half of the gross profit on their balance sheet. So Nvidia is not necessarily like round-tripping. But effectively what’s happening is OpenAI gets the opportunity to pay for a big chunk of it in equity and Nvidia’s lowering their prices without lowering their prices effectively. And they’re getting ownership of a company who very likely could… but Nvidia comes out great because they’re getting the capex dollars up front. All they’re really doing is they’re saying half of my money that’s in this, sure it does make its way to me somehow, but in reality I still made half of that gross profit and the other half is equity in a company that may or may not be worth something. A company that may or may not be able to pay hundreds of billions of dollars of compute deals that they’ve signed.

Token demands doubling every two months. The thing I like to call is tokconomics. It’s the economics of the tokens. How much compute is being spent, how much is the gross profit, what’s the value being created by these tokens. Nvidia keeps saying AI factory which produces intelligence. That intelligence has value. Let’s say you have a gigawatt of capacity. What can I serve? I could serve a thousand times of a model that’s really shitty. Or I could serve one times amount of a model that’s good. And I could serve like 0.1 times of a model that’s amazing. Multiply that by whatever factor of how many users, what’s the number of tokens outputted.

This is sort of where the whole GP5 thing comes around. OpenAI had a challenging thing. They had a couple gigawatts of capacity. How do they maximize their serving capacity? One avenue is we continue to serve big models and we make bigger models and the tokens are more expensive. But this log scale is really challenging because yes the value is an order magnitude more but the cost is way more and then the real whammy is the user experience is way worse. If I serve a massive model it’s slow. And users are fickle.

For GP5, what could they have done? They could have gone big. They actually tried that with 4.5. They screwed up some things because it was really hard to get 100,000 GPUs to work properly. Also, they hadn’t figured out the whole reinforcement learning paradigm at that time. The scaling laws are a chart of quality versus compute, but that compute breaks down into how much bigger do I make the model, how much more data do I put in the model, and if the internet only has so many tokens, you’re kind of screwed. So it took… there was potentially a cliff until reinforcement learning happened where you can generate data and train the model to be better without the internet having that data.

So they kind of had this problem of you have X amount of compute, you can service your users, but today if people want to use my API I rate limit them because I can’t actually serve them all. I have multiple ChatGPT accounts because I use deep research. You kick off a bunch, you read it, and you’re like, “Wow, I learned a ton. Move on.”

So as OpenAI, what’s your choice? Do you make the model way bigger and not be able to serve anyone? Or do you make the model the same size, which is what they did for GP5. It’s basically the same size as 4o and roughly the same cost. That’s actually a little bit cheaper potentially, and then you just serve way more users. And then instead of putting them on a bigger model, you put them on models that do thinking. So if you’ve used GP5 thinking or GP5 Pro, there’s more intelligence there.

I’m not doubling my hardware every two months, but I’m doubling my tokens every two months. So there has to be enough of a cost decrease and there is, with at a given level of intelligence.

If you could snap your fingers and change a dial that would most unlock and unleash more development, is it just inference latency? I think capacity/cost is more important than latency. I think existing levels of latency are fast enough for a lot. Now if the latency was 10x lower for GP5 then they could have made a model that was 10x bigger and served it at this speed. But then you would have the same capacity issue.

There’s this whole concept of overparameterization. If you just throw more parameters in a neural network… when humans had a vocab test, you memorized before you understood and it wasn’t until you did multiple repetitions and in different forms that you actually understood the content rather than just memorized. When you do an LLM it’s the same thing. If you throw some data at it, it will memorize it before it generalizes. This concept called grokking. The aha moment. The models do the same thing. They memorize up until they understand at some point. If you make the model bigger and bigger without the data changing, you just memorize everything. And actually it starts to get worse again because it never had the opportunity to generalize.

The challenge today is not necessarily make the model bigger. The challenge is how do I generate and create data that is in useful domains so that the model gets better at them. Nowhere on the internet to show you how to fly through a spreadsheet using only your keyboard. There’s no data on the internet about this. So how do you teach a model that? It’s not going to learn it from reading the internet over and over.

That’s where this whole reinforcement learning paradigm happened. Giving it environments, specific environments to learn and then fold back in. There’s sort of a challenge in terms of building those environments. There’s like 40 startups now in the bay doing these environments. These startups are making environments for OpenAI, Anthropic and others. As simple as here is a fake Amazon. Figure out how to click around and purchase items. Figure out how to compare the two items and pick the right one. Eventually it’s bought the right deodorant and you’ve succeeded and you fold it back in.

