Chinese companies today are not that far

behind in AI models at least perceivable

by the public relative [music] to the

amount of compute they have, right? The

leading Chinese labs have 100-200

megawatts total of compute at [music]

most, whereas Anthropic is, you know,

nearly 5 gigawatts by the end of the

year. You know, the question is sort of,

well, does it matter? And I think

[music] right now it doesn't matter that

much because, you know, when we break

down the compute [music] ratio or budget

of a lab, historically it's been like

60% training, 40% inference, but that

training gets broken down further. And

then it's [music] actually like 50% of

the compute is research, like 10% of the

compute is development, and then 40% is

inference. And what I mean [music] by

research and development is, you know,

researchers are generating ideas,

testing new architectures, testing new

data mixes, [music] but ultimately when

Anthropic trains Methos, it's sub-200

megawatts, right?

>> They didn't pre-train the whole thing.

>> The pre-train. It's sub-200 megawatts

for, call it, 2 months, and then the RL

is even less.

>> But total computer was probably higher,

right?

>> But it's like sequential, right? The

most they ever used at one point in time

was maybe 200 megawatts. And then in

reality, [music] they had multiple

gigawatts, so most of their compute was

going to the research, not the

development or the model.