AJ

Googler

[AJ @_37636045]

@zephyr_z9 @HyperTechInvest Ok, my bad seems I missed it your previous comment was primarily on HBM. Offload not only to NAND, but to dram pool via cxl as well which I have seen very large production deployments

[AJ @_37636045]

@zephyr_z9 @HyperTechInvest Do you know how much memory is used for weights versus kv cache currently? Why does it even matter when parameters are not scaling in the future?

[AJ @_37636045]

@zephyr_z9 @HyperTechInvest hmmm, why would that reduce demand?

[AJ @_37636045]

@deedydas You are using Harker to represent Bay Area? Clickbait! Go find a real job and spend less time on X.

[AJ @_37636045]

@zephyr_z9 It only affects the parameters, KV cache usage not changing much.

[AJ @_37636045]

@leeminceman @munster_gene oh, just give up investment...

[AJ @_37636045]

@Kay2289123 nvda已经被完美定价了？

[AJ @_37636045]

@BenBajarin That is unit price not volume

[AJ @_37636045]

@bigeagle_xd 长者🎂

[AJ @_37636045]

@IvanaSpear Google’s engineer did a presentation about HBF in FMS, so Google must already be in the game. You did great job on optical companies, but you might want to do more research on storage companies before commenting

[AJ @_37636045]

@ShanghaoJin @azarashi_doggo $1500你真是张嘴就来啊。。。

[AJ @_37636045]

@paurooteri How do you know it’s hbf?

[AJ @_37636045]

@fi56622380 @TiezhuCrypto @ShanghaoJin

[AJ @_37636045]

@Rickylu88 @ShanghaoJin 砍了存储gpu只能空闲着 jin这哥们但凡对llm的inference了解一点也不会发这种暴论

[AJ @_37636045]

@zephyr_z9 时间已经过了，没发生

[AJ @_37636045]

@ShanghaoJin 增加的capex又不是用来train model的。。。

[AJ @_37636045]

@pequityresearch what's vague? are you expecting they share the exact number?

[AJ @_37636045]

@freearkshaw source?

[AJ @_37636045]

@freearkshaw Lumentum?

[AJ @_37636045]

@labubu_trader @FundaAI I wish I could play calls of $SNOW