Or it could be clean this data. Here’s this table. Ton of dirty data. How do I separate out the columns? You slowly iteratively teach it. Another example is you’re in a game whether it’s tic-tac-toe or Call of Duty or a math puzzle. These things hill climbed up math puzzles like crazy from Q4 of last year to Q2 of this year. A lot of that was here’s how I use Python to write something that does the math for me. And now these things are actually quite good at math.

The environments can be super varied. It can be here’s a medical case, what’s wrong with it? And then you have another model say, here’s your instructions on how you would grade the result. You can feed these models into very complicated environments.

Mid innings on pre-training? I think we’re early on text. We’re quite early. And then the other angle is just because you’ve used the text doesn’t mean you can’t learn faster. Take a class, give them all a book, tell them to read it once, test them all. One kid’s going to get 100 and one kid’s going to get a 40. If you read the book out loud to them, the kid who got a 100 might get a 30 and the kid who got a 40 might have got a 60. When we talk about model architecture, the same thing happens.

In fact pre-training is the base of everything. You want to keep having gains because any gains on pre-training, the model learns a little faster or the model’s a little bit smaller for the same quality, feeds into the next stage which is this whole post-training side which will subsume the majority of the compute at some point.

And inning wise for reinforcement learning? I think we’ve like thrown the first ball. Think about how much we see throughout our life and how much of that information we throw away. I don’t remember anything about what I had for lunch yesterday. But if it was amazing or bad, I would have remembered that. These models, these environments, we’re generating tons of data and throwing most of away and training the model, but it’s infantesimal compared to what humans have done.

There’s people who even think you don’t get to the magical AGI until you embody it, i.e. you put the model in something that can interact in the real world, as in a robot. Elon and XAI think embodiment is required to get to artificial general intelligence because you need the model to be able to pick this up or oh wow this is a rotating thingy which you could never get from just watching a video about it.

I think we’re so early in the reinforcement learning because that’s what humans are. We’re reinforcement learners.

As for reasoning, RL, it’s a lot about how the human psyche and intelligence works. There’s a caution of trying to make it too much like humans because the fundamental substrate is not like humans. The processing is not like humans. Our brain is very different from how these ALUs on a chip work. The scaling of these things is very different. But at the same time, it’s important to reckon back to what actually makes people smart.

The magic of transformers was attention. I calculate everything in my context length. King and queen are actually exactly the same on a ton of stuff, but then it’s the opposite on one number because one’s a male, one’s a female. What we’re terrible at is exact recall. But you get the gist. Models very different. Fundamentally transformer attention has been calculating the attention to everything to each other and getting the models to actually be able to recall. You can get the model to repeat exactly what you want, anything in its context length. But what they really suck at is having infinite context because it’s sparse. You have sparse, you’ve taken this entire world and you’ve encoded it in such a small amount of data that lives in your brain and it’s so sparse but you understood how to grab the fundamental reason and put it down there. Whereas models they haven’t been able to create something sparse yet.

How do you reason over the context of infinity? Models, there’s a ton of research going on in long context. How do I get longer and longer context without blowing up my model cost. The fundamental algorithm needs to change and improve over time. That doesn’t necessarily mean the model has to work like we do. Why can’t the model just reason and have a database that it writes stuff in? We refer to our notes, we refer to our calendar, our shopping list.

One of the first things OpenAI did was deep research. Deep research is working for 45 minutes. It’s outputting millions of tokens and creating this amazing thing. They enabled it to be able to write something down elsewhere and have this recall and effectively use language to compress information, put that off to the side, use language to compress other information off to the side, and then looking at all this compressed information and writing something.

The upper limit I think I’m among the most bullish you can get because the upper limit of this is that this will just be smarter than humans. I don’t think that will happen anytime soon. Even if that doesn’t happen anytime soon, there’s so much valuable stuff that can be done with these models that economically we will skyrocket. If the models know how to do COBOL to C and Python migration of mainframes, just migrate everything. The world is how much more efficient? Making all these random applications and automated reports and stop using Excel as a database. We could literally just pause it at 6 months from now time frame of how good it is at software development and it would be godsend in terms of how much efficiency and value can be created for the economy and it doesn’t ever have to get to digital god level.

