Felipe Paletta

Partner @invistamonett | Equity Research, CNPI

[Gavin Baker @GavinSBaker]

1) DeepSeek r1 is real with important nuances. Most important is the fact that r1 is so much cheaper and more efficient to inference than o1, not from the $6m training figure. r1 costs 93% less to *use* than o1 per each API, can be run locally on a high end work station and does not seem to have hit any rate limits which is wild. Simple math is that every 1b active parameters requires 1 gb of RAM in FP8, so r1 requires 37 gb of RAM. Batching massively lowers costs and more compute increases tokens/second so still advantages to inference in the cloud. Would also note that there are true geopolitical dynamics at play here and I don’t think it is a coincidence that this came out right after “Stargate.” RIP, $500 billion - we hardly even knew you.

Real: 1) It is/was the #1 download in the relevant App Store category. Obviously ahead of ChatGPT; something neither Gemini nor Claude was able to accomplish. 2) It is comparable to o1 from a quality perspective although lags o3. 3) There were real algorithmic breakthroughs that led to it being dramatically more efficient both to train and inference. Training in FP8, MLA and multi-token prediction are significant. 4) It is easy to verify that the r1 training run only cost $6m. While this is literally true, it is also *deeply* misleading. 5) Even their hardware architecture is novel and I will note that they use PCI-Express for scale up.

Nuance: 1) The $6m does not include “costs associated with prior research and ablation experiments on architectures, algorithms and data” per the technical paper. “Other than that Mrs. Lincoln, how was the play?” This means that it is possible to train an r1 quality model with a $6m run *if* a lab has already spent hundreds of millions of dollars on prior research and has access to much larger clusters. Deepseek obviously has way more than 2048 H800s; one of their earlier papers referenced a cluster of 10k A100s. An equivalently smart team can’t just spin up a 2000 GPU cluster and train r1 from scratch with $6m. Roughly 20% of Nvidia’s revenue goes through Singapore. 20% of Nvidia’s GPUs are probably not in Singapore despite their best efforts. 2) There was a lot of distillation - i.e. it is unlikely they could have trained this without unhindered access to GPT-4o and o1. As @altcap pointed out to me yesterday, kinda funny to restrict access to leading edge GPUs and not do anything about China’s ability to distill leading edge American models - obviously defeats the purpose of the export restrictions. Why buy the cow when you can get the milk for free?

[Felipe Paletta @felipepaletta]

a boca quer fechar...

[Felipe Paletta @felipepaletta]

Tesla (TSLA): vale a pena COMPRAR? | Lucro cai 55% e ações sobem 11%

https://t.co/NOdE5LOvXX

[Felipe Paletta @felipepaletta]

O dólar vai subir mais?
Entenda as razões por trás da depreciação do real

https://t.co/THnfqBBcA7

[Monett @invistamonett]

Ao acessar a Monett você já visualiza os conteúdos que são compatíveis com o seu perfil de investidor.

A nova funcionalidade te ajuda a encontrar as melhores recomendações para o seus objetivos e te proporciona uma experiência totalmente personalizada.

Baixe o app e aproveite!

[Bloomberg Markets @markets]

Sen. (D) Elizabeth Warren says about 2 million people will lose their jobs as a result of the Fed’s anticipated monetary tightening.

She asked Fed Chair Powell to explain why these people must lose their jobs as part of the effort to bring down inflation https://t.co/XZjOTm3oVR

[Felipe Paletta @felipepaletta]

Reflexão do dia:

Buffett on Buybacks, by @RogerLowenstein https://t.co/HkQqxROHPc

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: One Shot, One Kill https://t.co/Z0YkeFt88V via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Pavlov, toque o sino mais uma vez! https://t.co/dAFOt8HXGn via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Beco sem saída https://t.co/IytF7ek9nt via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Bora FII do Bill https://t.co/gaVj7e04JI via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Bolsolula: os FIIs pós eleições https://t.co/g12SAUiAcK via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Dividendo parado pega pó https://t.co/kJRnwUwqvn via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: O pior está por vir https://t.co/moV9kZhrKB via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: R$ 10 mil por mês com FIIs https://t.co/LIkadFjajA via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Figurinha repetida não completa álbum https://t.co/FWxXh2m6a7 via @LinkedIn

[Felipe Paletta @felipepaletta]

@luizcesta yields uppp

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Os dividendos cairão, mas você vai ganhar dinheiro https://t.co/MrgXTXOJYK via @LinkedIn

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: A insegurança do lucro e o medo do prejuízo: com renda variável subindo, será hora de vender? https://t.co/GdbRcwaW6T via @LinkedIn

[Felipe Paletta @felipepaletta]

@Celito1961 @LinkedIn https://t.co/LMgjrjDsYa

[Felipe Paletta @felipepaletta]

Confira o artigo mais recente da minha newsletter: Imóvel em Portugal, vale a pena? https://t.co/T87fVvMLUg via @LinkedIn