Ep. 263 - Nilesh Shah: From Dirt to Tokens and Everything in Between
カートのアイテムが多すぎます
カートに追加できませんでした。
ウィッシュリストに追加できませんでした。
ほしい物リストの削除に失敗しました。
ポッドキャストのフォローに失敗しました
ポッドキャストのフォロー解除に失敗しました
-
ナレーター:
-
著者:
Nilesh Shah has spent his career following the interesting problems, from chip packaging at Intel to the fast-moving world of neoclouds and AI infrastructure. Recorded at YOTTA in Las Vegas, we get into what makes a neocloud different, why AI has turned liquid cooling into a must-have, how Hosted.AI connects compute with demand, and what it really takes to optimize the stack from dirt to tokens. We also make room for token factories, orbital data centers and a few avocados that never fulfilled their potential.
Chapters:
00:00 From baker jokes to Intel chip packaging
02:05 How memory, storage and silicon led to AI infrastructure
05:14 Following the future and learning to fail fast
07:53 Curiosity, focus and the shiny-object problem
09:55 Family, GPUs and discovering technology for yourself
11:47 The “not a polite roundtable” at YOTTA
13:18 What is a neocloud?
16:11 How AI workloads changed compute requirements
18:18 Why transformer models need interconnected GPUs
19:07 Density, networking and the shift to liquid cooling
23:01 Not all neoclouds are created equal
23:47 Three ways neoclouds deliver AI compute
27:31 Dirt to tokens: the full AI infrastructure stack
31:11 Optimizing tokens, GPUs, racks and power in real time
35:15 Where neoclouds compete and differentiate
40:17 Hardware, software and financing shape the market
42:36 How neocloud contracts and offtake models differ
49:23 How much resiliency does AI really need?
53:04 Can high-density AI ever operate without UPS?
57:28 Why telemetry could make compute bankable
59:07 Agent-to-agent AI, humanoids and orbital compute
01:02:28 The scarce resource may be permission to build
01:04:08 When AI agents run the household
01:06:48 What comes next
Links:
Nilesh Shah on LinkedIn: https://www.linkedin.com/in/nilesh-n-shah
AI: https://hosted.ai/
YOTTA: https://www.yotta-event.com/
QTS Data Centers: https://q.com/
Thank you for visiting our channel, be sure to Like, Comment, and Subscribe!
For more information, go to https://www.Q.com
Learn More about QTS Data Centers on the QTS YouTube Channel: https://www.youtube.com/users/qtsdatacenters
Think you have a great topic for a podcast or would like to schedule an interview with Dave? Send in your requests to podcast@q.com