Now I do believe we’re going to get to digital god eventually. Is that 10 years? Is that 5 years? Is that 100 years? Is that a thousand years? I don’t know. Because there’s so many unknown unknowns.

What physical intelligence is doing, what are they actually doing today is like holy crap it’s so simple in terms of to a human. To models it’s like picking this up is freaking hard. How much do I squeeze my pinky versus this finger? But you pick up a glass of water and you tilt it. Think about how simple that is. You don’t even think about it. But instinctually pick up a wine glass and swish it and it lets the aroma out and you smell it. These models can’t do that yet. Nowhere close. But it doesn’t need to be that good. It doesn’t need to be able to swish a wine glass and not break it. What it needs to be tremendously valuable is pick this up and put it down here after knowing what it is.

More than 10% of Etsy’s traffic is straight from GPT. Amazon blocks GPT but otherwise it would be really high. People make purchasing decisions through GPTs they just don’t make the purchase. OpenAI’s head of applications was at Shopify and created the shopping agent. The models are going to purchase for you. They’re going to do actions for you and the model and then the company that does those actions for you will be able to take some sort of take rate. Even if it’s 1% or 2%. It’ll be like a credit card transaction. Visa is the most amazing business in the world because of this.

On talent wars: I think it’s tremendously hilarious that people are like, “Oh my god, this person’s getting paid a billion dollars.” It is infeasible. How could this person possibly be worth that much? Well, they’re running the experiments on chips that cost hundred billion. If every wasted experiment they do, if they just used a third of the compute… there’s so much wasted compute. These things are so complicated. There’s a group of people just trying different stuff on the existing data. How do you mix it? What order do you feed it into the model? How do you filter it? What’s the architecture? If you just make them a little bit more efficient, you just saved not only 5% of your compute time, you also save 5% across my entire inference fleet. And we’re so far away from these models being anywhere near as efficient as a human brain. Adding more people to the problem doesn’t make it faster.

One thing Jensen told me: “Dylan, the reason America is rich, people have it all wrong. The reason we’re rich is because we’ve exported all the labor, but we’ve kept all the value.” And that’s what Nvidia does. They’ve exported the labor of making their chips. It’s done in Asia. Those companies make money, not as much money as Nvidia and Apple. All the gross profits are going to them.

ML research is the exact same as semiconductor manufacturing. Tune a thousand different knobs. You frankly cannot test everything. It’s too large of a search space. You just have to have enough intuition, pick that point, see the data. The whole idea of ML research is you spend a lot of time on compute training doing what effectively were useless things besides teaching yourself what’s the right thing to do and what’s the wrong thing to do.

On power dynamics: Does Anthropic have all the power in the Cursor relationship? Cursor has like nearly a billion dollars of revenue now. But their margins are what they are and they’re sending most of it back to Anthropic. But then Anthropic is taking all the gross profit dollars and putting them into compute for training. So then all those gross profit dollars are going to the hardware layer.

But Anthropic only makes the model that’s generated the code. There’s a lot more in this system. Cursor gets all of the data, all of the users, how they interact. Anthropic gets a prompt and sends a response. Cursor is training embedding models, made the autocomplete model, can switch the Anthropic model to OpenAI model whenever they want.

The Microsoft OpenAI one is absurdly interesting. 2023 it was like Microsoft’s going to own the world. H2 2024 Microsoft backed down. Amy Hood and whoever else were like, “Maybe we don’t need to be on the hook for $300 billion.” They paused a bunch of data centers. “Oh, you know, we don’t need to be the exclusive compute provider.” They relinquished this power.

On Nvidia: The king. All of the gross profit is going to them today. They can’t buy anything. They weren’t even allowed to buy ARM. What do they do with all this cash flow? They’re doing demand guarantees, backstopping clusters. Using their balance sheet to win.

On the bubble question: If the models don’t improve, yes, we will overbuild. If the models don’t improve, we’re absolutely screwed. In fact, the US economy will go into recession. But compared to historical bubbles: UK spent 6% of GDP on railroads for a decade. We’re nowhere close to 6% of our GDP. The strongest balance sheets in the world can all pull the plug at any point.

If I could have an intelligence as smart as a Google senior engineer, that’s $2 trillion of software value. We don’t need to get digital god for there to be immense value. All of these other revolutions have been capital goods that reduce the amount of human capital you need. Whereas this is just creating human capital.

