Nathan Lambert

But like the experience when you're inside of Google and you're training on TPUs as a researcher, you don't need to know anything about the hardware in many cases. Right. Like it's like pretty beautiful. But as soon as you step outside, they all go. A lot of them go back.

But like the experience when you're inside of Google and you're training on TPUs as a researcher, you don't need to know anything about the hardware in many cases. Right. Like it's like pretty beautiful. But as soon as you step outside, they all go. A lot of them go back.

they leave google and then they go back yeah yeah they're like they leave and they start a company because they have all these amazing research ideas and they're like wait infrastructure is hard software is hard and this is on gpus or if they try to use tpus same thing because they don't have access to all this code and so it's like how do you convince a company whose golden goose is search where they're making hundreds of billions of dollars from to start selling gpu or tpus uh which they used to only buy a couple billion of you know i think in 2023 they bought like

they leave google and then they go back yeah yeah they're like they leave and they start a company because they have all these amazing research ideas and they're like wait infrastructure is hard software is hard and this is on gpus or if they try to use tpus same thing because they don't have access to all this code and so it's like how do you convince a company whose golden goose is search where they're making hundreds of billions of dollars from to start selling gpu or tpus uh which they used to only buy a couple billion of you know i think in 2023 they bought like

they leave google and then they go back yeah yeah they're like they leave and they start a company because they have all these amazing research ideas and they're like wait infrastructure is hard software is hard and this is on gpus or if they try to use tpus same thing because they don't have access to all this code and so it's like how do you convince a company whose golden goose is search where they're making hundreds of billions of dollars from to start selling gpu or tpus uh which they used to only buy a couple billion of you know i think in 2023 they bought like

um like a couple billion and now they're buying like 10 billion to 15 billion dollars worth but how do you convince them that they could they should just buy like twice as many and figure out how to sell them and make 30 billion dollars like who cares about making 30 billion dollars won't that 30 billion exceed actually the search profit eventually

um like a couple billion and now they're buying like 10 billion to 15 billion dollars worth but how do you convince them that they could they should just buy like twice as many and figure out how to sell them and make 30 billion dollars like who cares about making 30 billion dollars won't that 30 billion exceed actually the search profit eventually

um like a couple billion and now they're buying like 10 billion to 15 billion dollars worth but how do you convince them that they could they should just buy like twice as many and figure out how to sell them and make 30 billion dollars like who cares about making 30 billion dollars won't that 30 billion exceed actually the search profit eventually

I mean, like, you're always going to make more money on services than... Always. I mean, like, yeah. Like, to be clear, like, today people are spending a lot more on hardware than they are the services, right? Because the hardware front runs the service spend.

I mean, like, you're always going to make more money on services than... Always. I mean, like, yeah. Like, to be clear, like, today people are spending a lot more on hardware than they are the services, right? Because the hardware front runs the service spend.

I mean, like, you're always going to make more money on services than... Always. I mean, like, yeah. Like, to be clear, like, today people are spending a lot more on hardware than they are the services, right? Because the hardware front runs the service spend.

If there's no revenue for AI stuff or not enough revenue, then obviously, like, it's going to blow up, right? You know, people won't continue to spend on GPUs forever. And then NVIDIA is trying to move up the stack with, like, software that they're trying to sell and license and stuff, right? But... Google has never had that DNA of like, this is a product we should sell, right?

If there's no revenue for AI stuff or not enough revenue, then obviously, like, it's going to blow up, right? You know, people won't continue to spend on GPUs forever. And then NVIDIA is trying to move up the stack with, like, software that they're trying to sell and license and stuff, right? But... Google has never had that DNA of like, this is a product we should sell, right?

If there's no revenue for AI stuff or not enough revenue, then obviously, like, it's going to blow up, right? You know, people won't continue to spend on GPUs forever. And then NVIDIA is trying to move up the stack with, like, software that they're trying to sell and license and stuff, right? But... Google has never had that DNA of like, this is a product we should sell, right?

Google Cloud, which is a separate organization from the TPU team, which is a separate organization from the DeepMind team, which is a separate organization from the search team, right? There's a lot of bureaucracy here.

Google Cloud, which is a separate organization from the TPU team, which is a separate organization from the DeepMind team, which is a separate organization from the search team, right? There's a lot of bureaucracy here.

Google Cloud, which is a separate organization from the TPU team, which is a separate organization from the DeepMind team, which is a separate organization from the search team, right? There's a lot of bureaucracy here.

Technically TPU sits under infrastructure, which sits under Google Cloud, but like Google Cloud, like for like renting stuff and TPU architecture are very different goals, right? In hardware and software, like all of this, right? Like the Jaxx XLA teams do not serve Google's customers externally. Whereas NVIDIA's various CUDA teams for like things like Nickel serve external customers, right?

Technically TPU sits under infrastructure, which sits under Google Cloud, but like Google Cloud, like for like renting stuff and TPU architecture are very different goals, right? In hardware and software, like all of this, right? Like the Jaxx XLA teams do not serve Google's customers externally. Whereas NVIDIA's various CUDA teams for like things like Nickel serve external customers, right?

Technically TPU sits under infrastructure, which sits under Google Cloud, but like Google Cloud, like for like renting stuff and TPU architecture are very different goals, right? In hardware and software, like all of this, right? Like the Jaxx XLA teams do not serve Google's customers externally. Whereas NVIDIA's various CUDA teams for like things like Nickel serve external customers, right?