On power and data centers: AI data centers are like 3-4% of US power. We just haven’t built power in like 40 years. We’ve transitioned from coal to natural gas. It’s a supply chain thing, a lack of labor thing. There’s a company putting bunch of truck engines in parallel like diesel truck engines because the capacity for diesel truck engines is huge. Elon buying power equipment from Poland. Electrician wages have doubled for mobile electricians that can work on data center stuff.

On China: They’re a very formidable competitor. If we didn’t have the AI boom, the US probably would be behind China and no longer the world hegemon by the end of the decade. Without AI, we’re definitely just going to lose. Our supply chains are slower. They cost too much. Our debt is unsustainable. AI has to dramatically accelerate GDP growth. Once you start talking about dividing the pie you’re screwed. It has to be growing the pie.

China doesn’t necessarily need AI to win. They’ve always played this long game. They did it with steel, rare earth minerals, solar panels, phones, PCBs. They’ve dumped at least $400-500 billion into the semiconductor ecosystem. If you take any country in isolation, China is the one that has everything at the highest level on average. Sure, they’re 30 years behind on jet engines, but they don’t need to go outside of China for any of the materials besides raw materials.

ByteDance is the third largest user of GPUs in the world. DeepSeek engineers make a lot more than other engineers, but they’re not making $10 million even though they may be worth it. China could build way faster than us. If they wanted to build a 10 gigawatt data center, they could probably build it in a few years. They don’t have the best chips. They don’t have the best memory. They do have the most power. They can build stuff way faster. We’re impressed at how fast Elon does stuff. Elon’s slow compared to China.

Speed round - OpenAI: Super awesome. Anthropic: Actually more optimistic on Anthropic than OpenAI. Their revenue is accelerating way faster because what they’re focused on is more relevant to that $2 trillion software market. AMD: Love them but they’re pretty mid. XAI: In real danger of not being able to raise capital. They’ll have the biggest individual data center. If he doesn’t figure out a business model besides Pornbot (Annie)… Oracle: Going to make so much money if you believe OpenAI is successful. In most worlds where Oracle gets paid $300 billion by OpenAI, OpenAI is like a $5-10 trillion company.

Meta: Got the cards to potentially own it all. The next paradigm of human computer interface is we don’t actually have to touch it at all. We tell the AI what we want. The only company in the world who has the full stack from good hardware (glasses with screen) plus good models plus capacity to serve them plus knowledge around recommendation systems plus the capital.

Google: Was pretty bearish two years ago but super bullish now. They’re waking up on every front. Taking TPUs, selling them externally. Models are competitive. Aggressive on infrastructure investments. They have the hardware business, Android, YouTube, search. Well positioned to capture both consumer and professional markets.

On the future of software: The era of SAS businesses is really tough in the age of AI. AI software development tanks the cost of building competing software. If you are a SAS business, your customer acquisition cost remains the same but now you’ve added a humongous COGS. China doesn’t have much of a SAS business because the cost of developing software was so much less that the SAS business model didn’t work. Software developers in 2015 in China were getting paid maybe a fifth of the US and were maybe twice as good. 10x lower cost of software and so SAS never happened. Many software businesses will have a reckoning with the fact that their COGS is going to soar, their customer acquisition cost isn’t going to fall and they have a lot more competitors and so they don’t hit that escape velocity.

Startups: Periodic Labs. Mostly OpenAI people, a Google guy, couple material scientists. Taking the RL paradigm but doing it with real world chemistry and physics. Test chemistry, feed that feedback back into the model. Physical flywheel is slow compared to digital, but there’s a ton of low hanging fruit.

On hardware startups: Not a big bull on accelerator companies competing with Nvidia. Too hard, too capex intensive, no revolutionary leap. But individual parts of the supply chain which are not space-aged. Transformers haven’t changed in 50-100 years. Solid state transformers. Networking between chips. As we extend context length the memory requirements become bigger. Optics space bridging electrical connectivity and optical connectivity.

Closing: My brother. Everything he’s done in my life. I’ve been an asshole my whole life. Every time he pulls me back on path. He corrects me. He loves me unconditionally. I’m very bad at task orientation, remembering to do specific things. I vibe really hard and I’m in the moment really hard and I digest tons of information, but I’m bad at being considerate of what other people are thinking.